navigating adtech

10

Click here to load reader

Upload: giorgiogarrido6

Post on 18-Jul-2016

42 views

Category:

Documents


1 download

DESCRIPTION

Understanding AdTech

TRANSCRIPT

Page 1: Navigating AdTech

A guide for Marketers

Navigating Planet Ad Tech

data

analysis

dis

tribu

tionval

uation

identify

IN THIS REPORT

2 Part One: The Ad Tech Sectors Ad Exchanges

6 Part Two: Pricing Cost Per Mille

7 Part Three: Tactics Audience Targeting

8 Part Four: Big Topics Cookies

T he promise of ad

technology is to get

marketers closer to

their customers via

data analysis, immediate valu-

ation and distribution. This

means using data to accurately

identify audiences, determine

the value of those audiences,

and deliver the right messages to

them instantly. The problem is

there’s an abundance of ad tech

firms all trying to capture their

portion of the media budget by

offering these services. With so

many players to choose from

marketers are finding it frustrat-

ing to navigate the space. The

purpose of this report is to out-

line the various players in the

ad technology sector and clearly

communicate their value propo-

sition, business model and

history in order to help marketers

ask the right questions.

+

Page 2: Navigating AdTech

2 NAVIGATING PLANET AD TECH MIT TECHNOLOGY REVIEW + DIGILANTPART 1: THE AD TECH SECTORS

Ad ExchangesAn ad exchange is a marketplace where media is bought and sold, with complete transparency and fluidity, modeled after the same principals as the stock exchange. It is in the exchange where bidding from agency trading desks and demand side platform (DSPs) is done.

Who has an ad exchange? Appnexus, Google Adx, OpenX and Yahoo RMX. Supply Side Platforms (SSPs) such as PubMatic, Admeld and Rubicon also act as exchanges.

History of ad exchanges Ad exchanges were created as venues for the real-time exchange of media inventory to be executed by multiple buyers. When introduced in 2007, real-time bidding (RTB) allowed advertisers to bid auction style on an impression-level basis and load an ad in milliseconds.

How ad exchanges make moneyAd exchanges collect a percentage of each transaction (e.g. 20%); thus their incentive is to conduct as many transac-tions as possible, sometimes resulting in transactions that may not be transparent to the publisher.

Benefits of ad exchangesExchanges provide publishers and adver-tisers with an open and automated online advertising market, promising liquidity, operational efficiency, the ability to cherry-pick impressions and transparency.

Concerns with ad exchanges In reality there is no such thing as a true exchange for media. Every publisher and advertiser has proprietary relationships with various platform providers. Every exchange has its own relationships with various Demand Side Platforms (DSPs) and SSPs. Advertisers should split their buys across multiple platforms to make sure they are exposed to all available publisher inventory.

Today most of the inventory found on exchanges is of varying quality, including a mix of long-tail and premium inventory offered side by side.

Many exchanges may not give the context of where an impression is being served (e.g. the publisher) as much value as it should. Qualitatively, it can be difficult to score or value an impression.

There are potential challenges around brand safety and fraud if impressions are not served as the buyer intends them to be: for example, ads might be served on disrepu-table sites or may be fraudulently generated.

Some ad exchanges offer their own buy-ing platforms so that advertisers can pur-chase media directly through the exchange. Advertisers should know this can result in a conflict of interest, since the exchange is now representing both the advertiser (buyer) and seller (publisher) of media under one roof.

Questions to ask before employing an ad exchange

n How do I ensure I am not buying blindly in a large category?

n What is the level of visibility into the impressions I am buying?

n What is the percentage of your inventory by category?

Agency Trading DesksAgency trading desks are the audience-buying divisions of agency holding companies. Their promise is to help match the right audiences to the right ad impressions efficiently at scale.

● Trading desks do not typically ownproprietary technology. They are made up of people who specialize in analytics and ad trafficking, and they often license one or more demand side platforms (DSPs), which collect and analyze audience data, to deliver on this promise. The trading desk team may also include ad operations teams from each of the agency DSPs.

Where are trading desks found? Most large holding company owned adver-tising agencies have trading desks and a few independent ones exist also. Those owned by agency holding companies typically serve their own clients, and include WPP’s Xaxis, MDC’s Varick Media, Omnicom’s Accuen, also Publicis’s AOD , IPG’s MAP, Dentsu’s AmNet and, Havas’ AffiPerf. Accordant Media is an independent trading desk.

History of agency trading desks Agency trading desks emerged for reasons of operational efficiency. By 2007 it was clear that advertising productivity could be enhanced using technology, making it easier for the agency staff to pick and choose the best media placements on behalf of their clients.

