A weekly review is a chance to turn scattered signals into a small number of decisions. Connected tools can make those signals easier to find, but a dashboard cannot tell you on its own why a customer acted or what to build next.

The temptation with a weekly review is to make it comprehensive. Resist it. A review that gathers everything produces a feeling of being informed and a list of nothing in particular, and the two are easy to confuse.

Choose one outcome from your plan: activation, reliability, retention, or another measure that matters now. Write down the target and what you expected to happen. This makes it easier to distinguish a useful signal from a busy week of activity.

One outcome, not four. A review with four questions is a dashboard, and dashboards are for people who are not deciding anything today.

The current Founder2 integrations can retrieve snapshots from PostHog, Sentry and a configured Slack channel. PostHog data includes activity and event summaries; Sentry provides issue context; Slack provides channel context. Fields, access and completeness vary by provider. Check the sync time and go to the source when the decision needs more detail.

For a fictional example, a founder may notice fewer active users and an increase in errors during the same week. Those observations suggest a question, not a proven cause. Check whether the affected users and time periods overlap, and look for alternative explanations before deciding what to fix.

That check is the whole practice. Two lines going in the same direction is a coincidence until you have looked at the actual records, and the cost of acting on a coincidence is a week of building the wrong thing.

What a weekly review can pull in today.
  • PostHogActivity and event summaries.
  • SentryIssue context and error frequency.
  • SlackContext from a configured channel.
  • GitHubOpen work and recent activity.

Each source returns a snapshot. Freshness and completeness vary by provider.

Write the observations down as sentences that cannot be disputed: "1,240 active users, down from 1,410 last week", "38 unresolved errors, up from 12". Then write the hypotheses separately: "the new onboarding step is dropping people", "a release on Tuesday broke the checkout path".

The two lists look similar and behave differently. An observation can be checked by opening a tool. A hypothesis can only be tested. Putting them in one list is how a hypothesis becomes a finding without anyone noticing the moment it happened.

Write one finding, its source, what remains unknown and the next test. Give the task an owner and a date for review. If the evidence is weak, say so. If the information does not change a decision, it may not belong in the weekly review.

Then do the last part, which is the part that gets skipped: update the plan. A review that produces a decision but leaves the plan untouched has not changed anything except your mood. The plan is the document that the next week reads, and it is the only reason a review is cumulative rather than a weekly slot in the calendar.

Continue with the GitHub workflow to turn a chosen next step into a reviewable issue, and the Stripe guide for the revenue side of the same review.