Pa

Pa

b

a

ARRAA

KBIU

1

fGmca22stntu2oi2

r

m

0h

Research Policy 42 (2013) 101– 116

Contents lists available at SciVerse ScienceDirect

Research Policy

jou rn al h om epage: www.elsev ier .com/ locate / respol

rofiting from business model innovation: Evidence from Pay-As-You-Driveuto insurance

anos Desyllasa,∗, Mari Sakob,1

Manchester Business School, University of Manchester, Booth Street West, Manchester M15 6PB, United KingdomSaid Business School, University of Oxford, Park End Street, Oxford OX1 1HP, United Kingdom

r t i c l e i n f o

rticle history:eceived 22 December 2010eceived in revised form 3 May 2012ccepted 17 May 2012vailable online 25 June 2012

eywords:

a b s t r a c t

The emergent business model literature, revolving mainly around the mechanisms through which newbusiness models create and deliver value, has left the value capture challenge under-explored. This paperexamines how an incumbent firm profits from business model innovation through the study of Pay-As-You-Drive auto insurance. Although business models do not warrant formal intellectual property(IP) protection, their constituent components (e.g. business methods and brands) often do. Drawing onthe profiting-from-innovation framework, we find that formal and strategic IP protection methods play

usiness model innovationntellectual propertysage-based insurance

complementary roles. Initially, formal IP rights are used primarily as a defensive strategy, as vehicles forpackaging and trading know-how, and most importantly as a means of “buying time” to build specialisedcomplementary assets. Long-term competitiveness, however, depends on whether the innovator buildsa strong position in specialised complementary assets and is capable of reconfiguring them over time inline with changes in the market environment. Thus, we explicate the complex mechanism and dynamiccapability for capturing value from business model innovation.

. Introduction

The notion of business model is attracting increasing attentionrom academics and practitioners alike (Baden-Fuller et al., 2010;ovindarajan and Trimble, 2005; Rappa, 2004). Despite disagree-ent among scholars on what a business model is, there is some

onsensus that it describes the design of the value creation, deliverynd capture mechanisms to be employed by the firm (Chesbrough,010; Chesbrough and Rosenbloom, 2002; Teece, 2010; Zott et al.,011). The choice of an appropriate business model is increasinglyeen as a crucial business decision due to the post-industrial rise ofhe knowledge economy and digital technology. As firms developew products and services, they often require a new business modelo realign their systems and processes to support the novel prod-cts or services (Chesbrough and Rosenbloom, 2002; Zott and Amit,010). More generally, a firm’s business model is an important locusf innovation and a crucial source of value creation for the firm andts stakeholders (Amit and Zott, 2001; Chesbrough, 2011; Teece,

006).

Although the emergent business model literature has elabo-ated on the mechanisms for value creation and delivery when new

∗ Corresponding author. Tel.: +44 0161 275 6469.E-mail addresses: panos.desyllas@mbs.ac.uk (P. Desyllas),

ari.sako@sbs.ox.ac.uk (M. Sako).1 Tel.: +44 01865 288925.

048-7333/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.respol.2012.05.008

© 2012 Elsevier B.V. All rights reserved.

business models are conceived and implemented, it has left theissue of value capture relatively under-explored. This is surpris-ing given that the adoption of a novel business model has beenacknowledged as an important element of a firm’s intellectualproperty (Zott et al., 2011). The extant innovation literature hasfurthered our understanding of the means firms have at their dis-posal to protect innovative products or processes. In particular, astream of work that has come to be known as the “profiting frominnovation” (PFI) literature examines this central question for tech-nological innovation (Teece, 1986, 2006; Chesbrough et al., 2006).This body of work suggests that both formal (patents, trademarksand copyrights) and strategic (e.g. secrecy, lead time advantages)means of IP protection exist, and that their relative effectiveness asmeans of value capture depends on the IP appropriability regime(i.e. the efficacy of legal protection and the inherent replicabilityof the innovation). However, no-one has yet examined systemati-cally whether the strategies that are employed for profiting fromtechnological innovation are also effective in the case of businessmodel innovation.

Business model innovation has its own distinctive features(see Baden-Fuller et al., 2010). Most importantly, a new busi-ness model is itself unlikely to qualify for formal IP protection.However, specific business methods underlying it may be pro-

tectable by obtaining formal IP rights. This applies particularly toinnovative business methods reflecting novel applications of Infor-mation and Communication Technology (ICT) (Gambardella andMcGahan, 2010; Salter and Tether, 2006; Tether, 2005). In specific

1 arch P

icsbHdaanp

i(itaWioiWyf

tmttotfoTtasUai

vattbciHfimbtem2ttbcaecMicr

assets can be seen as a special case of strategic means of IP protec-tion, which also include secrecy and lead time advantages (Cohenet al., 2000; Levin et al., 1987).2

2 The relative importance of formal and strategic IP protection methods alsodepends on the evolution stage of the product or service market. At the pre-

02 P. Desyllas, M. Sako / Rese

nstances, an entire business model can be embedded in digitalode (Ovans, 2000). More recently, what constitutes patentableubject matter has been extended to cover novel ways of doingusiness in the USA and some other countries (Blind et al., 2003;all and MacGarvie, 2006; Wagner, 2008). Thus, in these juris-ictions new business methods that underpin a business modelre now patentable. This provides an apt institutional setting forddressing the question: how can incumbent firms profit from busi-ess model innovation with the availability of business methodatenting?

This paper addresses this question through the study of the autonsurance industry, using the prominent case of Pay-As-You-DrivePAYD) auto insurance, which was introduced by the Americannsurer Progressive Corporation (henceforth called Progressive) inhe late 1990s. PAYD is a novel method of determining insur-nce premiums on the basis of when and how a car is driven.e compare the PAYD business model with the conventional auto

nsurance business model, study the evolutionary process of PAYDver a fifteen-year period, and compare the competitive dynam-cs in the US and the UK (without business method patenting).

e utilise multiple research methods, involving archival anal-sis, company interviews, and analysis of Progressive’s IP port-olio.

Our analysis of the PAYD case provides the following contribu-ions to the emergent literature on how firms profit from business

odel innovation. First, Progressive sought formal IP protection forhe novel business methods central to a new business model, ini-ially, not as a means of capturing the resultant value, but as partf a pre-emptive strategy (so not to be inhibited by a competitorhat obtained IP protection). We also find that different forms oformal IP protection complemented rather than of substituted forne another since they protect different aspects of the firm’s IP.hese findings are in line with IP protection strategies related toechnological and service innovation (Cohen et al., 2000; Tethernd Massini, 2007). Our analysis comparing the UK and US marketshows that the strengthening of the IP appropriability regime in theS facilitated Progressive to slow down imitation and new entry,nd maintain some control over the development of usage-basednsurance.

Second, we find that patents in licensing agreements serve asehicles for the packaging and trading of not only codified butlso tacit knowledge by facilitating an active collaboration betweenhe participants (Arora et al., 2001). We suggest that this func-ion of patents is particularly important for the transfer of newusiness models because business model imitation is subject toausal ambiguity and depends on obtaining control over a system ofnterdependent assets and activities (Henderson and Clark, 1990;ughes and Scott Morton, 2006; Zott and Amit, 2010). Third, wend that the complementary assets that are required for the imple-entation of a new business model do not necessarily exist ex ante

ut are built and configured as the innovation process unfolds. Inhe PAYD case, formal IP rights contributed to the initial phase ofxperimentation and exploration, which are inherent in businessodel development (Govindarajan and Trimble, 2005; McGrath,

010; Sosna et al., 2010; Teece, 2010), and bought Progressive timeo build the specialised complementary assets. Thus, we argue thathe requisite specialised complementary assets may be enhancedy the possession of formal IP rights. Finally, we find that the spe-ialised complementary assets that contribute to the success of

particular business model change as the market environmentvolves with the emergence of new competitors, suppliers, andomplementors (Demil and Lecocq, 2010; Jacobides et al., 2006;cGrath, 2010). Hence, the longer-term competitive position of

nnovators depends on their dynamic capability to reconfigure their

omplementary asset base in line with changes in the market envi-onment.

olicy 42 (2013) 101– 116

2. Theoretical framework

Notwithstanding significant disagreements on what a businessmodel is (Zott et al., 2011), this study adopts a well-formulated def-inition that sees a business model as a description of a company’slogic of value creation, delivery and capture (Amit and Zott, 2001;Chesbrough, 2010; Ghaziani and Ventresca, 2005; Teece, 2010).Specifically, a business model articulates the value proposition ofthe firm; identifies a market segment and specifies the revenuegeneration mechanism; defines the structure of the value chainrequired to create and distribute the offering and the complemen-tary assets needed; details the revenue mechanism by which thefirm will be paid for the offering; estimates the cost structure andprofit potential; describes the position of the firm within the valuenetwork linking suppliers, customers and complementors; and for-mulates the competitive strategy by which the innovating firm willgain and retain advantage over rivals (Chesbrough, 2010).

The early extant literature emphasised the importance ofcoupling technological innovation with the development of anappropriate business model (Chesbrough and Rosenbloom, 2002;Johnson et al., 2008). More recently, new business models areconsidered central in their own right for both product and ser-vice firms to ensure business growth (Chesbrough, 2011). Eitherway, committing resources to business model innovation is acomplex investment decision, particularly for established firms(Govindarajan and Trimble, 2005; Sosna et al., 2010). This is becausethe total opportunity cost that is involved in the adoption of a novelbusiness model includes not only the development or acquisitionof new assets, but also, for incumbents, the possible cannibalisationof a firm’s existing businesses and the obsolescence of its core com-petencies (Chesbrough, 2010; Christensen, 1997; Leonard-Barton,1992). Other barriers to business model innovation are cognitiveand organisational inertia, and internal resistance (Sosna et al.,2010; Teece, 1980; Zott et al., 2011). But ultimately, a major deter-minant of such an important investment decision is the firm’sability to profit from the new business model by protecting it fromimitation.

The PFI literature offers a useful theoretical framework to anal-yse the issue of how pioneering firms profit from being first toinnovate (Teece, 1986, 2006). The PFI framework identifies twomechanisms for protecting the innovating firm’s core knowledge.The first consists of formal means of intellectual property protec-tion, such as patents, trademarks and copyrights. However, in somecases, as in business model innovation, firms may not possess IPrights or, even when they do, the appropriability regime may beweak. This is the case when the core knowledge involved cannot beeasily codified, IP rights are not easily enforceable, imitation is easy,scale is not important, or the switching costs between technologiesare low. Then, a second protection line, namely specialised com-plementary assets, becomes vital for successful appropriability.These firm-specific assets complement the innovation throughoutthe firm’s value chain, and include production capabilities, distri-bution networks, brand names, and after-sale services which areneeded for the commercialisation of the innovative idea. What dis-tinguishes these assets from generic complementary assets is thatthey are not available in competitive supply and they are subject tounilateral or bilateral dependence with the core innovation. These

paradigmatic phase – where multiple product designs are competing for marketdominance – innovation in product characteristics and the possession of formal IP

arch P

iittvftibiig2btbe(S

tTiobbpecePDadCoaota

ntPoTsbTnt

3

AiPa

rd–s

policyholders for incurred losses and the underwriting expenses;and (2) by investing the premiums an insurer collects in commonequities and fixed-income securities. As Table 1 illustrates, in 2003,

3 Our analysis of formal IP rights focuses on patents and trademarks. Progressivealso developed proprietary software as part of the PAYD offering and employedexclusivity agreements with the suppliers of the monitoring device but the companywas not willing to disclose further details.

4 The patent application date is the time when the information about an inventionis first released to the public domain. A possible problem with application dates is

P. Desyllas, M. Sako / Rese

The lens of the PFI framework can improve our understand-ng of different strategies for capturing value from business modelnnovation (Teece, 2010). Extension to business model innova-ion may even enrich and generalise the PFI framework. However,he PFI framework, originally developed with technological inno-ation in mind, should be applied with caution. The traditionalormal and strategic IP protection methods may not be as effec-ive with business model innovation as they are with technologicalnnovation. In relation to strategic means of IP protection, it maye hard to exploit advantages based on secrecy and speed. For

nstance, many new business models represent ICT-implementednnovations which reflect novel applications of existing technolo-ies and a deep understanding of customer needs (Amit and Zott,001; Johnson et al., 2008; Rappa, 2004). The transparency of theuilding blocks of business models can make them susceptibleo imitation by competitors (Teece, 2010). Nevertheless, the rightusiness model may not be apparent upfront, and has to be discov-red through an iterative trial-and-error and idea-testing processBirkinshaw et al., 2008; Doganova and Eyquem-Renault, 2009;osna et al., 2010).