Buying and delivering online media is more expensive and labor intensive than traditional media, so trading desks were created to streamline the system and keep the costs in check.

Benefits of agency trading desks Agency trading desks crunch and analyze audience behavioral and demographic data to uncover the insights needed to help a client reach its optimal audience and deliver on its key performance indi-cators (KPIs).

The Ad Tech Sectors

Part One

Page 3: Navigating AdTech

3 NAVIGATING PLANET AD TECH MIT TECHNOLOGY REVIEW + DIGILANTPART 1: THE AD TECH SECTORS

Data Management PlatformsA growing number of ad tech vendors are incorporating data management platforms (DMPs) as part of their service to clients. DMPs are data warehouses for gathering and analyzing both first and third-party data.

● Some of this data is collected offline, or at least off the Internet; for example supermarket data collected from loyalty cards, or customer registration data. The rest is collected online, either from past campaigns or via cookie data.

DMPs can now also track mobile data assets, as well as unstructured audience data collected online from sources like video, social and web analysis tools, and points of sale.

DMPs are aggregators, collecting and managing data that enables marketers to scrutinize characteristics of an audience segment, which can then be translated into more focused and effective cam-paigns and content. This means clas-sifying audiences into categories, such as “Technology Enthusiast” or “Soccer Mom.”

DMP technology is typically deployed within a marketing department or inside the agency’s trading desk to help marketers understand their customers.

What DMPs are out there? Examples of stand-alone DMPs include Aggregate Knowledge, Adobe (Demdex, now called Adobe Audience), Audience Science and Lotame. Third-party data providers like Blue Kai and eXelate, who aggregate behavioral data to create thousands of targetable audiences, are now positioning themselves as DMPs/DSPs with integrated data management functionality.

Acxiom, Adchemy, Datalogix, Demdex (now owned by Adobe), Digilant, Epsilon, eXelate, Experian, Krux Digital, Lotame, Mediamath, Targetbase, Targusinfo,

They also operate within the existing advertising agency making it possible for the client to maintain fewer relationships. The agency trading desk may also benefit from the traditional media buying teams being under one roof because they can leverage and share information internally.

How agency trading desks make money Trading desks make their money on fees, commissions, or in some cases, arbi-trage—profits earned by reselling media to their clients.

Concerns with trading desks Most of the criticism regarding trad-ing desks is focused on transparency of pricing, and specifically on the issue of arbitrage, a term familiar to many from international currency trading. Arbitrage in this context is buying up media inven-tory space as principal to the transaction, then reselling it to an advertiser (agency client) usually at a higher price, violating the concept of “agency”. Not all trading desks engage in arbitrage.

The trading desks are justifying their practices by claiming that the “impres-sions” they are selling (each time an online ad is displayed constitutes an impres-sion) become more valuable once they

have layered on their own insights. They also claim that arbitrage profits are rein-vested into new technology, which in turn benefits the client.

It’s essential to ask questions before employing a trading desk. Here are a few examples.

Questions to ask before choosing a trading desk

n What is your business model? Do you practice arbitrage?

n How much money goes to the agency, how much to the trading desk?

n How do you set performance goals?n How can I be certain I am reaching

my target audience?

n How many inventory sources are you connected to?

n What technology platforms do you use and what criteria do you use to evaluate their performance?

n Do you use ad verification? n How are your partners evaluated

on an ongoing basis, and by whom?

n How do you protect my brand?n What inventory sources outside

of the ad exchanges are you connected to?

n How is my data protected?

Page 4: Navigating AdTech

4 NAVIGATING PLANET AD TECH MIT TECHNOLOGY REVIEW + DIGILANTPART 1: THE AD TECH SECTORS

Demand-Side PlatformsDemand side refers to the purchaser of media—or the advertiser. Demand-side platforms (DSPs) are systems that serve ads on behalf of the advertiser in real-time based on specified rules. Those rules are based on the data provided by advertisers and by inventory sources (Exchanges and SSP’s) when an impression is offered for auction or sale.

● The data stream, sometimes called bid stream, provides limited information about a user and the content where the impression will be displayed. A DSP may use behavioral targeting data to deter-mine if the audience segment and con-tent should be targeted. If the targeting criterion meets the advertiser’s objec-tives some DSPs determine how much it should bid for such impressions on behalf of their client using algorithms that analyze the incoming data stream.

DSPs are most often licensed by agency trading desks but they are also used directly by independent agencies that don’t have internal trading desk teams, ad networks and some brands that do their own media planning and buying. DSPs focus mainly on display advertising but are increasingly tapping into video and mobile inventory, as well as newly developed social opportunities.

