Product analytics
Know who uses your product, what they do first, where they give up and what brings them back. We set up the tracking, read the behaviour and help your product team decide what to build or fix next.
Signs you could use this
- You ship features and can't say afterwards whether they worked.
- Your funnel has a big drop and you're guessing why.
- Event names in your analytics tool no longer make sense to anyone.
- Product, marketing and engineering each report different user counts.
What we do
- Get the tracking right
- Good analysis needs good events. We review what you collect, agree naming rules and fill the gaps, so the data tells the whole journey and not half of it.
- Map the journey
- We trace how new users move from sign-up to their first useful moment, and where they leave. This is usually where the clearest improvements sit.
- Look at retention and cohorts
- Who comes back after a week, a month, three months, and what did they do differently in their first session?
- Measure feature impact
- Before and after comparisons, or proper experiments where they are possible, so a launch is judged on evidence.
What you receive
A first round, covering the tracking review and a core set of product views, normally takes three to five weeks.
- A tracking plan and the fixes to put it in place
- Funnel, retention and cohort views for your product
- A prioritised list of friction points with supporting data
- A repeatable way to evaluate new features
Tools we use: Mixpanel, Google Analytics, Adobe Analytics, BigQuery, AWS Redshift, SQL, Python.
Common questions
We already use Mixpanel or Google Analytics. Can you work with that?
Yes. Most of our product work starts from an existing set-up. We audit it, fix what's broken and add what's missing.
Can you analyse raw event data in our warehouse?
Yes. For deeper questions we usually go to the raw events in BigQuery, Redshift or PostgreSQL, where there is more freedom than in a point-and-click tool.
How much traffic do we need?
Less than people think for journey and funnel work. Experiments need more, and we'll tell you honestly when your volume is too low to test something.