Therefore, the key issue becomes what role formal IP protec-ion may serve in profiting from innovation in business models.his is a particularly topical issue in the light of the recent changesn what constitutes patentable subject matter in the US and a fewther countries (Japan, Australia and Singapore). In fact, a number ofusiness models have become closely intertwined with particularusiness methods protected by IP rights (e.g. Amazon’s “one-click”atent is seen as instrumental to its business model). The gen-ral view that business methods should not be patentable hashanged since the mid-1990s, a time of rapid ICT advances and the-business model boom. In 1997, the USPTO created Class 705 “Datarocessing: Financial Business Practice, management, Or Cost/Priceetermination”. This class includes novel practices in managingnd administering a business, finance practices and cost/priceetermination practices. A year later, in 1998, the United Statesourt of Appeals for the Federal Circuit reinforced the patentabilityf business methods with its decision about the State Street Banknd Trust Co. vs Signature Financial Group case. After a long seriesf legal iterations, in June 2010, the US Supreme Court clarifiedhat a claimed method may be patentable as long as it is not just anbstract idea.

Despite scepticism about the quality and enforceability of busi-ess method patents (Dreyfuss, 2000; Hall, 2009; Lerner, 2002),heir availability has created a new IP appropriability regime. As theFI framework predicts, a firm’s competitive strategy is conditionaln the IP appropriability regime (Pisano, 2006; Teece, 1986, 2006).he enhanced toolkit of formal IP protection methods warrants atudy of whether or not it has indeed strengthened the ability ofusiness model innovators to appropriate returns from innovation.hus, we explore the ways in which firms can profit from new busi-ess models in the contemporary IP appropriability regime throughhe study of the PAYD case.

. Case and research methods

Pay-As-You-Drive auto insurance was first introduced in themerican market by Progressive (one of the top five US auto

nsurers) in the late 1990s. We have tracked the development ofAYD during the period from 1996 to 2010. A case study design isppropriate as this is a relatively new area of research which has

ights are the competitive imperatives. At the paradigmatic phase – where a stan-ard or dominant design has emerged and the technology moves towards maturitythe terms of competition change to cost efficiency through process innovation and

cale, and downstream value chain activities become crucial.

olicy 42 (2013) 101– 116 103

only recently received attention in the management literature (seeBaden-Fuller et al., 2010). In such cases, new causal linkages maybest be uncovered through an iteration of inductive and deductivetheorising (Eisenhardt, 1989; Yin, 2009).

The analysis is based on multiple sources of data and com-bines qualitative and quantitative research methods to strengthentheory grounding by triangulation of evidence. The qualitativemethods involve archival analysis and sixteen interviews withmanagers and patent attorneys from Progressive, industry experts,and managers from auto insurance competitors and complemen-tors. The latter group includes managers from the insurance firmAVIVA, the licensor of PAYD for Europe and Canada, which offered“Pay-As-You-Drive” in the UK and Canada during the mid-2000s;IBM, AVIVA’s technology partner in the design of the monitoringdevice of PAYD; the Cooperative Insurance Group, which launched“Pay-How-You-Drive” in the UK in 2011; and Octo Telematics, anItalian-based global market leader in insurance telematics services(see Appendix A). The in-depth analysis allows us to understand thesalient features of innovative business methods and how patents,trademarks and complementary assets contribute to capturingvalue from business model innovation (if at all).3 The evidence thatis offered by the interviewees is cross-examined with the assess-ment of stock market investors, using an event study of the stockmarket’s reaction surrounding key dates in obtaining protection ofPAYD using business method patents and registered trademarks.Similar methods have been employed by earlier studies that assessthe private value of IP rights (e.g. Austin, 1993; Hall and MacGarvie,2006; McGahan and Silverman, 2006). As McGahan and Silvermanargue, by examining a firm’s market value, one relies on the incen-tives of investors to change their forecasts of the present values ofprofits in response to new information. In order to estimate thefull value captured from IP rights, we consider the incrementalvalue impounded at both the application and issuance dates forpatents, and the application and publication for opposition datesfor trademarks (see Appendix B).4

4. The auto insurance market

Insurance involves pooling funds from a large number of pol-icyholders to pay for the losses that some of them incur. Autoinsurers write insurance for private passenger or commercial vehi-cles. They use actuarial science to quantify and select the risks toinsure and decide the premium to charge. The insurance policies aresold either directly to the customer or indirectly through the use ofagents or brokers. The insurers’ profit comes from two sources: (1)through the underwriting process, which is the difference betweenthe earned premiums and the payments made to or on behalf of its

that they are not necessarily publicised and the application outcome is uncertain.The patent issuance date is the time when the information about an invention isformally publicised. A possible problem with issuance dates is that they are partiallyanticipated and the relevant information may be already known. For trademarks, wefocus on the date of “publication for opposition” rather than the date of issuance.This is because trademarks rarely raise opposition by other parties once the USPTOattorneys approve the mark for publication and the trademark award is subject tofewer hazards compared with patents. Finally, there are practical constraints thatmake the estimation of the financial impact of events relating to the building ofcomplementary assets impossible. Most complementary assets are built over timeand there are no discrete dates to be considered in an event study.

104 P. Desyllas, M. Sako / Research P

Table 1Auto insurance profitability in 2003.

US auto insuranceindustry

Progressive CasualtyInsurance Group

Expense ratio 25.2 19.9Loss ratio 69.1 67.4Combined ratio 94.6 87.3Investment income ratio 3.9 8.7Operating ratio 90.7 78.6

Notes: Authors calculations using data from Graham and Xie (2007) and Progres-sive’s 2003 Annual Report. Expense ratio = underwriting expenses over premiumswritten; loss ratio = loss and loss adjustment expenses incurred over premiumsera

fotityp

rsaiaciaepaldaattpesfraatw

cihluecd

5

5

tv

arned; combined ratio = loss ratio plus expense ratio plus policyholder dividendatio; investment income ratio = investment income over premiums earned; oper-ting ratio = combined ratio minus the investment ratio.

or every $100 the industry collected in premiums, an insurer paidut $69 in claims and spent $25 on underwriting expenses. Thus,he industry profits $5 for every $100 in premiums. But it also prof-ted another $4 by investing the available financial reserves, leadingo an operating margin of 9%. For Progressive, 2003 was a very goodear leading to better-than-average profitability due to both betterrofit and investment returns.

The US auto insurance industry is characterised by intenseivalry, resulting in significant pressures on premium rates and ahift in market shares towards the more efficient insurers (Grahamnd Xie, 2007; Webb and Pettigrew, 1999). There are three waysn which insurers can improve performance. The most consider-ble one is to reduce the claims costs (67% of premiums). Theseosts are defined by loss severity and frequency. Although cost sav-ngs can be achieved by building a superior underwriting capabilitynd hence superior assessment and pricing of risks, the scope forfficiency improvement is bound by two fundamental insuranceroblems: the adverse selection (Akerlof, 1970) and moral haz-rd (Arrow, 1963). Because drivers differ, “poor” drivers are moreikely to have an accident. Insurers try to identify and avoid insuringrivers that are perceived to be high risk drivers. Also both “good”nd “poor” drivers that become insulated from risk through insur-nce will be less concerned about the negative consequences ofhe risk than they otherwise might be. To address these informa-ion asymmetries between insurers and policyholders, insuranceroviders determine premium rates on the basis of their claimsxperience related to driver demographic characteristics and theafety record of the make and model of vehicle being driven. Theseactors include, driver age, sex, occupation, home address, priorecord, as well as vehicle crashworthiness, safety features (e.g.irbags, anti-lock brakes), popularity with thieves, cost to repair,nd age. However, these factors are imperfectly correlated withhe likelihood of having an accident, which is one of the reasonshy profitability varies among insurers and with time.

The other two sources for higher profits are improving the effi-iency of the underwriting process or improving the return onnvestment. The former source can be the result of savings in over-eads and underwriting expenses, including better policy handling,

ower commissions to agents and customer service costs, etc. Vol-me in business is required in order to spread fixed costs and reapconomies of scale (Cohen and Klepper, 1996). The latter sourcean be achieved by advanced investment strategies but it is heavilyependent on stock market cycles.

. A new business model for PAYD auto insurance

.1. Pre-paradigmatic phase

In 1994 Robert McMillan, an executive at Progressive, conceivedhe idea of PAYD insurance based on observations of OneStar ser-ice that used Global Positioning System (GPS) to track vehicles for

olicy 42 (2013) 101– 116

theft prevention, emergency response and navigation. He beganformulating an innovative method for calculating insurance pre-miums using data on a policyholder’s actual driving behaviour toenrich the traditional underwriting practices. In 1998 Progressivebegan a limited market test of PAYD in Houston-Texas, and a yearlater it formally launched a pilot version of PAYD throughout Texasunder the name of “Autograph”. Autograph relied on GPS satellites,mapping technology, a GPS receiver and a small onboard computerto determine when and how much a vehicle was driven. For thedesign and manufacturing of the monitoring device and the provi-sion of IT services, Progressive allied with ATX (a young telematicsTexan company). Data were recorded at 6-min intervals and wereuploaded to Progressive’s computers on a monthly basis via a callto the on-board modem. Premiums were calculated by combin-ing data on location, speed, mileage and time of day where thedriving occurred (75% weight) with traditional underwriting con-siderations (25% weight), such as driver age, sex and marital status.The approximately 1100 policyholders that participated reportedsavings of 25% on their premiums.

PAYD has several advantages compared with conventionalinsurance. According to Raymond Ling (Associate General Coun-sel at Progressive), PAYD is a method with a technical component,in the sense that it involves a fair amount of technical expertiseto enable the identification and the collection of the “right” kindof data, the reliable recording of information from the vehicle, andthe storage of the data by Progressive. Terry Buechner (Global P&CInsurance Segment Manager, IBM US) emphasised that, comparedwith the “static” auto insurance products that are offered on a six-month or one-year basis, PAYD changes the timeline of insuranceproducts to be “just in time” in the store, that is on a minute-by-minute or mile-by-mile basis. The use of more refined data leadsto an improved method for determining the cost of insurance thatreduces rating errors that arise from the grouping of drivers intogeneral actuarial classes. Thus, PAYD deals with the problems ofadverse selection and moral hazard that arise from the informationasymmetry between the insurer and the policyholders.

Since its birth, PAYD has passed through a long trial-and-errorprocess and several evolutionary phases until it reached its cur-rent form. The pilot testing of Autograph uncovered two problems.First, the implementation of Autograph involved high costs. Auto-graph cost Progressive $500 per vehicle, including the installationof GPS, plus cell phone and onboard computer hardware. To remaincompetitive, Progressive charged its customers only a $65 start-upfee and $5 monthly charge for the usage of GPS, cell phone andcomputer. Second, motorists saw the constant monitoring of theirvehicle as an invasion of privacy. As a result, Progressive had torethink its offering. In April 2000, Progressive stopped selling newAutograph policies and since July 2001 it also stopped servicing allAutograph policies.

After working on a re-configuration of Autograph that wouldaddress its drawbacks, Progressive introduced TripSense in 2003.This version exploited Internet communication to empower thepolicyholder to decide whether she wishes to share with Progres-sive the data record of her driving behaviour. Policyholders receivedfree of charge by post the Tripsensor which needed to be pluggedin the on-board diagnostics communications port of the car. TheTripSense setup cost was about $100 and avoided Autograph’sdata transmission costs. The Tripsensor was a simpler and cheaperdevice compared with its predecessor and recorded miles driven,times of the day in operation, and speed, but not the vehicle’s loca-tion. Then, policyholders needed to download from Progressive’swebsite the software that interfaces with the Tripsensor. Policy-

holders could plug in the devise to their PC’s USB port, upload andsend the data to Progressive. Progressive used these data to cal-culate the premium discount relative to the premium based ontraditional underwriting inputs, which acted as a baseline. The

arch P

dwu

5

arlcwwitssetssis

Smwoimgtsumttmi1Pwtect

5

oafcfvrTw

edew

P. Desyllas, M. Sako / Rese

river could have access to vehicle diagnostics, draw comparisonsith comparable cases and experiment with hypothetical vehiclesage scenarios.

