Using data to guide product development
Mat Clayton
Co-Founder
What is Mixcloud?
The Plan
The Plan
What does failure look like?
turntable.fm - MAU
Text
turntable.fm - DAU
Competitor analysis
• Sitemaps are an amazing resource
• Google - “site:competitor.com“
• OpenSiteExplorer, can be used to find links and any growth
tactics being used
• graph.facebook.com/<app_id> can give you monthly Facebook
actives for FB connected apps
• API’s
Key Performance Indicators
(KPI’s)
Pick a single KPI
• Easily understood
• Provides focus and align the team
• Represent the core of the product, and usually isn’t uniques
Mixcloud - Listener minutes / 28 days
AirBnb - Nights stayed / 28 days
Eventbrite - Tickets Booked / 28 days
SendGrid - Mail Sent / 28 days
Dashboards
• Distribute information amongst the team
• Make trends observable
• Give the team a solid understanding in
top level metrics
• We’re big fans of graphite and grafana
• Alert on changes
User Accounting
User accounting
Site A Site B
Active Users 1000 1000
Born +800 +100
Die -100 0
Sleep -300 -50
Awaken +100 +450
Next Period’s Active Users 1500 1500
Retention Curves
Identifying key
product features for retention
The Brainstorm
What correlates with retention?
First week Listens
Key User Metrics - Twitter
Twitter’s data
Correlation does not imply Causation
Cause or effect?
How do you choose
what to build?
What facts do we know about our users?
• What information did they give us? name, email, Facebook?
• Do they use mobile web/desktop/iOS/Android?
• Do they have email notifications enabled?
• Which core features have they used and to what degree?
• Which features did they miss?
• What was their last interaction with the service?
• Where did they come from? (Referrer)
• Do they use competitor products?
Evaluating potential product features
• Do we have the resources to build it?
• How many users will the feature be relevant to?
• How frequently will they interact with it?
• What is the conversion rate against our primary goal?
AB Testing
(Always Be Testing)
AB Testing Examples
Big changes == Big results
How big is big?
AB testing
• Continually run AATests
• Requirements
- Fast to implement, a single line of code
- Negligible cost to product performance, including SEO
- Results need to be available to the entire team
• Tools
- Django Experiments (we wrote it)
- A/Bingo (Ruby on Rails)
- Optimizely
- Google Website Optimizer
Failure is acceptable
“A good plan, violently executed now, is better than
a perfect plan next week.” - George S. Patton
Thanks for listening!
Any Questions?
@matclayton
mat@mixcloud.com

Using data to guide product development