Get a weekly product usage digest
Read your PostHog data and post a plain-language digest of what changed and why.
A product usage report should answer "what changed and should we care?", and most dashboards answer neither. This workflow reads your PostHog data every week and posts a plain-language digest to Slack: which metrics moved, which features are gaining or stalling, where new users drop off, and what's different from last week.
What does this workflow do?
The agent pulls your key metrics from PostHog: activation, retention, feature adoption, funnel conversion, whatever you've defined as the numbers that matter. It compares against previous weeks and writes the story of the week in a few short paragraphs, with the charts' conclusions in words.
It goes past summarizing. When a metric moves, the agent looks for the why in the data and in your other tools: signups dipped the week a pricing change shipped (it knows from Linear), a feature's usage jumped after it appeared in onboarding (it knows from your changelog). Correlation gets flagged as correlation, and when the data can't explain a move, the digest says so and asks the product owner rather than inventing a narrative.
How does it work?
- Define the metrics once. Point the workflow at your PostHog insights and dashboards, or describe what you care about and let it propose the set.
- It reads the week's data. Metrics, trends, funnels and cohort movement, compared against the previous weeks.
- It looks for explanations. Cross-referencing what shipped from Linear and GitHub, what changed in onboarding, and known seasonality from its own memory.
- It writes the digest. Five to ten sentences of what happened, a short list of movements with numbers, and one or two suggested questions worth digging into.
- It posts and archives. To your product channel in Slack, with the full version in Notion. Anyone can ask follow-up questions in the thread.
Why a written digest when dashboards exist?
Dashboards answer questions people remember to ask. A weekly narrative puts the answer in front of the team without the asking, which is how non-analysts actually stay informed. The digest links back to the underlying PostHog insights for anyone who wants the raw view.
Memory does the calibrating. The workflow learns what your team considers signal ("we don't care about weekend dips"), keeps metric definitions stable, and tracks which past explanations turned out right, so its reads get more trustworthy over time.
Works with
PostHog, Slack, Notion, Linear. Runs weekly; a daily variant works for launch weeks.
Frequently asked questions
What do I need to set up in PostHog first?
Working event tracking and ideally your core insights or dashboards. The workflow reads what exists. If your tracking has gaps, the digest will tell you which questions it can't answer, which is a useful gap list in itself.
Can it explain why a metric moved?
It looks for candidate explanations in your data and your other tools, and it labels them as hypotheses with the evidence. When it can't explain a move, it says so and can ask the feature's owner.
Can non-technical people ask follow-up questions?
Yes. Reply to the digest in Slack ("which plans drove the retention dip?") and the agent answers from the data, in plain language, with a link to the underlying insight.
Does it replace our analyst or PostHog dashboards?
No. It replaces the weekly assembly and narration work. Deep analysis stays with humans, who now start from a shared, current picture.
Can it feed other reports?
Yes. The investor update and weekly status report templates can pull from the same metric definitions, so every report tells the same story with the same numbers.