.2. Paradigmatic phase

Despite the progress made with Tripsense, the rapid technologydvances that took place in the meantime resulted in significanteductions in the cost of PAYD’s technologies. Progressive estab-ished a new alliance with Xirgo (a young telematics Californianompany) and developed a new PAYD version called MyRate thatas formally launched in 2009. Like Autograph, MyRate used aireless device for data transmission to Progressive without the

ntervention of the driver, but it made use of much more advancedechnologies and hence achieved a much lower cost. Fifteen yearsince the initial conception of the underlying idea, Progressivetarted to expand further afield covering a total of 19 states. How-ver, MyRate did not deal with the perceived invasiveness dueo the real-time vehicle monitoring. Moreover, the rollout to newtates was problematic. Auto insurance is regulated on a state-by-tate basis and Progressive would have to seek approval from thensurance commissions of every state. Problems emerged in statesuch as California where GPS tracking is illegal.

Eventually, in August 2010, Progressive replaced MyRate withnapshot. Snapshot is similar to MyRate, but instead of the ongoingonitoring of a vehicle, policyholders receive a premium discounthich is based on a 30-day monitoring period. The data recorded

ver this limited period of time are used as a guide to the driv-ng behaviour of a particular driver. After the termination of the

onitoring period the policyholders also return the device to Pro-ressive. This version of PAYD has the advantage that it addresseshe three drawbacks of its predecessors: the cost issue is addressedince the monitoring device is returned to Progressive and can besed by other policyholders; the privacy issue is addressed since theonitoring device is not permanently attached to the car (although

his may incentivise drivers to fake their driving behaviour)5; andhe state regulation issue is addressed since the limited window of

onitoring does not violate state laws. In 2011, Snapshot is offeredn 30 American states but it still represents a small part (just over00,000) of Progressive’s 7.8 million policies. Nevertheless, so farrogressive has not marketed its PAYD offering aggressively. Theyere careful at the experimentation so that they do not jeopardise

heir conventional insurance business (Richard Hutchinson, Gen-ral Manager of Usage Based Insurance at Progressive). In 2012 theompany is expected to roll Snapshot out across all the 50 states inhe country.

.3. PAYD as a new business model

Progressive built a new business model around the offeringf PAYD auto insurance. The new business model offers a novelpproach along several key building blocks. First, it represents a dif-erent value proposition to drivers. PAYD empowers the drivers toontrol their “consumption” by transforming the cost of insurancerom a fixed to a variable cost (David Robson, Partner, Financial Ser-

ices and Insurance, IBM UK). The transparent and simple pricingules offer customers a sense of fairness for the service they receive.he technology involved in PAYD allows the insurer to engageith policyholders and enhances the sense of security through the

5 This problem could be relatively easily solved if the monitoring period isxtended and some degree of randomization in the monitoring is introduced so thatrivers do not know exactly when they are monitored. It should also be acknowl-dged that a portion of one’s driving behaviour is “inelastic” (e.g. commuting toork, driving to school, etc.).

olicy 42 (2013) 101– 116 105

bundling of complementary services. These services include directcommunication with a help-centre in the case of an emergencyusing an emergency/panic button, automated accident notification,roadside assistance and theft protection utilising GPS technology.

Second, the characteristics of PAYD insurance enable Progres-sive to tap into underexploited market segments. The ability toprice-discriminate depending on driving patterns attracts driversthat are underserved and penalised with conventional insurance,such as good and low-mileage drivers (Hutchinson; Todd Litman,Victoria Transport Policy Institute). Potential policyholders includeall the drivers that could not afford to own a car because of highinsurance fees, and “occasional drivers” in large cities that are wellserved by public transportation. The fact that PAYD is based onactual driving behaviour attracts cautious drivers that belong tohigh-premiums groups, i.e. drivers that are young; with criminalconvictions; and with low credit score ratings. The fact that PAYDdoes not require that the identity of the driver is known attractscar rental/leasing companies owning vehicles driven by multipledrivers; car sellers offering test drives; and companies owning afleet of vehicles driven by their employees (e.g. cabs or trucks).In addition, the complementary services that can be offered withPAYD attract parents of young drivers who obtain tools for moni-toring them; and adults with elderly parents whose driving skillsrequire precautionary assessment.

Third, the implementation of PAYD challenges Progressive’sconventional value chain and value network. PAYD allows Progres-sive to improve the efficiency of several operations along its valuechain e.g. the data collection process, the underwriting efficiencyand actuarial accuracy, the claims handling and loss settlement. Atthe same time, some operations, such as IT software and hardwaresystems, become considerably more important compared withconventional insurance operations. Another surprising change for atraditional insurance firm is that Progressive finds itself involved instock control and warehousing for the monitoring devices. Finally,since most PAYD policyholders are Internet literate, insurance poli-cies can be sold directly to consumers and serviced online, reducingthe underwriting expenses. In relation to Progressive’s value net-work, PAYD requires the elimination of old and the introduction ofnew complementors. On the one hand, the Internet-based distribu-tion channel enables Progressive to eliminate the agents and bro-kers. On the other hand, the technology inherent in PAYD requiresinputs and capabilities that fall outside Progressive’s core compe-tences, such as the manufacturing, shipment and installation of themonitoring device, and advanced IT and communication services.Recognising these challenges, Progressive established long-termexclusive agreements with the insurance telematics experts, ATXand Xirgo. Progressive has had exclusivity with the suppliers ofthe monitoring device so that they could not provide a device of asimilar nature to another entity for insurance purposes (Ling).

Fourth, Progressive has had to rethink the sources of its revenueand cost structure. On the revenue side, PAYD can solve the adverseselection problem and allow for almost perfect price discriminationon the basis of actual driving behaviour. In each actuarial class, anumber of drivers might prefer to switch to PAYD instead of payingthe standard insurance premiums that are averaged across all theiractuarial class peers (e.g. young male drivers who drive cautiouslywould be penalised). More generally, as more and more better-than-average drivers are switching to PAYD insurance, the averagepremiums for their former peers that are left behind will rise, hencetriggering a vicious circle for conventional insurance offerings. Fur-thermore, although the self-selection of “good” drivers can havea negative impact on total revenue, this can be outweighed by a

higher business volume achieved by accessing new market seg-ments and higher policy renewal rates.

On the cost side, PAYD can improve the actuarial accuracy of theunderwriting process by basing calculations on individual driving

1 arch P

brslociO2sdwiPatfCytftu

itdEoaTiiokcpiAr

PainacplMOc(macotocc

1c

06 P. Desyllas, M. Sako / Rese

ehaviour. In addition, PAYD has the potential to bring about mate-ial cost savings in claims compared with conventional insurance ineveral ways. First, it offers incentives to drive less and more safely,eading to lower incidence frequency and less severe damage. Sec-nd, it can eliminate the costs arising from fraudulent or impreciselaims and legal disputes with third parties. Third, using the mon-toring device, Progressive can track and recover stolen vehicles.verall, PAYD is estimated to reduce the loss ratio by more than0% (Harbage, 2011). However, PAYD was (and to a lesser extenttill is) suffering from a major handicap: the cost of the monitoringevice. Because, most vehicles do not come off the assembly lineith the necessary technology built in, the monitoring device cost

ncreases underwriting expenses. In this sense, one can argue thatAYD was developed before its time. That said, IT hardware costsre now about a third of those in the early 2000s, telecommunica-ion carrier costs have fallen by more than 80% and IT software costsell as well (Trautsch). According to Graham Jackson (IBM Insuranceonsulting Business), “as the cost of the infrastructure turns down,ou move more to the left of the average premium matrix becausehe percentage savings in expected claim costs becomes more relevantor lower premium policies when you get lower cost of the infrastruc-ure”. Further, as the numbers of PAYD policyholders increase, thenit cost that the insurers pay to telematics suppliers fall.

Finally, Progressive has to reconsider its sources of profit. Dur-ng the initial period of PAYD, Progressive relied more on profithrough underwriting than premium investment due to goodriver self-selection and hence fewer resources for investment.ssentially, Progressive exploited PAYD’s superior risk pricing byvercoming the “omitted variable bias” of conventional insurancend the advantage of attracting better risks. As can be seen fromable 2, that elaborates on the conditions under which usage-basednsurance results in higher profits compared with conventionalnsurance, the overall profit potential of PAYD is determined bypportunities for gaining business from rivals, accessing new mar-et segments, savings on underwriting and loss expenses, andross-selling. At the same time, Progressive’s patent and trademarkortfolio has generated a new source of income through licens-

ng agreements. In 2002 it established licensing agreements withVIVA for the right to launch PAYD in Europe and Canada, and moreecently granted a license to Allstate in the US.

Interestingly, the new business model was embraced throughrogressive’s existing organisational structure and developedlongside its conventional auto insurance business. In general,t is argued that the relationship between an emerging busi-ess model project and the umbrella organisation should strike

balance between the need for experimentation with differentonfigurations of business functions, management teams, com-etencies, performance metrics and the potential to borrow and

everage resources from the parent organisation (Koen et al., 2010;arkides and Charitou, 2004; Govindarajan and Trimble, 2005).ur analysis suggests that initially, Progressive’s executives wereoncerned with how PAYD would fit their overall business modelBuechner). However, it was soon realised that the two business

odels did not involve strategically incompatible or conflictingssets. PAYD could capitalise on Progressive’s existing competen-ies. Historically, Progressive has been known for its capability tout-segment the competition by employing advanced statisticalechniques to analyse driving behaviour, adopting an enhanced set

f underwriting variables to better model risks, and using sophisti-ated underwriting software programs.6 In terms of distributionhannels, Progressive was one of the first auto insurers to sell

6 Progressive dominated the non-standard, high risk driver segment between the960s and 1990s by being able to better assess risks and set rates at finer levels thanompetitors.

olicy 42 (2013) 101– 116

policies and accept payments in real time over the Internet in the1990s. Further, mutual synergies between the two offerings wereexpected to emerge. In particular, the driving data sourced by themonitoring device can be used to refine the actuarial classes thatare employed in conventional insurance (Jackson). Finally, Progres-sive’s organisational culture was conducive to experimentation andinnovation as documented by its innovation record. Overall, thePAYD insurance was seen more as an opportunity to grow the over-all business rather than as a threat to the conventional insurancebusiness model.

6. Profiting from PAYD

Following Gillan et al. (2000), we begin by offering a very crudeassessment of the contribution of PAYD insurance to Progressive’smarket value. If we invested $100 in Progressive’s stock on January1, 1996 (close to the time of Progressive’s first patent applicationfor PAYD), this investment would worth $587 (an annual rate ofapproximately 15%) by 30 October 2010 (see Fig. 1). Alternatively,if we invested $100 in the S&P 500 index, it would be worth $241(an annual rate of approximately 7%), or $160 (an annual rate ofapproximately 4%) had the investment been made in the S&P 500Insurance Index. Although, Progressive has clearly outperformedboth the market and its industry (also see Table 1), only a frac-tion of this superior performance can be attributed to PAYD. Inparticular, we are interested in assessing the effectiveness of Pro-gressive’s formal IP rights and its specialised complementary assetsas mechanisms for value appropriation from the PAYD innovation.

6.1. Business method patents

Progressive realised early on the commercial potential of PAYDand sought to obtain formal IP protection (see Table 3 for a chronol-ogy). This decision is justified by Ling: “We have been working onthis for over 15 years, we sunk a lot of money for developing and,on the innovation side, we filed for patent protection to protect that”.On 29 January 1996, a short time before Progressive introduceda limited market test, it submitted its first PAYD-related patentapplication entitled “Motor vehicle monitoring system for deter-mining a cost of insurance” (US5797134), which was issued on 18August 1998. The patent is uniquely classified in US Class 705/4Insurance. Although the marketed versions of PAYD relied on a lim-ited number of variables (speed, mileage, location, time), the patentclaims reserved the right to the company to monitor and recordmany other aspects of a driver’s behaviour and the state of thevehicle (e.g. radio volume, braking patterns, airbag deployment).Progressive strengthened its IP position by filing for three otherpatents that are linked to patent US5797134 and form a family ofpatents: issued patents US6064970 (continuation of US5797134)and US6868386 (continuation-in-part of US6064970), and thepending patent application US20040153362 (continuation-in-partof US6868386).7 Progressive’s patenting strategy has followed theevolution of its inventive activity, while the continuation patentsenable claiming priority back to the initial patent application.Progressive strengthened its position on the interface betweenthe policyholders and the insurer by filing the issued patentUS7124088 (citing US5797134) and the pending patent applicationUS20070038488 (continuation of US7124088) in order to protect

an online insurance policy service system using the Internet thatfacilitates real-time automated calculation, adjustment and com-munication of premiums.