Who has a DSP? Platforms include Dataxu, Digilant, Invite (acquired by Google) MediaMath, Rocketfuel, The Trade Desk, Turn, and [x+1].

History of DSPs DSPs were created to facilitate real-time bidding on ad exchanges. Each DSP must hold a seat on each exchange in order to receive the bid stream infor-mation. In response to slow adoption by agencies, the first DSPs (Turn, Rocket-fuel, [x+1]) positioned themselves as ad networks, a perception that persists in the marketplace.

and [x+1] are DSPs with DMP functionality, sometimes referred to as consolidated media buying platforms, while Collective and Turn offer similar services plus addi-tional publisher offerings.

History of DMPs Demdex, the first DMP, was launched in 2008 as an improvement upon cobbled-together in-house systems working to collect and ana-lyze the large volume of new online data. Publishers started using DMPs to manage and segment their audience data, enabling them to profit from audience targeting and site personalization. Real-time audience buying prompted agencies and brands to use DMP technology to understand and oversee the customer journey.

Benefits of DMPsDMP technology exists to help marketers and publishers simplify data analysis to discover trends and deconstruct audience behavior so that clients can formulate optimal targeting strategies.

A DMP can tell you, for example, how many users bought children’s clothes in a certain range of sizes and also searched for video game software; it is also possible to estimate those users’ income levels, geographic location and other interests.

DMPs have a quite varied range of offerings. None of them provide all of the following services so it is important to know what you are trying to deliver before selecting a provider.

Here is a smapling of those varied services:

n Integrate reporting across real-time bidding (RTB), non RTB and other digital channels.

n Analyze and report on third-party data audiences that are right for the brand.

n Develop derivative audience or look-alike models of combined data assets.

n Distribute audience buys across RTB/non RTB inventory sources.

n Integrate advertiser first-party non-digital CRM data with digital network data.

n Integrate publisher first-party data with extended digital data.

How DMPs make money Pure-play DMP vendors charge a flat monthly fee based on volume of data assets. Companies that offer integrated DMP services may bake their fees into the cost of media.

Concerns with DMPs As ad tech vendors offer increasingly comprehensive solutions for clients, some traditionally stand-alone DMPs have begun to offer audience buying as well. But this can be problematic since it may incentivize the DMP to share data from various customers across the DMP organization to improve the overall performance of audience buys.

Another concern is whether some DMPs can provide the analytics to explain why an audience might be the right target to achieve a marketer’s KPIs. All too often the information provided doesn’t contain actionable insights into what influences customers. And it can still be difficult to show that the data improved ROI.

Questions to ask before employing a DMP

n How will you define the metrics that will best help me best understand my customer or potential customer?

n How will you measure the reach of those users across media channels and across touch points, both online and offline?

n How do you manage third-party data vs. first-party data?

n How is data exchanged between different platforms?

n How would you handle a client that wants to take a certain segment of data and cross-reference or combine it with another data set from another provider?

n How do you know what a good prospect looks like?

n How can I be sure the data you are using is accurate and current?

n Are you reselling or giving away any data that might be private?

n How can I generate new sales by using a DMP?

Page 5: Navigating AdTech

5 NAVIGATING PLANET AD TECH MIT TECHNOLOGY REVIEW + DIGILANTPART 1: THE AD TECH SECTORS

direct connections to DSPs. And all SSPs offer yield management tools for publishers to help them maintain and control price and volume of inventory.

Who has an SSP? The big ones are Pubmatic, Rubicon and Admeld (acquired by Google). Some ad networks are now referring to themselves as SSPs as well.

History of SSPs SSPs emerged as a more efficient way for larger publishers to navigate ad networks and fill as much remnant inventory space as possible.

How SSPs make money The SSP collects a percentage of each transaction (10-15%) from the DSPs.

Benefits of SSPsSSPs help marketers access more valuable inventory during the real-time bidding process, and they help uphold a publish-er’s brand standards by maintaining the right to block certain advertisers from accessing their inventory.

Concerns with SSPsSSPs, like exchanges, offer publishersyield management tools to manipulate the open market. Some examples include the ability to set floor prices, which give the publishers control over the minimum price of an impression sold via the SSP.

Another concern is that many SSPs offer DSP like services in addition to sell-ing publisher inventory, meaning they represent both the buyer of media and the seller. This is often seen as a conflict of interest. If a majority of the SSPs rev-enue comes from the sell side, then the SSP is typically offering tools that favor the publisher rather than the advertiser.

