Upload
masstlc
View
318
Download
0
Embed Size (px)
Citation preview
05/03/2023
Henry Cipolla
Data Driven Development
2
Title Clickbait
305/03/2023
Agenda: Define Data Drive Development Explore DDD in Product Management Explore DDD in Engineering Implementation
Last chance to leave!
405/03/2023
Good buzzword Means different things to different people
For the next 40 minutes it will be:
The grown up version of the LEAN
movement
Defining Data Driven Development
505/03/2023
Picking the right metrics and measuring [your output] against them to get to [your goal] in order to build better products.
Specifically:
605/03/2023
Product Manager
Output = Customer Experience
Goal = Delight
Engineer
Output = Product Code
Goal = Velocity and Quality
Again, different things to different people
DDD for PMs
05/03/2023
805/03/2023
KPIs are too high level– Users, revenue
Conflicting success metrics– Longer sessions, use of utility functions
Poor instrumentation– No chart instantly shows if goal was achieved
Limited AB Testing– AB is only used on the web for content testing
PM: Signs that you are not actually data driven
905/03/2023
1. Define success of this feature / story– Not just usage. Consider NPS, retention, and other behavior
2. Define the chart in your analytics tool that tracks success– Usually this means adding tagging to the story
3. Define the variations to test– Don’t test colors and positions. And do test expected failures
4. Revisit the results frequently
Recommended DDD approach
1005/03/2023
Onboarding Experience– First session length, sessions until successful task
Meal calorie estimator– Pounds of weight lost, meals entered per customer
Music shuffler– New exposures per listener, retention against control
Example feature success metrics
1105/03/2023
Relevant example of a descriptive chart
1205/03/2023
Ask a user how likely they are to refer the app on a scale of 1 to 10
Look at what features and attributes users share on either end
Thank the user by reacting to the feedback. If they score low, ask why. If high, have them rate you
Example NPS Survey
DDD for Engineers
05/03/2023
1405/03/2023
Dashboards only exist to impress investors & scare sales Features are not revisited after they are released Only bad retrospectives happen (postmortems) Eng process is not measured and tested like the product Velocity might be tracked but it isn’t improving and output
doesn’t match expectations
ENG: Signs that you are not data driven
1505/03/2023
Engineering has its own KPIs– Velocity and quality
“Product reflects the org”– Measuring the process can predict product troubles
Small time frames create easy measuring points– There is a reason LEAN and AGILE tend to come together
Your process is a product – be data driven
1605/03/2023
ENPS in disguise with some extras After every sprint (2 weeks) we ask engineers to
anonymously(ish) provide 3 1-5 ratings for:– The company
– Their crew / team
– Their work
Our approach: “Awesomeness”
1705/03/2023
The results are aggregated to describe how things are going
Provides insight into things you wouldn’t expect, like the efficacy of morale events
Makes a great dashboard
Overall Awesomeness
Implementation
05/03/2023
19
1. Business KPIs – track them visibly
2. Define success metrics per-feature, and track them - this will often require platform-specific tools
3. Expose these metrics – daily emails work great
4. Have a retrospective on feature launch. Did the feature hit its goals? Were multiple versions launched – who won?
5. Do this for every feature, frontend and backend as well as every team
How to get started
2005/03/2023
PMs feel more confident experimenting because there is a framework for measuring these experiments
Tradeoff conversations between feature and teams are easier because everybody speaks the same language
Problems are found faster because there are more opportunities for their discovery
Organizations scale easier because there is measurement on all efforts
Benefit of it all
Q&A ?
05/03/2023