7 These continuation patent applications cover related inventions, while theirancestor patent application was still pending.

P. Desyllas, M. Sako / Research Policy 42 (2013) 101– 116 107

0

100

200

300

400

500

600

700

800

900

01

/96

05

/96

09

/96

01

/97

05

/97

09

/97

01

/98

05

/98

09

/98

01

/99

05

/99

09

/99

01

/00

05

/00

09

/00

01

/01

05

/01

09

/01

01

/02

05

/02

09

/02

01

/03

05

/03

09

/03

01

/04

05

/04

09

/04

01

/05

05

/05

09

/05

01

/06

05

/06

09

/06

01

/07

05

/07

09

/07

01

/08

05

/08

09

/08

01

/09

05

/09

09

/09

Co

mp

ou

nd

valu

e o

f $10

0

Prog ress ive

Corpo ratio n

S&P 50 0Inde x

S&P 50 0

Insuran ceInde x

Fig. 1. Value of $100 invested in Progressive Corporation, S&P 500 Composite Index and S&P 500 Insurance Index on 01/01/1996.

Table 2Business models compared: conventional vs usage-based auto insurance.

Conventional insurance Usage-based insurance

Customer value proposition Underwriting to cover for accidents and otherunforeseen losses.

Improved underwriting capability which reduces adverse selectionand moral hazard, resulting in lower claims and/or lowerpremiums.Profit improves if claims reduction is greater than premiumreduction.

Market segments Markets segmented by categorical drivercharacteristics (gender, age, occupation, homeaddress, prior record) and vehiclecharacteristics (e.g. age, airbags, cost of repair).

Customers identified by individualised driver characteristics.Profit improves if new markets are served, for safe drivers (lowand infrequent claims), high risk drivers (high premiums) anddrivers attracted by complementary services.

Value chain Information collection mainly undertakenthrough standardised questionnaires answeredby policyholders at time of application.

ICT enables automation of information collection process anddeeper level of data.Profit improves if savings from actuarial accuracy and lowerloss expenses outweigh device and technology costs.

Insurance policies sold and serviced viaagents/brokers or directly to consumers.

Insurance policies sold, modified and renewed through theInternet (i.e. disintermediation).Profit improves due to process automation and reducedcommission to agents and brokers.

Positioning in the value network Value network consists of suppliers (actuaries,technology), intermediaries (agents/brokersand web-based aggregators most recently) andcustomers. State regulators are anotherpowerful stakeholder.

Telematics suppliers and telecommunication operators are themain new complementors. Bundling PAYD with other services (e.g.theft prevention, emergency response) results in other suppliersand complementors.Profit improves if the insurer can provide a platform in relationto current and future complementors.

Revenue generationand cost structure

Revenue consists of premium charged, basedon actuarial calculation of insurance risks forinsured customer groups.

Revenue consists of premiums calculated based on each driver’sbehaviour.Profit improves due to a move towards perfect pricediscrimination capturing greater consumer surplus.

Costs consist of underwriting expenses andoverheads, claims paid out, and commission toagents and brokers. Insurers try to select andmaintain good drivers (e.g. no-claimdiscounts).

Costs consist of lower underwriting and loss expenses (due togreater monitoring leading to less moral hazard), balanced againsthigher costs of monitoring device and digital technology (expectedto fall when technology becomes in-built).Profit improves with scale (lower monitoring device andtechnology unit cost) and with monitoring leading to betterdriving behaviour.

Competitive strategy Insurers try to obtain volume in business tomaximise revenues and benefit from scaleeconomies and excel in underwritingcapability to reduce loss expenses whilecontrolling for information collection costs.

A PAYD insurer can be seen as a modern, innovative and “green”company offering a fair, personalised and “flexible” insurance,appealing to “good” and high-risk drivers that are unfairly treatedwith conventional insurance. PAYD can be developed into aplatform for value-added services.Profit improves through gaining customers from conventionalinsurance, accessing new market segments, savings onunderwriting and loss expenses and cross-selling.

108 P. Desyllas, M. Sako / Research Policy 42 (2013) 101– 116

Table 3Chronology of Progressive’s IP rights for PAYD auto insurance.

Date Event description Abnormalreturn

Cumulativeabnormal return

29/01/1996 US5797134 filing: “Motor vehicle monitoring system for determining a cost of insurance”.Class: 705/4; Intl. Class: G06F 17/60.

−0.6%(−0.52)

0.5%(0.24)

17/01/1997 WO9727561 filling: “Motor vehicle monitoring system for determining a cost ofinsurance”. Intl. Class: G06F 165/00. Priority 29/01/1996, US5797134. Designated states:Japan, Korea, Mexico, EU states.

0.4%(0.25)

2.8%(1.00)

12/01/1998 TM75416261, TM75416262 filling: “AUTOGRAPH”. −0.2%(−0.14)

0.2%(−0.08)

11/02/1998 CA2229238 filing: “Motor vehicle monitoring system for determining a cost of insurance”.Intl. Class: G06F 17/60.

−0.1%(−0.04)

0.6%(0.22)

22/04/1998 CA2235566 filing: “Motor vehicle monitoring system for determining a cost of insurance”.Intl. Class: G06F 17/60. Priority 11/02/1998, CA2229238.

−0.6%(−0.39)

−0.8%(−0.30)

17/07/1998 BR9802520 filing: “Process and system for acquiring and recording data from theoperation of a vehicle”. Intl. Class: G05D27/00. Priority 29/01/1996, US5797134.

−0.6%(−0.39)

−3.4%(−1.30)

28/07/1998 MX199806067 filing: “Motor vehicle monitoring system for determining a cost ofinsurance”. Intl. Class: G06F 165/00. Priority 29/01/1996, US5797134.

1.2%(0.75)

−6.3%*

(−2.34)17/08/1998 US6064970 filing: “Motor vehicle monitoring system for determining a cost of insurance”.

Class: 705/4; Intl. Class: G06F 17/60. Continuation application of US5797134.−1.1%(−0.73)

−2.2%(−0.81)

18/08/1998 US5797134 issuance: “Motor vehicle monitoring system for determining a cost ofinsurance”. Class: 705/4; Intl. Class: G06F 17/60.

−1.1%(−0.74)

−2.2%(−0.84)

29/12/1998 TM75416261 publication for opposition: “AUTOGRAPH”. 0.7%(0.32)

5.8%(0.19)

22/06/1999 TM75416262 publication for opposition: “AUTOGRAPH”. 0.4%(0.18)

−2.3%(−0.55)

30/07/1999 US7124088 filing: “Apparatus for Internet on-line insurance policy service”. Class: 705/4;Intl. Class: G06Q 40/00. Cites US5797134.

1.3%**

(2.90)0.6%(1.56)

15/05/2000 US6868386 filing: “Monitoring system for determining and communicating a cost ofinsurance”. Class: 705/4; Intl. Class: G06F 17/60. Continuation-in-part application ofUS6064970.

10.5%**

(4.83)18.1%**

(4.84)

16/05/2000 US6064970 issuance: “Motor vehicle monitoring system for determining a cost ofinsurance”. Class: 705/4; Intl. Class: G06F 17/60. Continuation application of US5797134.

4.8%*

(2.22)12.2%**

(3.23)17/04/2001 EP1160707 filing: “Monitoring system for determining and communicating a cost of

insurance”. Intl. Class: G06F 17/60. Priority 15/05/2000, US6868386.−0.6%(−0.18)

−0.9%(−0.15)

24/04/2001 CA2344781 filing: “Monitoring system for determining and communicating a cost ofinsurance”. Intl. Class: G06F 17/60. Priority 15/05/2000, US6868386.AU762134 filing: “Monitoring system for determining and communicating a cost ofinsurance”. Intl. Class: G06F 17/60. Priority 15/05/2000, US6868386.

0.7%(0.21)

0.2%(0.04)

30/04/2001 KR1019980701948 issuance: “Motor vehicle monitoring system for determining a cost ofinsurance”. Inlt. Class: G06F 165/00. Priority 29/01/1996, US5797134.

−0.7%(−0.20)

−0.2%(−0.03)

09/05/2001 JP2002007718 filing: “Monitoring system for determining and communicating a cost ofinsurance”. Intl. Class: G06F 17/60. Priority 15/05/2000, US6868386.

−0.4%(−0.13)

−1.1%(−0.19)

14/05/2001 IL143130 filing: “Monitoring system for determining and communicating a cost ofinsurance” Intl. Class: G06F 17/60. Priority 15/05/2000, US6868386.

0.5%(0.16)

0.8%(0.15)

19/06/2003 AU762134 issuance: “Monitoring system for determining and communicating a cost ofinsurance”. Intl. Class: G06F 17/60. Priority 15/05/2000, US6868386.

−0.6%(−0.48)

−0.5%(−0.21)

07/11/2003 TM78324768 filling: “HOW UDRIVE”.TM78324761, TM78324745 filling: “UDRIVE”.

−0.1%(−0.06)

0.4%(−0.04)

21/11/2003 MXPA03010700 filing: “Motor vehicle monitoring system for determining a cost ofinsurance”. Intl. Class: G06F 17/60. Priority 29/01/1996, US5797134.

0.2%(0.15)

−0.9%(−0.44)

23/01/2004 US20040153362 filing “Monitoring system for determining and communicating a cost ofinsurance” Class: 705/10; Intl. Class: G06F 17/60. Continuation-in-part application ofUS6868386.

−0.6%(−0.52)

−3.8%(−1.79)

01/07/2004 TM78444755 filling: “TRIPSENSOR”.TM78444760 filling: “TRIPSENSE”.

0.3%(0.26)

0.2%(0.15)

10/08/2004 TM78324768 publication for opposition: “HOW UDRIVE”.TM78324761, TM78324745 publication for opposition: “UDRIVE”.

−1.3%(−1.20)

0.0%(−0.69)

20/01/2005 CA2494638 filing: “Monitoring system for determining and communicating a cost ofinsurance”. Intl. Class: G06F 17/60. Priority 23/01/2004, US20040153362.

−1.4%(−1.20)

−2.7%(−1.64)

24/01/2005 EP1557779, EP1557780, EP1746537 filing: “Monitoring system for determining andcommunicating a cost of insurance”. Intl Class: G06F 017/60; G06Q 40/00. Priority23/01/2004, US20040153362.

1.2%(1.29)

0.4%(0.25)

15/03/2005 US6868386 issuance: “Monitoring system for determining and communicating a cost ofinsurance”. Class: 705/4; Intl. Class: G06F 17/60. Continuation-in-part application ofUS6064970.

0.8%(0.78)

2.6%(1.50)

31/05/2005 TM78444760 publication for opposition: “TRIPSENSE”. 1.4%(1.41)

1.6%(0.81)

13/09/2005 TM78444755 publication for opposition: “TRIPSENSOR”. −0.3%(−0.27)

−0.3%(−0.16)

31/08/2006 TM78965099 filing: “A PERSONALIZED POLICY AT A PERSONALIZED PRICE FOR WHO YOUARE, WHAT YOU DRIVE, AND HOW YOU DRIVE IT”.

1.9%(1.71)

1.3%(0.65)

13/10/2006 US20070038488 filing: “Method and apparatus for Internet online insurance policyservice”. Class: 705/4; Intl. Class: G06Q 40/00. Continuation application of US7124088.

−1.2%(−1.03)

−3.2%(−1.60)

17/10/2006 US7124088 issuance: “Apparatus for Internet online insurance policy service”. Class:705/4; Intl. Class: G06Q 40/00. Cites US5797134.

1.4%(1.24)

1.4%(0.72)

05/06/2007 TM78965099 publication for opposition: “A PERSONALIZED POLICY AT A PERSONALIZEDPRICE FOR WHO YOU ARE, WHAT YOU DRIVE, AND HOW YOU DRIVE IT”.

0.4%(0.39)

3.5%*

(2.05)

P. Desyllas, M. Sako / Research Policy 42 (2013) 101– 116 109

Table 3 (Continued)

Date Event description Abnormalreturn

Cumulativeabnormal return

08/06/2007 TM77201265 filling: “PAY HOW YOU DRIVE”.TM77201266 filling: “PAY AS YOU DRIVE”.