Questions to ask before employing a SSPn What KPIs can you accommodate? n Which publishers contribute to your supply? n What is your price floor and how is that

price established?

n How much of your revenue comes from ad sellers versus ad buyers?

How do DSPs make money? DSPs collect a percentage (10-30%) of media spending and pay the exchanges, supply-side platforms (SSPs) and data providers. Often their fees are included as part of the cost of media, but some break it out as a separate line item.

Benefits of DSPsDSPs make it possible for advertisers to buy audiences on a one-to-one basis, rather than in bulk from a particular publisher’s website as was done in the past. They also facilitate buying from multiple inventory sources via one platform, thus making the process less complicated for the agency and client.

A DSP can constantly assess available inventory, refine campaign optimization models—both automated and manual—and make the best possible real-time bidding decisions. They allow for better price trans-parency and let buyers choose only their most desired impressions and then bid for those users dynamically.

Used well, a DSP can improve intel-ligence about customers, improve under-standing and engage more prospects, and offer potential to significantly boost market-ing performance.

Concerns around DSPsThere are complaints that some DSPs bury their costs in the price of media, which skews incentives for every vendor in the tech stack. But the counterargument to this is that percentage-based engagement gives incentive to the DSP to overpay for media and data. This can be manipulated by the DSP through their bidding strategies, algo-rithms and inventory curation practices.

All DSPs use the same third-party data and access the same inventory on the exchanges, and marketers are wary about the quality of exchange inventory. To ease some of those anxieties SSPs began enabling publishers to establish private exchanges, where inventory is only visible to agencies or DSPs that they have privately pre approved. DSPs can access premium inventory via private exchanges although the publisher must have already established its relation-ship with an SSP in order for this to hap-pen. But this is still not the panacea that was hoped for, because the private exchange system is still very small in scale.

Another concern is that DSPs some-times fall prey to click-bots and other fraudulent activity prevalent on exchange traded media.

Questions to ask before employing a DSP

n How many inventory sources are you integrated with?

n How quickly can I create a campaign using your platform? How quickly can I make changes?

n What kind of reporting do you provide? Is your reporting insightful and actionable?

n Is there an application program interface (API) that will pull reporting out of your system and integrate it with other systems?

n Do you have any managed services? What kind of support and training will my team receive?

n How do you integrate the data to meet my goals as a client?

n How will you provide insights and infor-mation on not only my presumed target audience but my potential audience?

n How will you help teach me to make smarter, more cost-effective decisions?

n How do you monitor whether a campaign is working? What kind of analytical methods do you use?

n What set of features do you offer? How do they compare to other DSP’s offerings?

n How much do you charge? n How do you select the quality of an

audience to bid on?

n How will you help me understand what drives marketing performance and who is interacting with my brand?

Supply-side PlatformsSupply refers to the seller of media—in most cases the publisher. Supply-side platforms are vehicles for publishers to make their inventory available for programmatic buying via an exchange type marketplace.

● They promise transparency and control over who can see their inventory and at what price. SSPs also act as exchanges, providing

Page 6: Navigating AdTech

6 NAVIGATING PLANET AD TECH MIT TECHNOLOGY REVIEW + DIGILANTPART 2: PRICING

buying of online advertising across the web, targeting specific audiences, often in real time. The term programmatic reflects the intended workflow improvements of automation. Prices may be determined by an algorithm, or pre-determined based on specified criteria.

Programmatic buying enables impres-sion by impression based buying as offered by automated marketplaces as opposed to the bundling of impression in CPM or thousands of impressions blocks. In order to achieve automated efficiencies the adver-tising units purchased by programmatic buyers are standardized with limitations on the variety of available units. By way of comparison to traditional buying, which takes place by negotiating directly with a publisher’s sales team in CPM blocks and can include non-standard or custom oppor-tunities, programmatic buying intends to be more efficient by offering standardized impressions. And if the data it is based on is right, that ad is more likely to be seen by someone who is more likely to find that brand, product, or service relevant because its impression driven rather than volume driven as traditional bulk buying is. Data-driven, programmatic advertising can be used when buying online display ads, mobile and video ads. The benefits are automation, workflow improvement and better results.

Programmatic selling This is the automated selling of a publish-er’s online inventory, usually ad space that was previously unsold; so-called “rem-nant” inventory ramped up for audience buying with the help of first and third party data. It usually takes place through RTB, but not always. Prices are usually less than for “direct sales” made between an adver-tiser and a publisher’s sales team. And although publishers can help drive up rev-enue from ad space that would have been left unfilled this way, some are wary of the system, out of fear that it drives down prices, introduces new middlemen, and cannibalizes their own inventory. There is also some use of private marketplaces in programmatic selling which, although automated, still allows for one-to-one communication between publisher and advertiser.

