10 ways to win at data analytics
TRANSCRIPT
2
Scale of
Approach
KPIs
10 Ways to WinAt Data Analytics
Industry Models
Data Lakes
Data Democratisation
Data
Science
Attribution
Customer Centricity
Retrofitting
YEARS OF EXPERIENCEBupa, Canon, Dyson, Hotel Chocolat, John Lewis, MTV, Tesco Bank, Waitrose…
COMMON CHALLENGESAcross technology, people and process
101010
PRACTICAL SOLUTIONSWhat in our experience works and doesn’t work
4
Opening ThoughtData Analytics Strategy
There is rather a lot going on in data analytics
Realistically no business can resource everything with equal priority
A lot of this is about staying focused and actively choosing not to do things
5
Hot Topics
Engineering4. Aggregate vs Raw Data5. Retrofitting Measurement6. (Too Many) Tools
Strategy1. Satisfying Thirst for
Data2. Smart KPIs
3. Cross Device/Channel
Analysis7. Attribution8. Customer Centricity9. Data Science10. Data Lakes
9
1. Thirst for Data?
1. Not everyone in an organisation is naturally thrilled by data
2. Some people just want answers3. Some people need help framing the
question4. Democratisation of data needs to be matched
by a process and dialogue for framing and answering questions
10
2. Smart KPIsTip 1: Steal from other industries
Retail
Category -> Product
Product -> Basket
Basket -> Checkout
Financial Services
Lead Product -> Up-Sell
Net Present Value
Logins per Month
11
2. Smart KPIs
Short Term ROI Long Term
LTV
Tip 2: Watch out for “going lean” on an inversely correlated KPI
12
3. Cross Device/Channel
• There is no Single Customer View
• There may be Substantial Customer Views
• All comes down to mapping opportunities to (re-)identify an individual
The Cross-Device/Cross-Platform Challenge
Email sent to [email protected]
Hashed email embedded in click-through URL:4279c1158bd90b61e60bb0a5d461f8bf
Hash 4279c1158bd90b61e60bb0a5d461f8bf becomes userid and passed to analytics tool
13
4. Data Processing: Scale of Approach
• Great for reporting• Standardised
metrics/approachesAggregate
Data• Great for basic customer
analytics• Requires database skills to use
Relational Data
• Great for modelling• Easily corrupted in translation!Raw Data
14
5. Retrofitting Measurement
• Two Harsh Truths– If measurement is an afterthought, the project has already
failed– If measurement can be deprioritised, the project has already
been written off in terms of success
• Measurement Design must be a core part of any project
6. (Too Many) Tools
Key Disciplines
Business case should drive technology need (always). Quickly becomes a false economy to have lots of tactical tools.
Unavoidable overlap in the functionality of different tools (e.g. CMS vs Personalisation). Beware of using two tools to do the same job
Be strict in identifying impact on knowledge spread across an organisation – will it get too thin if another tactical tool is added to the mix?
17
7. Attribution
Something we instinctively know as marketers:
Different channels behave differently
18
7. Attribution
Something we instinctively know as customers:
We naturally behave differently depending on what our first touchpoint was (e.g. email versus search)
21
8. Customer Centricity
Customer
Loads of A/B Testing
Behavioural Personalisatio
n
NPS Survey vs
Customer Analytics (e.g. Segmentation
Modelling)
Better Understanding of the Customer
Building better experiences for key
segments
22
Treating Segments Differently
8. Customer Centricity
No Personalisati
on Whatsoever
01
One-to-One Omnichanne
l Personalisat
ion02 03
24
9. Data Science
http://drewconway.com/zia/2013/3/26/the-data-science-venn-diagram
Winning Traits
1. Ability to continually interpret and communicate statistics in the context of the business
2. Ability to translate business requirements into (real) technology requirements
25
10. Data Lakes
Data Warehouse
ETLFinance
Telephony
EPOS
Data Lake
Finance
Telephony
EPOS
Clickstream
Ad Serving
27
10. Data Lakes
GovernanceClear Data DefinitionsDocumented SchemaDocumented ConsentOwnership of Data Quality
28
10 Ways to Win
1. Apply 80/20 rule for importance of framing questions versus answering them.2. Steal/adopt KPIs from other industries, but don’t “go lean” on inversely correlated
ones (CPA/LTV)3. Look out for opportunities to identify and re-identify users to build more substantial
customer views.4. Measurement Design must be a core part of any project. Full Stop.5. Use aggregate data for reporting, raw data for modelling.6. Have a skills and resourcing plan for any new tool and make sure you’re not spread
too thinly.7. Beware any “one size fits all” attribution model. It probably won’t fit.8. Don’t forget Customer Analytics for Customer Centricity!9. Look for key people that can bridge statistics and business, business and technology.10. Invest in governance to prevent your data lake turning toxic.
Thank [email protected]@lynchpinhttps://uk.linkedin.com/in/andrewhood