0.1%(0.07)

0.0%(0.04)

07/09/2007 TM77977541, TM77273925 filling: “MYRATE”. −0.5%(−0.40)

−2.5%(−0.23)

20/11/2007 TM77201265 publication for opposition: “PAY HOW YOU DRIVE”.TM77201266 publication for opposition: “PAY AS YOU DRIVE”.

−0.9%(−0.69)

−0.4%(−0.40)

09/01/2009 EP877992 refusal: “Motor vehicle monitoring system for determining a cost of insurance”.Intl Class: G06F 165/00. Priority 29/01/1996, US5797134.

−0.3%(−0.12)

2.4%(0.64)

17/03/2009 TM77977541, TM77273925 publication for opposition: “MYRATE”. 2.6%(1.16)

2.8%(0.67)

30/03/2009 EP1160707 refusal: “Monitoring system for determining and communicating a cost ofinsurance”. Intl. Class: G06F 17/60. Priority 15/05/2000, US6868386.

−0.9%(−0.38)

3.6%(0.90)

26/03/2010 TM77969553 filling: “SNAPSHOT”. 3.6%*

(2.43)3.6%(1.41)

17/08/2010 TM77969553 publication for opposition: “SNAPSHOT”. −0.5%(−0.49)

1.2%(−0.28)

US Patent Classes: 705/04: data processing: financial, business practice, management or cost/price determination, with applications to insurance; 705/10: as above, withapplications to market analysis, demand forecasting or surveying. Intl. Patent Classes: G06F 17/60 (G06Q 40/00 since January 2006): data processing systems or methodsfor administrative, commercial, managerial or forecasting purposes, with applications to finance and insurance; G06F 165/00: computing or data processing equipment ormethods for guiding a vehicle; G05D 27/00: systems for controlling or regulating multiple variables. Country coverage of patents: AU – Austria, BR – Brazil, CA – Canada, EP– EU states, IL – Israel, JP – Japan, KR – Korea, MX – Mexico, and US – USA, WO – World Intellectual Property Organization (WIPO) member states. TM stands for trademarks.

tcbsipisuLpu

dstbtoPgmarw

vipacstmPaft

U

* Significant at the 0.05 level. z-Tests in brackets.** Significant at the 0.01 level. z-Tests in brackets.

There is consensus among our interviewees that patent protec-ion was sought to ensure that Progressive “is not inhibited by aompetitor that patents first.” Patent filing is best characterised aseing a “defensive filing” (Ling and Hutchinson) or a “pre-emptivetrategy” (Buechner). Ex-post, however, Progressive has exploitedts IP rights to prevent imitation. In January 2011, Progressive filed aatent infringement lawsuit against Allstate and Liberty Mutual for

nfringing on its patent rights. An out of court settlement with All-tate has been reached: Progressive has granted a license to Allstatender Progressive’s PAYD patent portfolios. The lawsuit againstiberty Mutual has not concluded. Thus, at least to some extent,atent protection has safeguarded Progressive’s market share insage-based insurance.

The interviews provide evidence that patents contributed to theevelopment of PAYD as a marketable service offering. As Hutchin-on comments, “when you introduce something as radical you wanto understand the kind of customers, the retention of customers, theusiness economics etc. and that takes time to develop.” Accordingo several of our interviewees, the existence of patent protectionffered Progressive time to experiment until it got the design ofAYD right. Buechner explained that “the patent rights allowed Pro-ressive to hold off companies from entering into the market and gaveore breathing room to experiment and see what works in the market

nd what data they can get from the device. They had a little moreoom than otherwise, without pressure to do something before theyere ready to do it.”

The finding that patents were not particularly effective means ofalue capture is supported by the stock market’s reaction surround-ng the filing and issuance dates of Progressive’s PAYD-relatedatents. The last two columns of Table 3 report the percentagebnormal returns (ARs) on the event date and the percentageumulative abnormal returns (CARs) over the three-day periodtarting one day before the focal event date. The results suggesthat the majority of patents are associated with a neutral stock

arket reaction, including the ancestor of the patent family forAYD (US5797134). The statistically and economically insignificantbnormal returns imply that the investors judged that most of the

ormal IP protection events did not improve Progressive’s capacityo capture the value created from the PAYD innovation.

We find, however, evidence that two US patent filings and oneS patent issuance are associated with significantly positive ARs.

We estimate a statistically significant AR of 1.3% (p < 0.01), butan insignificant CAR of 0.6%, at the filing of patent US7124088(30/7/1999), which describes a method for Internet online insur-ance policy service. The invention relates to a system for automatedInternet online communication of changes to insurance policyparameters, assessing cost consequences of such changes, and test-ing of a range of variations in policy parameters. This invention,although not specific to PAYD, was instrumental to the develop-ment of PAYD insurance by allowing drivers to review their drivingdata and decide whether they wish to obtain the PAYD discountor revert to a conventional insurance service. The filing of patentUS6868386 (15/5/2000) is associated with a significant AR of 10.5%(p < 0.01) and CAR of 18.1% (p < 0.01). This patent, which builds onpatent US6064970, broadens the scope of the claimed invention bythe parent US5797134 and emphasises the generation of a databasecomprising data elements representative of driving characteristicsand the use of the Internet for communicating the recorded dataand the cost of insurance between the insured and the insurer. Italso covers the provision of value-added services (e.g. driving diag-nostic services). The magnitude of the abnormal returns for thefiling of patent US6868386 may be contaminated by the next dayissuance of patent US6064970 (16/5/2000), which led to a signifi-cant AR of 4.8% (p < 0.05) and CAR of 12.2% (p < 0.01). Acknowledgingthe difficulty of disentangling the impact of each of the two adja-cent events, the two related events jointly appear to bring abouta positive stock market response. It is interesting that the threeimpactful patent events are clustered during the period extend-ing from July 1999 to May 2000. We suspect that the presenceof considerable technical, demand and regulatory uncertainty asto the commercial potential of the PAYD method during the earlypre-paradigmatic evolutionary phase of PAYD, made the expectedwealth impact of the early patents negligible. The shareholdersappear to have revised their expectations, attaching a positive valueto PAYD-related patents, only after the commercial potential ofPAYD had been demonstrated through the pilot testing.

Progressive attempted to extend its territorial IP rights byobtaining international patent protection. It filed an application to

the World Intellectual Property Organization (WO9727561) cov-ering Japan, Korea, Mexico and the European Union for patentUS5797134, and adopted a similar strategy for its continuationpatents. However, the European Patent Office’s (EPO) examining

1 arch P

ditsiP

bgtalttraLhalpicttw

6

mdpfidHcetaiictefamiPPsciec

hpa

Hob

10 P. Desyllas, M. Sako / Rese

ivision rejected Progressive’s applications on the basis that thenventions described did not solve a “technical problem”. Despitehe legal hurdles and the fact that Progressive’s applications weretill under examination in Europe, Progressive exploited the pend-ng patent applications to license out the exclusive IP rights forAYD to AVIVA for Europe and Canada.

Thus, we find some support for the existence of a relationshipetween business method patenting and value capture, since Pro-ressive raised income through licensing fees by offering the righto foreign insurer to introduce PAYD in its markets. However, annonymous top strategy executive at AVIVA Europe justified theicensing fees that were paid to Progressive (even after the rejec-ion of Progressive’s patent application to the EPO) by arguing thathe agreements were not aimed so much at obtaining the exclusiveight to introduce PAYD in Europe, but at accessing Progressive’sccumulated know-how from having marketed PAYD in the US.ing explained: “We have a lot of knowledge in this area and weave a lot of people that have been exposed to the process. So therere many things that we can share and convey [. . .] This is where aot of the value comes in on our end.” It appears that Progressive’satent applications served as a vehicle for the packaging and trad-

ng of not only codified but also tacit knowledge, i.e. Progressive’somplementary resources and capabilities that had accumulatedhroughout the innovation process. As far as the stock market reac-ion is concerned, the foreign patent events are generally associatedith insignificant ARs.8

.2. Trademarks

Progressive complemented its patent portfolio with 13 trade-arks registered in the US. It trademarked the names of all the

ifferent marketed versions of PAYD as well as other relatedhrases, such as Pay-As-You-Drive and Pay-How-You-Drive. Ournding about defensive patenting also holds for trademarks: “weid not want to be ultimately locked out by others” (Hutchinson).owever, our interviewees also suggested that trademarks had aompounding effect together with business method patents thatnabled Progressive to protect its know-how relatively more effec-ively. Ling explained that, while patents “layout a variety of thinkingbout usage based insurance”, trademarks “convey what we are doingn a very short phrase or single word term; and build up some equityn that particular term so that if people hear it over time they willome to understand what we do.” Industry experts indicated thathe combination of IP rights, apart from ensuring some degree ofxclusivity over the underlying logic of PAYD, also prevented othersrom exploiting driver perception of PAYD by “inventing around”nd offering close variations of PAYD. Buechner pointed out thatany of Progressive’s competitors responded to PAYD by develop-

ng their own usage-based insurance and claiming that they offer aAYD offering. In fact, the disclosure of detailed information aboutAYD by Progressive through the filing for patent protection and fortate approval has facilitated imitation by competitors. “A lot of ourompetitors have accessed proprietary information through public fil-ng documents and they get the advantages of our learning without thenergy and the expense” (Hutchinson). Trademarks, at least partially,ounterbalanced this problem.

The results from the event study show that the share-

older reaction to trademark events is rarer compared withatent events. The filing and publication for opposition oflmost all the trademarks are associated with statistically and

8 We only obtain a significantly negative CAR for MX199806067 (28/7/1998).owever, the time profile of the 3-day abnormal returns (−4.6%, 1.2%, −2.8%) andur archival analysis using the Factiva database suggest that the estimated CAR isiased from confounding events.

olicy 42 (2013) 101– 116

economically insignificant abnormal returns. The only signifi-cant abnormal return is estimated for the filing of trademarkTM77969553 Snapshot (23/3/2010), which is associated with asignificant AR of 3.6% (p < 0.05), but an insignificant CAR of 3.6%.This trademark filing took place at a time when Progressive hadoptimised the key parameters of its PAYD offering, had expandedfurther afield and was about to aggressively roll it out to therest of the US.9 This explanation re-enforces our earlier argumentthat shareholder expectations as to the commercial potential ofPAYD took shape gradually as the technical, demand and regulatoryuncertainty was resolving.

6.3. Specialised complementary assets

Gaining control over the required complementary assets was amajor challenge for Progressive. This is reflected in the long timethat it has taken before PAYD was fully developed and rolled outacross the US. According to Ling: “It has taken a lot of effort to getwhere we are now where we are able to roll it out in a cost effective wayand to make sense not only for us as insurers but also for our customersand for them to accept what we are asking them, that is provide infor-mation for a potential discount”. Similarly, Hutchinson explained:“in insurance you do not know whether you are making a profit untilyears thereafter. When there are dramatic changes to the existingmodel it often takes a long time to really understand what is differ-ent and whether you are on the right track.” The complexity of PAYDrequired cross-functional expertise and required the modificationof several aspects of Progressive’s primary and support functions(Harbage, 2011). Our focus here is on the specialised complemen-tary assets that are expected to enable Progressive resisting thecompetitive pressures. Drawing on evidence from the interviewsand the archival analysis we identify these assets in Progressive’sinbound logistics, operations, and marketing and sales (see Fig. 2).

Inbound logistics. Progressive possesses a superior capacity toidentify and collect the right kind of data. Designing an effec-tive data collection process was a laborious undertaking becausethe process had to fulfil three conditions. It had to be financiallyaffordable; it had to avoid raising serious privacy concerns by thepolicyholders; and it had to avoid violating the state law. Consid-ering these conditions, Progressive determined the specificationsfor the monitoring device. Despite the fact that the technologiesinvolved in the monitoring device (the GPS and the odometer) werenot novel, Progressive had to determine what variables to monitori.e. variables that are highly correlated with accident risks. Afterspecifying the characteristics of the monitoring device, it had torely on exclusive partnerships with insurance telematics expertsacknowledging that the development and manufacturing of themonitoring device was outside its core competencies: “We do notsee ourselves as a technology company; we are interested in the appli-cations of it.” (Hutchinson).

Being a first-mover, Progressive now possesses a database withlarge amounts of data on miles driven and accidents that have beenaccumulated over a fifteen-year period. Thomson explains, “themore data you have about customers on what times of the day, at whatspeed, what sort of roads, and how aggressively they are driving, thisis how you really understand the customers and how to price them”.As Harald Trautsch (Chief Marketing Officer for Octo Telematics)puts it: “You can buy anything: you can collect as much money as youwant, you can buy resources and market research, but you cannot buy

a database. . .it takes you years to catch up with that”. In this way,Progressive has moved down the learning curve for usage-basedinsurance and compares favourably to its competitors.