Cost per milleCost per mille (CPM) actually represents cost per thousand, which is the price an advertiser pays to reach a thousand viewers. In digital advertising CPM is calculated by dividing the cost of an advertising placement by the number of impressions (expressed in thousands) that it generates.

How CPM is determined in RTB RTB providers (trading desks and DSPs) determine CPM by available impressions, market volatility, target audience parame-ters and historical clearing costs of media.

Benefits of CPM pricingCPM pricing carries less risk for adver-tisers because it enables them to secure a flat rate for cost of media regardless of fluctuations in the cost of media for the RTB provider.

Concerns with CPM pricing Due to the dynamic nature of clearing costs, a fixed CPM does not always guaran-teed accurate value for buyers and sellers.

Dynamic cost per milleDynamic cost per mille (dCPM) provides advertisers with the exact clearing cost of media plus a technology fee, allowing for variable daily pricing adjustments.

How dCPM is determined in RTB The dCPM pricing model requires that RTB providers make media clearing

PricingPart Two

costs transparent to buyers daily. They also charge a transparent technology fee for every thousand impressions served in addition to the cost of media.

Benefits of dCPM pricing Advertisers can determine actual cost of media and pay a set technology fee for every thousand impressions served mak-ing it possible to adapt pricing daily and seasonally depending on market trends.

Concerns with dCPM pricing Dynamic pricing requires more hands on participation by the advertiser because they must spend more time tracking and monitoring daily advertising costs.

Defining ad tech basicsReal—Time—BiddingRTB enables a marketer to participate in an auction for the purchase of individual online ad impressions in “real time,” i.e. during the milliseconds that elapse while a user downloads a web page.

The idea is to provide marketers with the ability to target key audiences wherever they might be online for an appropriate price, based on a range of consumer attributes and their potential worth to an advertiser. The marketer can then place a bid and select the most suit-able advertisement for that particular targeted consumer. For the first time in the history of advertising, advertisers no longer have to rely on the publisher as a proxy for audience—the consumer can now be found and targeted with relevant advertising directly.

Programmatic buyingProgrammatic buying is the automated

Page 7: Navigating AdTech

7 NAVIGATING PLANET AD TECH MIT TECHNOLOGY REVIEW + DIGILANTPART 3: TACTICS

Concerns with audience targeting Audience targeting relies on data avail-able via the browser cookie, and algo-rithms that analyze the cookie data, neither of which is a good proxy for actual human behavior. Algorithms can also be easily fooled by click-bots or inac-curate cookie data so advertisers who deploy them should ask what precautions are made to avoid this activity.

Some marketers also argue that con-text is more important than audience targeting.

Contextual Targeting

.

Contextual targeting is the targeting of advertisements based on the classification of inventory.

Most often the classification of inventory is provided by the publisher themselves with little concern for precision. Publishers and automated inventory providers mark inventory in categories of broad demand or highly sought after segments.

Inventory may also be classified for context by the use of “spiders” that crawl web pages and analyze the content for the purpose of ad targeting. Content is classified by category or taxonomy, so advertisers can target pages that are more likely to contain material around a given subject or idea. Negative contextual targeting ensures for example that an air-line ad will not end up running adjacent to an article about a plane crash. Positive contextual targeting allows brokerage firms to run ads on pages with content relevant to stock trading.

Big data When it comes to ad tech, “big data” means taking all the voluminous points of information on consumer behavior and then crunching and analyzing that data in order to pinpoint a target audience, per-sonalize messaging, and, ultimately, boost sales. Data has been around for a long time. What is new is its massive volume and variety, the speed at which it is cap-tured, and the ability to store and make sense of it in real time. A key part big data is the ability to analyze information and build predictive models.

Big data can help facilitate real-time decision making through the analysis of factors like audience behavior and by pre-dicting response rates.

Algorithms Algorithms, the so-called “secret sauce” for technology giants like Google, are similarly the secret sauce for ad technol-ogy firms, driving how bids are made in programmatic buying of online ad inven-tory wherever it is found within display, search, social media, or other venues. In pure computer terms an algorithm is the automated code that makes sense of a group of variables and data in order to make a decision. In ad tech, an algorithm’s job is to make the most efficient, targeted, relevant and customized marketing deci-sion, by combining instructions on bud-get, marketing performance, goals, the market itself, together with factors like a user’s location, the time of day, informa-tion on consumer history, and third-party data, in order to make a bid for a single ad impression. However, some ad tech providers are pulling back the curtains in favor of providing full transparency and control to clients. These companies include DataXu with their algorithm mar-ketplace and Digilant with its custom val-uation technology known as BOSS.

