How to Build a Pendo Feature Adoption Report That Drives a Decision
A useful Pendo feature adoption report has four numbers: the share of accounts that touched a feature, the share that used it twice or more, the retention gap between those groups, and the trend over the last 90 days. Everything else is decoration.
Why do most adoption dashboards fail?
Because they report usage without a comparison. “42% of accounts used Reports” is a fact, not a finding. It becomes a finding when you put it next to something: the same figure last quarter, a peer feature, or the retention of accounts that never touched it.
Pendo makes it easy to build tall dashboards. The discipline is building short ones.
What four metrics belong on the report?
| Metric | Definition | Why it earns a slot |
|---|---|---|
| Reach | Accounts with ≥1 event on the tagged feature in period | Distinguishes marketing from reality |
| Repeat use | Accounts with ≥2 events, or ≥2 distinct days | Single-use is curiosity, not adoption |
| Retention delta | 90-day retention, users vs non-users of the feature | The only metric that changes roadmap priorities |
| Trend | Same figures for the prior two 90-day windows | Turns a snapshot into a signal |
Everything else — total events, average clicks per session, time in feature — lives one level down. If someone asks for those, they are usually trying to justify a decision already made.
How do you define “active use” without lying to yourself?
Pick one definition and write it down. Two events on two different days is the most defensible default for B2B features that people would otherwise open once out of curiosity.
Avoid counting page views as feature use. A user who lands on the Reports page and bounces has not adopted Reports, and Pendo will happily count the page view for you.
For guidance-style features, count the downstream action instead. Did the account create the thing the feature exists to create? That drops your adoption rate by half and makes the number honest.
How do you compare adoption with retention?
Cohort by feature use, then measure retention at 30, 60 and 90 days. Pendo’s retention reports handle this if the feature is tagged; if not, export the account list and join it in a warehouse.
The result is usually a large gap — that is expected, because engaged accounts use more features rather than features causing engagement. To get closer to causation, look at accounts created in the same week with similar seat counts, then compare. It is not a randomised trial, but it stops the most obvious confound.
Read the gap as a hypothesis, not a causal claim. A feature with a 20-point retention delta might be genuinely valuable, or might simply be something only committed accounts bother to learn.
How often should you rebuild the report?
Once a quarter, and no more. Weekly rebuilds chase noise. Set the cohort window, freeze the definition, and only change it when the product changes enough that the definition stops meaning what it did.
If PostHog or Amplitude already holds your event data, run the same four metrics there and use Pendo for the tags and guides. Duplicating dashboards across tools is how teams end up with two different adoption numbers in one meeting.
Who should not read this report?
Sales, without context. An adoption figure travels badly into a deck — it becomes “customers aren’t using X” when the real finding is “X is a power feature used by 40% of accounts that never churn”. Share adoption reports with product and CS, and share a narrative with everyone else.
Teams under 500 MAU: the numbers are too small to cohort. Wait.
What does a good first version look like?
One page. Four numbers with 90-day trends. One line of interpretation. A tagged feature list of five to eight items, not forty. That version takes about two hours in Pendo and answers the question leadership actually asks: what should we build next, and what should we stop maintaining.
Key Takeaways
- Four metrics are enough: reach, repeat use, retention delta and 90-day trend.
- Define repeat use explicitly — two events on two different days is a defensible default.
- Never count page views as feature adoption; count downstream actions instead.
- A retention delta is a hypothesis, not proof of causation.
- Rebuild quarterly, not weekly, and keep the tagged feature list under ten items.
