3. Themes in retail
● The bar for personalization is rising
● Competing on anything other than price and logistics
requires offering something more
4. Themes in retail
● The bar for personalization is rising
● Competing on anything other than price and logistics
requires offering something more
● Data science makes it possible to know your clients at
scale
5. Knowing your clients
Three lessons from Stitch Fix
Lesson 1: Feedback loops unlock personalization
Lesson 2: Client incentives matter
Lesson 3: Data (science) enables personalization at scale
21. You need the data!
Personalization depends on knowing your clients
22. Broken feedback loops
Why don’t traditional retailers have this data?
● They don’t know they need it?
● Their clients have no good reason to give it to them
Recall, the bar is high. You want to know every client!
23. Broken feedback loops
?
Weak or missing altogether
Unclear what the customer thinks about what they bought, much less others items
https://openclipart.org/
24. You need the data!
The best way to get data from your clients is for them to want to give it you!
Compelling self-interest drives feedback loops
● First order benefit (very compelling): your experience gets better!
● Second order benefit (less compelling): your feedback helps Stitch Fix /
other clients
25. You need the data!
Reduces the need for “out of channel” feedback
Survey Enter to
win!
https://openclipart.org/
27. Lesson 3: Data (science) enables
personalization at scale
28. A long history of personalization
In some ways, personalization is old-fashioned
http://cliparts101.com/
29. Scaling personalization
But how do you scale this?
And, we want more!
● Uniformly high quality
● Virtuous cycles and feedback
● Iterability
http://cliparts101.com/
30. Technology and data science
Having the data is not enough - data science is required to bring it to life
Algorithms are
● iterable
● testable
● replicable
http://openclassroom.stanford.edu/
31. Technology and data science
Having the data is not enough - data science is required to bring it to life
Algorithms are
● iterable
● testable
● replicable
A / B
32. Having the data is not enough - data science is required to bring it to life
Algorithms are
● iterable
● testable
● replicable
Technology and data science
Algorithmic
recommendations
35. Let the data scientists out of the lab!
Using data to run a business
What are we optimizing?
How should a business make tradeoffs?
Making decisions in the presence of uncertainty
How do I tell if this is working?
https://www.evanmiller.org/bayesian-ab-testing.html