Because of proprietary issues, the exact workings of a vendor’s algorithm and why it works can be tricky to ascer-tain. The algorithm is not just a machine-automated task, but something that needs to be tended and expertly tweaked by pro-fessionals in order to best leverage their results for a better and more relevant cus-tomer experience.

Audience Targeting

.

“Audience” is a euphemism for an advertiser’s target market, consumers or people.

Every advertiser is seeking to commu-nicate directly with audience prospects that are most likely to improve its busi-ness goals. Ad technology promises that ads can be delivered to the right audi-ence, at the right time.

As David Verklin, an early media agency player said: “No more diaper ads if you don’t have or are not expecting a baby. No more dog food ads if you don’t own a dog. And no more denture cream ads if you have a really good set of choppers.”

Benefits of audience targeting Audience targeting utilizes data to assess who is behind each impression and how valuable that impression is. An auto brand for example will value each of the follow-ing impressions differently: someone in the market for a car, a car enthusiast, and some-one who fits the target profile but is not car shopping at the moment.

Another benefit of ad technology is the ability to deliver dynamic creative content that reaches each audience —changing the message at different stages of the buying funnel for more per-sonalization and optimization to control the response.

TacticsPart Three

Page 8: Navigating AdTech

8 NAVIGATING PLANET AD TECH MIT TECHNOLOGY REVIEW + DIGILANTPART 4: BIG TOPICS

CookiesA cookie is a tiny text file planted by web sites on browsers to track what users are reading, buying and searching for online.

History of cookies Originally, Web pages did not allow stor-age of a user’s history or requests. Each page would load in isolation of the one that came before it. Netscape designed the cookie in 1995 to provide a trail to keep track of where a user was in the midst of a web application, allowing features like shopping carts or memory of a user pref-erences or login information.

First-party cookies This is information collected by a brand or an online publisher for its own use. In the case of Amazon, for example, such cookies store information about customer purchases and preferences.

Benefits of first-party cookies First-party data is considered especially valuable because it provides the most up-to-date and clear snapshot of a customer’s

Look-alike Targeting

This is the attempt to classify the behaviors and characteristics of people who have already shown interest in a product or service, and use probabilities to find new customers with similar behavior patterns who are also likely to be interested in the same products or services.

Benefits of look-alike targetingLook-alike targeting is a good way to reach out and prospect for new custom-ers who are likely to convert.

Concerns with look-alike targeting By definition, look-alike targeting is a tactic to reach prospective customers who may not be familiar with a brand or product, so it doesn’t necessarily result in immediate conversion. This is why it is often considered a higher funnel tactic.

Retargeting When a consumer visits a website a pixel often is triggered and a cookie is dropped on the user’s browser (typical browsers are Internet Explorer, Chrome and Firefox).

● The retargeting cookie enables ad technologists to track and analyze a user’s behaviors online so they can reach them with additional advertising message across the different online properties.

Benefits of retargetingIt works. Retargeting has been more success-ful in boosting clicks and transactions than any other digital advertising strategy thus far. It also provides for the best possible cost per acquisition, (CPA) because the prospect is already clearly interested in the brand. Retar-geting reliably produces measurable ROI for clients when using last touch/last click mea-surement, especially compared to display advertising overall, where success can be more difficult to trace.

Retargeting can be presented in dif-ferent ways, such as an additional ad impression, an offer for free shipping, or a discount with an expiration date.

Concerns with retargeting The success of retargeting is wholly depen-dent on the volume of traffic to a site. It may be effective in capturing “low hanging fruit” because those customers are already aware and therefore more likely to buy any-way. But it does not help marketers achieve their overall sales goals, because the pool of potential targets is small and it does not attract new business.

Big TopicsPart Four

It is also risky to repeatedly place the same ad in front of consumers and overexpose them to the same idea or product. Often consumers experience burnout or may start forming a neg-ative association with the brand, even getting feelings of being “creeped out” by the sensation they are being followed around the Internet.

Since retargeting reaches a limited audi-ence, it is also possible for marketers to over-expose users to the same messages when they engage too many retargeting providers on a plan. Negative brand perception could also be created if excessive retargeting con-tinues after the person has already made a related purchase.

behavior. First-party cookies enable pub-lishers and brands to personalize content.

Issues with first-party cookiesPersonalizing content based on the cookie trail limits the user experience and deters consumers from other paths that could lead to discovering other interests and making additional purchases.