9 We also obtain a positive CAR for the filing of TM78965099 (5/6/2007), but ouranalysis suggests that the estimated CAR is affected by confounding events.

P. Desyllas, M. Sako / Research P

A

edobefteseooasoa

ea“(

twuaiawdpnthg

AVIVA (at the time, Norwich Union, a leading UK insurer) estab-

Fig. 2. IP rights and complementary assets for PAYD auto insurance.dapted from Teece (2006).

Operations. Progressive has developed a superior capacity tomploy advanced rating and underwriting processes utilising theeeper level of data offered by the PAYD technology. It has devel-ped a deep understanding of the use of data on actual drivingehaviour and, using appropriate technology solutions, it hasnriched the traditional underwriting practices and dealt withraudulent claims. This process required extensive testing in ordero ensure that everything works properly and is integrated prop-rly, e.g. the database management systems and the administrativeystems for calculating premiums. Now, Progressive has developedxpertise in operationalising variables with high predictive powerf risks; modelling risks; designing appropriate pricing models (e.g.ffering rewards for good driving and/or penalties for bad driving);nd designing an appropriate profit model accounting for the self-election of good and low mileage drivers. As a result, Progressive’sperations are characterised by higher underwriting efficiency andctuarial accuracy compared with its competitors.

Marketing and sales. Progressive possesses the advantage ofstablished relationships with the stakeholders of the auto insur-nce industry. As a first-mover, Progressive had to go through aneducation process” for both the drivers and the state regulatorsHutchinson).

As far as the drivers are concerned, Progressive had to shape cus-omer perception of this very different type of insurance comparedith conventional offerings. Selling PAYD was not “business as

sual” for several reasons. First, Progressive had to convey to drivers simple and clear message as to the type of driving behaviour thats rewarded (e.g. drive outside the rush hours; avoid rapid acceler-tions and sudden stops). An important challenge for Progressiveas to incentivise drivers to improve their driving behaviour usingriving diagnostics and tips for improvement through the PAYDortal. Second, several drivers were concerned with the intrusive-ess of PAYD. Progressive had to find the right balance between

he perceived costs of compromised privacy and the benefits fromigher consumer control over insurance costs. This is why Pro-ressive offered a long familiarisation period: “consumers who have

olicy 42 (2013) 101– 116 111

applied for auto insurance to Progressive are informed that they havethe ability to add this to their policy if they choose. It is voluntary”(Hutchinson). Third, Progressive had to deal with driver concernsas to the uses of information on their driving behaviour. Driverswere concerned that their data would not be passed to the Policeor the Government and would not be used to turn down a claim. Forthis purpose, Progressive developed specific terms of business anddata governance rules. During these fifteen years, Progressive hasbuilt a considerable reputation capital as a usage-based insurancepioneer, which will be hardly matched by followers.

Progressive has also established important relationships withthe state regulators. This was necessary because in some statesregulation challenges the insurers’ ability to monitor driving usingtelematics. For instance, the state of California limits the parame-ters that companies can use in pricing risks. The pattern of the initialexpansion of PAYD to 19 US states has been explained by Hutchin-son: “We had conversations with individual states departments ofinsurance and we proposed and explained PAYD and got a feeling oftheir willingness to allow this.” State regulators gradually overcometheir initial scepticism and are becoming much more positively dis-posed toward usage-based insurance. They understand that saferdriving leads to fewer and less severe accidents and casualties andimproved driving conditions. They also appreciate that fewer milesdriven reduce significantly emissions and are now keen to incor-porate usage-based insurance in their climate plans. Progressive’sties with the state regulators might prove critical in the devel-opment of a favourable regulatory environment for Progressive’sversion of usage-based insurance across the 50 American states.The forthcoming federal regulation on auto insurance as part of the“Highway bill” has the potential to streamline how states receiveusage-based insurance which can facilitate the rollout of PAYD.

6.4. Competitive dynamics and the future of PAYD insurance

Progressive conceived, designed, developed and launched PAYDinsurance in the US market, which represents one of the mostpatent friendly in relation to formal IP protection. Despite Pro-gressive’s best efforts to protect its IP, ten of the top 25 insurancecompanies in the US began piloting usage-based insurance in theearly 2010s, including American Family Mutual Insurance (TeenSafe Driver, a program that uses technology to capture drivingevents in need of correction and communicated to the driver),GMAC (a low mileage discount program), Liberty Mutual (OnboardAdvisor, a program helping fleets identify and reduce at-risk driv-ing behaviour), State Farm (Drive Safe & Save, a program offeringdiscounts on the basis of mileage and acceleration/deceleration),Allstate (Drive Wise, a program similar to Progressive’s MyRate),SafeCo (Teensurance, a program offering discounts for satellitetracking enabling parents to disable their teens’ vehicle if drivenoutside agreed boundaries/time). With the exception of LibertyMutual (which is in legal dispute with Progressive) and Allstate(which received a license from Progressive), the other insurers havepurposefully structured their PAYD offerings in such a way so thatthey do not directly infringe Progressive’s IP rights. Although, thesealternative offerings are close substitutes to Progressive’s PAYD,Progressive has maintained control over its own versions of usage-based insurance in the US.

An interesting question is how PAYD insurance would havedeveloped if patent protection was not available. Since this ques-tion cannot be answered for the US market, it might be instructiveto look at the development of PAYD insurance in the UK wherebusiness methods are not eligible for patent protection. In 2002,

lished a cooperative agreement with Progressive and was the firstUK insurer to start developing a PAYD pilot scheme. AVIVA collab-orated with IBM and Orange and launched a PAYD pilot with about

1 arch P

5AAosDasp(nPyadfritewoCclcacw

mtaPatHwp

hisoaltbj

aiubFthhriTfec

t

12 P. Desyllas, M. Sako / Rese

000 drivers in 2004. However, the program was unprofitable andVIVA withdrew from PAYD in 2007. Industry experts attributedVIVA’s bitter experience to an ill-chosen business model: it did notutsource enough technology-related activities and its marketingtrategy targeted older drivers that tend to be pay low premiums.espite a discouraging start, we are currently witnessing multiplettempts with PAYD pilots or programs by established UK players,uch as the Cooperative Insurance Group (Pay-How-You-Drive, arogram similar to Progressive’s MyRate for young drivers), AXAat a development stage), AA (at a development stage), but also aumber of new entrants, such as Insurethebox (a program similar torogressive’s MyRate), i-kube (a program that monitors driving ofoung drivers offering discounts for not driving between at night)nd Ingenie (a program similar to Progressive’s MyRate for youngrivers). Many of these attempts build on AVIVA’s experience andocus on the young driver segment, who normally represent highisks and face very high premiums with conventional insurance. Annteresting case is the launch of “Pay-How-You-Drive” in 2011 byhe established insurer Cooperative Insurance Group. The companyxploits its underwriting and marketing expertise and collaboratesith Cobra’s UK subsidiary, Wunelli, which offers all the technol-

gy required. Having developed the offering in about a year, theooperative has been very successful in terms of numbers of poli-ies sold and profit margins achieved in the first six months sinceaunch (Ansley and Thomson). Another interesting aspect of thisase, is that Wunelli has also launched its own usage-based insur-nce version under the name “Coverbox” by partnering with otheronventional auto insurers and hence simultaneously competingith the Cooperative Group.

Thus, it seems that fast imitators and new entrants have hadore freedom to develop their own PAYD insurance offerings in

he UK compared with the US. In other words, the availability of relatively stronger IP appropriability regime in the US facilitatedrogressive to slow down direct imitation, exploit its first-moverdvantage to build the requisite complementary assets, and main-ain some control over the development of usage-based insurance.owever, the sustainability of Progressive’s competitive advantageill likely depend on the ways in which Progressive and its com-etitors deal with three challenges in a changing environment.

First, the technology involved in PAYD has matured and thereas been a significant reduction in the operating costs. Also, an

ncreasing number of car manufacturers now embed telematicsolutions in new vehicles. Therefore, the imitators of PAYD are notnly avoiding the development costs that Progressive incurred, butlso do not have to worry about the cost of manufacturing, instal-ation and maintenance of the monitoring device. As the cost ofhe technology is coming down, the provision of PAYD insuranceecomes financially attractive for a larger number of drivers, not

ust those with paying high premiums with conventional insurance.The second change has been the emergence of a new industry

rchitecture with industry dis-integration and vertical special-sation of the participants. The development and growth ofsage-based insurance has been followed by a rise in the num-er and role of specialised insurance telematics service providers.or instance, Octo Telematics has emerged as a global leader inhe development of cutting-edge telematics applications. Theyave partnered with 60 insurance companies across the world andave collected information on 60 billion miles driven and 200,000eal crashes. They claim to possess the largest insurance telemat-cs database in the world (Trautsch). Similarly, Cobra Automotiveechnologies is a telematics expert offering all the services requiredor usage-based insurance across Europe. The services of these firms

nable established firms and new entrants to minimise the handi-ap of a weak initial positioning in complementary assets.

Third, there is a remarkable change in the public attitudesowards privacy, leading to lower perceived costs arising from the

olicy 42 (2013) 101– 116

intrusiveness of PAYD. Over the last decade the spread of socialnetworking sites and the advances in mobile telephone devices andapplications have reduced peoples’ inhibitions to sharing their per-sonal information and data. As Jackson explains, people will soonunderstand that “the telematics device in the vehicle is no differentfrom the mobile phone in your pocket”.

The declining costs of the provision of PAYD insurance also coin-cides with growing restrictions on insurers’ ability to set premiumson the basis of identifiable driver characteristics (see EuropeanCourt of Justice’s ruling in February 2011 that it is illegal to pricediscriminate on the basis of gender). Now, there is agreementamong most insurers that “the box is the way to go” (Thomson).But as the telematics technologies employed by different usage-based insurers are converging, it will be increasingly harder to avoidcommoditisation. This does not necessarily imply that there is noscope for differentiation, but that the terms of competition andthe sources of differentiation are changing. An important differ-entiating factor will be how the insurer interacts and influencesdrivers (Jackson). Further, Trautsch is convinced that “insurancefirms can add various factors to their offerings to develop a commodityinto a product, or service, or even experience.” Ansley explains, “thequestion is how you pitch it and present it to the customer and howthe customer interacts with the product.” Several firms are explor-ing opportunities to bundle PAYD insurance with complementaryvalue-added services, including breakdown service, navigation,speed cameras notification or even real-time accident preventionservices by alarming drivers of unsafe driving. Progressive has beenone of the first ones to offer complementary services that are tightlylinked to its core business by leveraging its advantages (e.g. it offersreal time response to accidents, theft recovery, remote door unlock-ing, travel directions, emergency assistance). Other insurers havestarted considering opportunities to offer broader complementaryservices, some of which diverge from traditional insurance-relatedofferings. For instance, there are services utilising the rich informa-tion content about the PAYD policyholders to offer advertising oflocation-based services or online social networking activities (e.g.interact with other drivers by sharing and comparing driving data).Here, there will be strategic decisions to be made. Progressive’smanager explains: “We have approached this initially purely from aninsurance standpoint. Others have approached it in a different manner.But that adds risk and complexity and cost to the whole equation”.

7. Discussion

The analysis of the PAYD case allowed a detailed examination ofhow Progressive sought to capture value from its innovative busi-ness model. Similar to IP protection strategies for technologicalinnovation (e.g. Cohen et al., 2000), formal IP protection mainlyserved purposes that are in discordance with the basic rationale ofthe IP system. Business method patents and trademarks were pri-marily sought by Progressive as a pre-emptive strategy. It is arguedthat Progressive would have found a way to commercialise thePAYD idea even in the absence of formal IP protection: “If you lookat Progressive’s history you will see that we have introduced severalinnovations some of which were not patentable. When we can do thiswe try to patent” (Ling).