Third-party cookies These cookies are used to follow the same browsers across the entire internet. Third-party cookies are placed by a party other than the site the user is currently visiting. When a user is detected visiting a pub-lisher’s site, the publisher sets the first-party cookie and the advertiser sets the third-party cookie.

Many other parties including ad networks, exchanges and DSPs, set third-party cookies.

Benefits of third-party cookiesThird-party cookies allow the marketers to track user behavior for retargeting and audience targeting purposes within the whole universe of websites.

Issues with third-party cookiesWhile users recognize the value of

Page 9: Navigating AdTech

9 NAVIGATING PLANET AD TECH MIT TECHNOLOGY REVIEW + DIGILANTPART 4: BIG TOPICS

action, including every touch point along the way. The challenge for a marketer is to track those touch points and determine how they ultimately contributed to a sale. Marketers need to use tools such as ad verification services to ensure that an impression was indeed delivered to the intended page. For branding managers especially, it has become increasingly important to focus on dwell time and “viewability,” that is, whether ads are actually seen and for how long, among the audience they have targeted. This also has technical limi-tations. Dwell time can be tracked, but viewability currently cannot.

The predominant approaches to viewability are to

n Sample domain viewability by sending a crawler that follows some of the ad impressions in order to determine a probability of viewability. This method does not measure actual viewability.

n Sample actual impression viewability by exploiting browser security gaps, and then extrapolating this sample on the domain/seller as a whole.

n Sample the probability of viewability by the propensity of a mouse to move over an ad.

None of these methods is an actual view-ability measure that is acceptable as an auditing function. They are—debatably—good as relative measures that can allow an ad buyer to filter/optimize towards viewability.

Concerns with attribution The user experience is fragmented among multiple devices, online and offline, on websites and social media. A purchase often occurs only after multiple site vis-its. Tracking multi-touch-point journeys and attributing them is challenging.

What to consider when choosing an attribution modelHow does my success rate change when I use a different attribution model, such as equal weight or a model that accounts for decay?

Connecting The DotsThe ultimate goal of every digital marketer is to get closer to the customer. Ad technology can help but there are points you should consider.

● Most importantly all must realize that the “Ad Tech” space is still in its infancy. The associated technologies are still in early stages of release and its business models are still evolving. Substantial prog-ress has been made towards shortening the gap between consumers and brands at scale but things are far from perfect.

The distinctions between roles are blurry, for example a DSP and an agency-run trading desk may perform essentially the same function – the management of a campaign and its media buying. A deci-sion about which to work with will prob-ably depend on the marketer’s desire for expertise, proximity to the technology provider, contractual structure, budget allocated to digital channels and knowledge.

Advertising Audit and Risk Manage-ment (AARM), a provider of independent advertising audit and consulting services, advises advertisers to ask many questions. These should include whether or not the trading desks benefit from rebates or dis-counts from publishers and others, and if those reductions are passed on to clients. AARM also suggests establishing up front guidelines to monitor campaign efficiency and effectiveness.

Agency Trading Desks offer the con-venience of being closely aligned with a brand’s agency although as previously discussed the ATD’s business model may cause marketer’s concern. Working directly with a DSP can provide more transpar-ency and control, and cut out the costs of media agency and the trading desk. They offer managed service contracts and some offer self-service contracts or technology licenses depending on how your market-ing team chooses to operate.

Meanwhile, the publishers, particu-larly top-tier ones, appear caught in the middle. RTB was pitched as a good

first-party cookies that hold items in their shopping cart and remember their login info, tracking by third parties is seen by many as an invasion of privacy.

Retargeting in particular is seen as an egregious practice that creates the impres-sion of being “chased” by ads. Countries in the EU have already passed legislation to ensure that users opt in to tracking cook-ies (fact check), and DNT (do not track) legislation is being discussed in the United States. There is also a higher likelihood of data leakage with third-party cookies than there is with first-party cookies.

Ad-tech privacy issuesPrivacy appears to be the Achilles heel of ad tech. The assumption has been that people accept the collection of data regarding their online behavior because it is not personally identifiable and enables them to receive more relevant advertising, but users are increasingly aware of the issue and beginning, in some cases, to push back.

AttributionAttribution is the ability to identify the effect of every single impression across every medium during the customer journey.

● When a brand employs a mix of TV, search, and digital display advertising, to which does it attribute each customer action? The most common attribution model used in digital advertising is “last touch/last click.” One hundred percent of the customer action is attributed to either the last ad clicked or the last ad served. This can be either a display ad, search ad or other type of digital ad.

But last click/last touch is an imper-fect measure. For example search tends to get a disproportionate amount of credit towards the end of the sales cycle, where display advertising has clearly played a role in creating interest for a particular product or brand. For this reason attribution has begun to focus on the more complete story of a cus-tomer’s journey towards a purchase or

Page 10: Navigating AdTech

10 NAVIGATING PLANET AD TECH MIT TECHNOLOGY REVIEW + DIGILANTACKNOWLEDGMENTS

This white paper was produced by MIT Technology Review Custom. Thank you to the following people who contributed their opinions and expertise:

Jeremy Hlavacek, VP, Strategy and Business Operations, WeatherFX at The Weather Channel (formerly at Varrick)

Jason Fairchild, Chief Revenue Officer, Open X

Ed Montes, CEO of Digilant

Joe Zawadzki, CEO, Media Math

Paul Alfieri, VP Marketing, Turn

Joshua Koran, Senior Vice President of Product Management, Turn

Adam Kasper, Chief Media Officer, Havas Media, North America

Sandro Catanzaro, VP Strategy, Dataxu

Omar Tawakol , CEO, BlueKai

Pascal Bensoussan, Chief Strategy Officer, Aggregate Knowledge

Brian O’Kelley, CEO, AppNexus

Ari Paparo, Senior Vice President of Media Prod-ucts, Baazarvoice

Kirk McDonald, President, Pubmatic

Duc Chau, Founder and SVP of Research and Development, Rubicom

Christina Beaumier, VP of Product Development, Xaxis

Perianne Grignon, Chief Marketing Officer, X plus One.

Matt Greitzer, Chief Operating Officer, Accordant

Arthur F. Muldoon, Jr. Co-founder & CEO, Accordant

Meredith Levien, Chief Revenue Officer, Forbes

Matt Prohaska, Programmatic Advertising Director, The New York Times / NYTimes.com

John Slade, Commercial Director, Digital Advertising, The Financial Times

Brian Morrissey, Editor-in-chief, Digiday

Scott Neville, Chief Marketing Officer, IPONWEB

Tom Hespos, President, Underscore Marketing.

Marc Goldberg, CEO, dailyRx (formerly CEO, About.com)

Alex Sutton III, Acquisition Manager, My M & M’s, Mars Retail Group

Debi Kleiman, President, MITX

Eric Hjerpe, Partner, Kepha Partners, VC Firm

About MIT Technology Review CustomBuilt on over 100 years of excellence in technology journalism, MIT Technology Review Custom is the arm of global media company MIT Technology Review that’s responsible for creation and distribution of custom content. Our expert staff develops meaningful and relevant content from concept to completion, distributing to users when and where they want it in digital, print, online, or in-person experiences. The turn-key solutions offer everything from writing and editing expertise to promotional support, and are customized to fit clients’ content marketing goals—positioning them as thought leaders aligned with the authority on technology that matters.

Copyright © 2013, MIT Technology Review. All Rights Reserved

deal for them, a “win-win” because it was thought that people would pay more for targeted ads. But that does not appear to be happening yet. Publishers are also evolving in the Ad Tech era learning to balance revenue achievement between programmatic and direct sales. However the oversupply of inventory makes this a challenging situation. The pendulum of power has swung a bit to favor market-ers who can take advantage of data and proprietary information to better value advertising inventory and react to mar-ket prices of an oversupplied system. Yet new tools and developments in the space developed for publishers are pushing back against the buyer’s incursion. Applications such as private deals and marketplaces empower publishers to create layers of scarcity and drive up prices. Both buyers and seller are also learning to make distinc-tions between “programmatic” efficiencies and auction or automated marketplace benefits.

Typically advertisers use a mixed approach to achieve their goals: a combi-nation of premium content directly bought from the publisher along with inventory auctioned through RTB.

Some publishers are seeing an oppor-tunity in programmatic buying; a chance to package segments themselves, like secur-ing rare high-net-worth-specific audi-ences by blending their own data with third-party data. Some UK publishers are already thinking this way, including The Daily Mail. The Financial Times is explor-ing the option and in the United States The New York Times and Conde Nast are offering similar services.

Meredith Levien, Forbes Media’s group publisher and chief revenue officer, argues that publishers can lead in both worlds; direct sales and programmatic buying. She says the intelligence gathered on the exchange about which consumers are buying what products can translate into smarter strategies by the publisher’s direct sales team.

Ultimately the move towards automa-tion in digital media is upon us, but it's up to the advertisers and publishers to deter-mine how to make it work for them.