We do find, however, some evidence which is broadly consis-tent with the value appropriation rationale. First, Progressive hasfiled a lawsuit against competitors that it considers infringe its IPrights. Second, Progressive’s patents and pending patent applica-tions facilitated the establishment of licensing agreements with

AVIVA in Europe and Allstate in the US. In the case of AVIVA, whichwas not a direct competitor of Progressive, the evidence suggeststhat AVIVA obtained a licensing agreement primarily in order toobtain access to Progressive’s accumulated tacit know-how, rather

arch P

taetamet1tiRo(a

tpppaSptpkTmft

edobTifbtafPd

fktpwaLrtTtrrhaucscit

P. Desyllas, M. Sako / Rese

han exclusive rights of PAYD. It appears that such agreements serves vehicles for the packaging and trading of not only codified knowl-dge (here encapsulated into the patented PAYD method) but alsoacit knowledge (Progressive’s complementary assets that it hasccumulated throughout the innovation process). Thus, such agree-ents enable the licensee to build an absorptive capacity to better

xploit the know-how codified in patent records by drawing onhe licensor’s accumulated tacit know-how (Cohen and Levinthal,990, 1989). This patent function is likely to be particularly impor-ant for the transfer of new business models because imitations constrained by causal ambiguity (Barney, 1991; Lippman andumelt, 1982) and depends on obtaining control over a systemf interdependent assets and activities (see Henderson and Clark1990) on “architectural innovation” and Zott and Amit (2010) onn “activity system perspective” of business models).

Like technological innovation studies, our analysis shows thatrademarks had a compounding effect on Progressive’s attempt torevent competitors from inventing-around and exploiting drivererception about what PAYD auto insurance is about. If anything,atents alone facilitated the diffusion of information and gener-ted knowledge spillovers (see Cohen et al., 2000; McGahan andilverman, 2006). This finding suggests that different forms of IProtection are complements rather than independent forms of pro-ection. They protect different aspects of a firm’s IP: while patentsrotect PAYD as a novel service product, trademarks protect mar-eting and branding (Amara et al., 2008; Tether and Massini, 2007).he complementarity of different formal IP protection methodsay be stronger in the case of business models due to their multi-

aceted nature and the fact that business model innovation typicallyakes place in a rather weak appropriability regime.

The case shows that, as Progressive’s PAYD methods werentering the paradigmatic phase, it was the development andeployment of specialised complementary assets that effectivelyffered Progressive an advantage over rivals. These assets wereuilt gradually over the long gestation period of the PAYD offering.he interviews revealed that Progressive’s formal IP rights allowedt to hold off competitors until it got PAYD’s design right. Thisunction was important because the implementation of the newusiness model required extensive experimentation and a longrial-and-error process over a fifteen-year period (Govindarajannd Trimble, 2005; McGrath, 2010; Sosna et al., 2010). There-ore, formal IP rights contributed to the capturing of value fromAYD indirectly, by offering Progressive the vital space required toevelop the necessary specialised complementary assets.

This influence of formal IP protection methods is confirmed byurther evidence. First, by an event study analysis using stock mar-et data. We estimate a generally neutral stock market reactiono formal IP rights events, particularly during the early pre-aradigmatic phase of PAYD in the 1990s (this is consistent with theell-documented poor private value of patents in general by Arora

nd Ceccagnoli, 2006; Austin, 1993; Cohen et al., 2000; Hall, 2009;evin et al., 1987). However, the results show that the investorsevised their valuation expectations about the PAYD insurance withime and reacted positively to particular patents and trademarks.he positive stock market reaction to patents took place only afterhe uncertainty as to the commercial potential of PAYD had beeneduced through the pilot testing in the late 1990s; the positiveeaction to trademarks took place much later, when Progressivead optimised all the key parameters of its PAYD offering and wasbout to roll it out across the US. Thus, stock market investors eval-ated positively formal IP rights only after there were tokens for theommercial potential of PAYD and the possession of the required

pecialised complementary assets by Progressive. This finding isonsistent with the estimated positive relationship between annnovator’s market value and the possession of IP protection whenhe innovator also owns the required specialised complementary

olicy 42 (2013) 101– 116 113

assets (McGahan and Silverman, 2006). Second, by comparing theevolution of usage-based insurance over time in the US and theUK. The strengthening of the IP appropriability regime with theexpansion of formal IP protection for business methods in the USinfluenced the landscape in favour of Progressive, the first-moverin usage-based insurance. Progressive has been able to deal withdirect imitators and maintain some control over the developmentof usage-based insurance in the US.

An important insight is that Progressive’s possession of thenecessary complementary assets was a result of the innovation pro-cess, rather than a strong initial position. Thus, the complementaryassets that are required for the implementation of a new businessmodel do not necessarily exist ex ante but are built as the innova-tion process unfolds and are often endogenous to the possession offormal IP protection. Our case suggests that the outcome of an inno-vation battle is not determined by the initial positions in specialisedcomplementary assets of the innovator, the complementors andpotential imitators, but by their relative capability to develop theseassets over time.

Last but not least, the PAYD case demonstrates that, as thetechnology central to PAYD matured and as the auto insuranceindustry architecture evolved through vertical dis-integration andfirm specialisation, there were noticeable changes in the specialisedcomplementary assets that mattered for competition (Coombset al., 2003; Jacobides et al., 2006). The emergence of a number ofspecialised insurance telematics suppliers led to multiple attemptsto imitate PAYD insurance by both incumbents and start-ups. In thisnew market environment, the terms of competition and the sourcesof differentiation shifted towards downstream activities. Severalinsurers have begun experimenting with the bundling of servicesbeyond insurance in order to differentiate. More generally, servicebundling changes a firm’s positioning in the value network fromone that relies on a “commodity business model” to one anchoredin a “platform business model” (attracting companies to invest invalue-enhancing activities for the platform) (Chesbrough, 2011: p.105; Gawer et al., 2008). As other scholars point out, the businessmodel construct emphasises the dynamism of competitive advan-tage (Demil and Lecocq, 2010; McGrath, 2010; Teece, 2007, 2010).Thus, in the realm of dynamic capabilities, we argue that a com-petitive imperative for business model innovators is to constantlyfine-tune or reconfigure the assets that are complementary to theirbusiness models as their market environment changes.

8. Conclusion and implications

This paper contributes to the emergent work on how incum-bent firms innovate and profit from new business models. ThePAYD case shows that conceptualising and implementing a novelbusiness model is not sufficient for a firm to achieve competitiveadvantage and above average returns. An integral part of a businessmodel innovator’s competitive strategy should be the capacity tocapture the resultant value. But this is a challenging undertakingbecause innovative models can hardly be tested in laboratories, andbusiness model innovators tend to operate in a weak IP appropri-ability regime–even after the recent expansion of what constitutespatentable subject matter to cover novel business methods in somejurisdictions, including the US.

The PAYD case demonstrates that formal and strategic IP protec-tion methods play distinctive but complementary roles. Securingformal IP protection for various business model components isimportant in the early exploratory period as a defensive strat-

egy, as a vehicle for the packaging and trading both codified andtacit know-how, and most importantly as a means of “buyingtime” in order to build a strong position in specialised complemen-tary assets, thus raising barriers to entry for competitors and new

1 arch P

etmstcbcc

Pitatitpclc

tmfcattutfipftfims

rtctgehsocePvplcccFutddIPai

returns in the absence of unusual events) are computed using thestandard market model technique. In this model, the returns to a

14 P. Desyllas, M. Sako / Rese

ntrants. Even when the key business model components are pro-ected through formal IP rights, the multifaceted nature of business

odels implies that competitors may be able to invent around usingtrategically equivalent assets and eventually develop a close varia-ion of the focal business model. Thus, the innovator’s longer-termompetitiveness depends on whether it develops superior capa-ilities (relative to competitors) to build the requisite specialisedomplementary assets and constantly reconfigure them in line withhanges in the market environment.

An additional contribution of our work is that it enriches theFI framework, which was originally developed with technologicalnnovation in mind. The findings from this study about the poten-ial endogeneity of the requisite specialised complementary assetsnd the need for constantly reconfiguring them may also apply toechnological innovation. Thus, these two findings represent mod-fications of the PFI framework which may well be generalisableo some categories of technological innovation. The finding thatatent licensing can be employed as a vehicle to convey not onlyodified but also tacit know-how may be less relevant to techno-ogical innovation compared with business models that typicallyonsist of multiple components and are subject to opacity.

Our findings have important implications for management prac-ice. With the expansion of formal IP protection for business

ethods in some countries, there is greater opportunity to profitrom business model innovation as long as firms pay attention toomplementary assets. We suggest that managers should adopt

sequence of interrelated actions in order to effectively pro-ect their novel business models. In the short-term, they shouldry to ring-fence the constituent parts of their business modelsing a strategically orchestrated portfolio of formal IP rights. Inhe medium-term, they should exploit their IP rights and otherrst-mover advantages in order to develop the specialised com-lementary assets that can act as an effective isolating mechanismrom competition. As for the longer-term, managers need to ensurehat their organisations develop the capability to constantly recon-gure their complementary asset base in line with changes in thearket environment in order to resist the competitive parity pres-

ures from imitators.We acknowledge the usual limitations that apply to case study

esearch. An assessment of the generalisability of the findings fromhe present study will require further work in different empiricalontexts. PAYD can, however, be seen as a particular version ofhe utility or “on-demand” model which is based on a “pay as youo” approach (Rappa, 2004, 2010). This family of business mod-ls are based upon metering usage and cost/price calculations andave applications to a wide cross-section of services, ranging fromimple car rental services (e.g. City Car Club, UK’s largest networkf hourly rental cars) to smart phones and more complex cloudomputing services (e.g. IBM’s on demand computer services). Thexample of cloud computing, in particular, has many parallels withAYD insurance. There is a growing reliance on computing fromarious types of users, including individuals, businesses and theublic sector. The current business models based on purchasing or

easing of IT products is expected to give way to a “usage-basedomputing” business model that will deliver infrastructure, appli-ations and business processes in a secure, shared and scaleableomputer environment over the Internet for a usage-based fee.urthermore, the cloud computing innovation space is already pop-lated with numerous software and Internet patents as well asrademarks and copyrights. Formal IP protection is likely to haveifferential effects on the capability of the players in this space toevelop and profit from business model innovation (e.g. AKAMAI’s

ntelligent Platform, Amazon’s Web Services, Fujitsu’s Global Cloudlatform). Thus, we expect that our findings to be largely generalis-

ble and most directly relevant to “pay as you go” business modelsn other sectors of the economy.

olicy 42 (2013) 101– 116

Acknowledgements

The authors gratefully acknowledge financial support from theUK’s Advanced Institute of Management. They are grateful to DavidSimoes Brown, Lex Donaldson, Annabelle Gawer, Alan Hughes, PaulJackson, Silvia Massini, Ian Miles, Marcela Miozzo, Michael Mol,Robert Pitkethly, Yannis Skulikaris, Kostas Stathopoulos, ChanderVelu, Mo Yammin, and the participants at the Oxford Strategy work-shop, the Academy of Management, Strategic Management Society,and Sloan Industry Studies conferences for helpful comments.

Appendix A. Further details on the archival analysis andthe interviews

The archival analysis is based on information from Progres-sive Corporation’s annual reports, searches of the business andpopular press using Factiva database with appropriate keywords,experts’ industry reports on the auto insurance industry, andcase studies on Progressive and PAYD insurance (Frei, 2004;Hansen and Vandenbosch, 2004). Data on Progressive’s patentsand trademarks were collected from the Delphion database andthe online search engines of the EPO and the USPTO. Our inter-viewees are managers and patent attorneys from Progressive, aswell as external stakeholders. The latter group includes managersfrom AVIVA, the licensor of PAYD for Europe and Canada; IBM,which was AVIVA’s technology partner in the design of the mon-itoring device of PAYD; the Cooperative Insurance Group, whichhas introduced “Pay how you drive” for young drivers in theUK; and Octo Telematics, an Italian-based international insur-ance telematics services company. The interviewees are: FromProgressive Group of Insurance Companies – Richard Hutchinson,General Manager of Usage Based Insurance (6/7/2010); RaymondLing, Associate General Counsel (29/3/2007 and 12/7/2010); SusanRouser, Public Relations (6/7/2010). From IBM – David Robson,Partner, Financial Services and Insurance (UK) (16/7/2007); MarkMcLaughlin, Director of Strategy and Solutions, Insurance Indus-try (23/7/2007); Terry Buechner, Global P&C Insurance SegmentManager (23/7/2007); Roger Burt, Patent Attorney (24/04/2006);Steve Wood, Vice President, Strategy UK & Ireland and GrahamJackson, IBM Insurance Consulting Business (2/11/2011). FromAVIVA – a top Strategy Manager at AVIVA Europe, who wishedto remain anonymous (23/6/2008). From the Co-operative Insur-ance: Nick Ansley, Motor Product Manager, General Insurance(12/10/2011); Andrew Thomson, General Insurance Pricing Man-ager (12/10/2011). From Octo Telematics – Harald Trautsch, ChiefMarketing Officer (11/09/2011). Two other experts that we inter-viewed are: Richard Lawrence, Patent Attorney at Hewlett-Packard(20/12/2005); Todd Litman, Researcher at the Victoria Trans-port Policy Institute (13/3/2007). Interviewing both internal andexternal stakeholders enabled us to cross-check and enhance thereliability of the interview-based data. Four interviews were con-ducted in-person, while the remaining twelve interviews wereconducted through the telephone. Each interview lasted between60 and 90 min. The interviews were recorded, transcribed and ana-lysed.

Appendix B. Further details on the event study analysis

Stock price data were collected from Datastream database. Fol-lowing previous studies (Denis, 1994; Gillan et al., 2000; Westonet al., 2004), the “normal” returns on Progressive’s equity (i.e. the

firm’s equity in excess of the risk-free return are explained by thereturns on a market index. We employ the S&P 500 index to proxy

arch P

fto5edttw1WboitlrItsenf

R

A

A

A

A

A

A

A

B

B

B

B

CC

C

C

C

C

C

C

C

C

D

D

D

P. Desyllas, M. Sako / Rese

or the market returns. The parameter estimates for the compu-ation of the normal returns are derived from an OLS regressionf the returns on Progressive’s equity on the returns on the S&P00 index over the 250-day period beginning 290-day prior toach event date. The abnormal returns (ARs) are calculated by theifference between the actual returns to Progressive’s equity andhe predicted “normal” returns on the event day. In order to testhe null hypothesis that the AR on the event day is equal to zero,e calculate a test statistic as defined in Weston et al. (2004, pp.

66–7), which has approximately a standard normal distribution.e also derive a measure of the cumulated abnormal return (CARs)

y cumulating the ARs over a three-day period, including the datef the IP event (day 0), the previous day (day −1) and the follow-ng day (day +1). Similarly, we test the null hypothesis that thehree-day CAR surrounding the event day is equal to zero by calcu-ating a test statistic as defined in Weston et al. (2004, p. 167). As aobustness check, we re-estimated ARs and CARs using the S&P 500nsurance Index as a proxy of the market portfolio in order to con-rol for industry-specific conditions. The results were qualitativelyimilar and hence they are not presented here. Finally, because thestimation of the abnormal returns is based on the assumption thato other major event occurs during the event window, we check

or the presence of confounding events using Factiva database.

eferences

kerlof, G.A., 1970. The market for “lemons”: quality uncertainty and the marketmechanism. Quarterly Journal of Economics 84, 488–500.

mara, N., Landrya, R., Traoréb, N., 2008. Managing the protection of innovations inknowledge-intensive business services. Research Policy 37, 1530–1547.

mit, R., Zott, C., 2001. Value creation in e-business. Strategic Management Journal22, 493–520.

rora, A., Ceccagnoli, M., 2006. Patent protection, complementary assets, and firms’incentives for technology licensing. Management Science 52, 293–308.

rora, A., Fosfuri, A., Gambardella, A., 2001. Markets for technology and their impli-cations for corporate strategy. Industrial & Corporate Change 10, 419–451.

rrow, K.J., 1963. Uncertainty and the welfare economics of medical care. AmericanEconomic Review 53, 941–973.

ustin, D.H., 1993. An event-study approach to measuring innovative output: thecase of biotechnology. American Economic Review 83, 253–258.

aden-Fuller, C., Demil, B., Lecoq, X., MacMillan, I., 2010. Special issue on businessmodels. Long Range Planning 43, 143–145.

arney, J.B., 1991. Firm resources and sustained competitive advantage. Journal ofManagement 17, 99–120.

irkinshaw, J., Hamel, G., Mol, M., 2008. Management innovation. Academy of Man-agement Review 33 (4), 825–845.

lind, K., Edler, J., Schmock, U., Andersen, B., Howells, J., Miles, I., Robers, J., Hipp,L., Green, L., Herstatt, C., Evangelista, R., 2003. Patents in the Service Industries.European Commission, Brussels.

hesbrough, H., 2011. Open Services Innovation. Jossey-Bass, San Francisco.hesbrough, H., 2010. Business model innovation: opportunities and barriers. Long

Range Planning 43, 354–363.hesbrough, H., Birkinshaw, J., Teubal, M., 2006. Special issue on “Profiting from

Innovation” by David J. Teece. Research Policy 15, 1091–1099.hesbrough, H., Rosenbloom, R.S., 2002. The role of the business model in capturing

value from innovation: evidence from Xerox Corporation’s technology spin-offcompanies. Industrial & Corporate Change 11, 529–555.

hristensen, C.M., 1997. The Innovators Dilemma: When New Technologies CauseGreat Firms to Fail. Harvard Business School, Boston, MA.

ohen, W.M., Klepper, S., 1996. A reprise of size and R & D. The Economic Journal106, 925–951.

ohen, W.M., Levinthal, D.A., 1989. Innovation and learning: the two faces of R&D.The Economic Journal 99, 569–596.

ohen, W.M., Levinthal, D.A., 1990. Absorptive capacity: a new perspective on learn-ing and innovation. Administrative Science Quarterly 35, 128–152.

ohen, W.M., Nelson, R.R., Walsh, J.P., 2000. Protecting their intellectual assets:appropriability conditions and why U.S. manufacturing firms patent (or not).NBER Working Paper.

oombs, R., Harvey, M., Tether, B.S., 2003. Analysing distributed processes of provi-sion and innovation. Industrial & Corporate Change 12, 1125–1155.

emil, B.t., Lecocq, X., 2010. Business model evolution: in search of dynamic consis-tency. Long Range Planning 43, 227–246.

enis, D.J., 1994. Organizational form and the consequences of highly leveragedtransactions Kroger’s recapitalization and Safeway’s LBO. Journal of FinancialEconomics 36, 193–224.

oganova, L., Eyquem-Renault, M., 2009. What do business models do? Innovationdevices in technology entrepreneurship. Research Policy 38, 1559–1570.

olicy 42 (2013) 101– 116 115

Dreyfuss, R.C., 2000. Are business method patents bad for business? Santa ClaraComputer and High Technology Law Journal 16, 263–280.

Eisenhardt, K.M., 1989. Building theories from case study research. Academy ofManagement Review 14, 532–550.

Frei, F., 2004. Innovation at Progressive (A), Pay-as-you-go insurance. Harvard Busi-ness School Case Study 9-602-175.

Gambardella, A., McGahan, A.M., 2010. Business-model innovation: general purposetechnologies and their implications for industry structure. Long Range Planning43, 262–271.

Gawer, A., Cusumano, M.A., 2008. How companies become platform leaders. MITSloan Management Review 49, 28–35.

Ghaziani, A., Ventresca, M., 2005. Keywords and cultural change: frame analysis ofbusiness model public talk, 1975–2000. Sociological Forum 20, 523–559.

Gillan, S.L., Kensinger, J.W., Martin, J.D., 2000. Value creation and corporate diver-sification: the case of Sears, Roebuck & Co. Journal of Financial Economics 55,103–137.

Govindarajan, V., Trimble, C., 2005. Building breakthrough businesses within estab-lished organizations. Harvard Business Review 83, 58–68.

Graham, L., Xie, X., 2007. The United States insurance market: characteristics andtrends. In: Cummins, D., Venard, B. (Eds.), Handbook of international insurance.Springer, Boston, pp. 25–120.

Hall, B., 2009. Business and financial method patents, innovation, and policy. NBERWorking Paper 14868.

Hall, B.H., MacGarvie, M., 2006. The private value of software patents. Unpublishedmanuscript.

Hansen, S.W., Vandenbosch, B., 2004. Progressive insurance: TripSense and the chal-lenges for innovation. Unpublished manuscript.

Harbage, R., 2011. Usage-based auto insurance (UBI): a revolution is underway. Isyour company ready? Towers Watson Presentation, December 2011.

Henderson, R., Clark, K.B., 1990. Architectural innovation: the reconfiguration ofexisting product technologies and the failure of established firms. Administra-tive Science Quarterly 35, 9–30.

Hughes, A., Scott Morton, M., 2006. The transforming power of complementaryassets. MIT Sloan Management Review 47, 50–58.

Jacobides, M.G., Knudsen, T.r., Augier, M., 2006. Benefiting from innovation: valuecreation, value appropriation and the role of industry architectures. ResearchPolicy 35, 1200–1221.

Johnson, M.W., Christensen, C.M., Kagermann, H., 2008. Reinventing your businessmodel (cover story). Harvard Business Review 86, 50–59.

Koen, P.A., Bertels, H., Elsum, I.R., Orroth, M., Tollett, B.L., 2010. Breakthrough inno-vation dilemmas. Research Technology Management 53, 48–51.

Leonard-Barton, D., 1992. Core capabilities and core rigidities: a paradox in manag-ing new product development. Strategic Management Journal 13, 111–125.

Lerner, J., 2002. Where does State Street lead? A first look at finance patents, 1971to 2000. Journal of Finance 57, 901.

Levin, R.C., Klevorick, A.K., Nelson, R.R., Winter, S.G., 1987. Appropriating the returnsfrom industrial Research and Development. Brookings Papers on EconomicActivity, 783.

Lippman, S.A., Rumelt, R.P., 1982. Uncertain instability: an analysis of interfirmdifferences in efficiency under competition. Bell Journal of Economics 13,418–438.

Markides, C., Charitou, C.D., 2004. Competing with dual business models: a contin-gency approach. Academy of Management Executive 18, 22–36.

McGahan, A.M., Silverman, B.S., 2006. Profiting from technological innovation byothers: the effect of competitor patenting on firm value. Research Policy 35,1222–1242.

McGrath, R.G., 2010. Business models: a discovery driven approach. Long RangePlanning 43, 247–261.

Ovans, A., 2000. Can you patent your business model? Harvard Business Review 78,16.

Pisano, G., 2006. Profiting from innovation and the intellectual property revolution.Research Policy 35, 1122–1130.

Rappa, M.A., 2004. The utility business model and the future of computing services.IBM Systems Journal 43, 32–42.

Rappa, M.A., 2010. Business Models on the Web.,http://digitalenterprise.org/models/models.html.

Salter, A., Tether, B.S., 2006. Innovation in services: through the looking glass ofinnovation studies. Unpublished manuscript.

Sosna, M., Trevinyo-Rodriguez, R.N., Velamuri, S.R., 2010. Business model innovationthrough trial-and-error learning: the Naturhouse case. Long Range Planning 43,383–407.

Teece, D., 1980. The diffusion of an administrative innovation. Management Science26, 464–470.

Teece, D., 1986. Profiting from technological innovation: implications for integra-tion, collaboration, licensing and public policy. Research Policy 15, 285–306.

Teece, D.J., 2006. Reflections on “Profiting from Innovation”. Research Policy 35,1131–1146.

Teece, D.J., 2007. Explicating dynamic capabilities: the nature and microfounda-tions of (sustainable) enterprise performance. Strategic Management Journal28, 1319–1350.

Teece, D.J., 2010. Business models, business strategy and innovation. Long RangePlanning 43, 172–194.

Tether, B.S., 2005. Do services innovate (differently)? Insights from the EuropeanInnobarometer Survey. Industry & Innovation 12, 153–184.

Tether, B.S., Massini, S., 2007. Services and the innovation infrastructure. DTI Occa-sional Paper No. 9.

1 arch P

W

W

W

Inc., London.

16 P. Desyllas, M. Sako / Rese

agner, S., 2008. Business method patents in Europe and their strategic use – evi-dence from franking device manufacturers. Economics of Innovation and New

Technology 17, 173–194.

ebb, D., Pettigrew, A., 1999. The temporal development of strategy: patterns inthe U.K. insurance industry. Organization Science 10, 601–621.

eston, J.F., Mitchell, M., Mulherin, J., 2004. Takeovers, Restructuring and CorporateGovernance, 4th ed. Pearson Prentice-Hall, New Jersey.

olicy 42 (2013) 101– 116

Yin, R.K., 2009. Case Study Research: Design and Methods, 4th ed. Sage Publications

Zott, C., Amit, R., 2010. Business model design: an activity system perspective. LongRange Planning 43, 216–226.

Zott, C., Amit, R., Massa, L., 2011. The business model: recent developments andfuture research. Journal of Management 37, 1019–1042.