Products That Learn After Release

Resources

• The HEART framework — Google’s original paper connects user-centered goals to signals and metrics across happiness, engagement, adoption, retention, and task success. Helps compare intended and actual outcomes after release rather than mistaking activity or adoption alone for lasting value. https://research.google/pubs/measuring-the-user-experience-on-a-large-scale-user-centered-metrics-for-web-applications/

• PostHog product analytics — Official documentation introduces behavioral analysis through trends, funnels, retention, and paths. Useful for locating drop-off and segment-specific regressions, then forming questions for investigation rather than reading causal explanations directly from aggregate charts. https://posthog.com/docs/product-analytics

• PostHog session replay — Official documentation for inspecting recorded product interactions and their debugging context. Complements analytics with observed friction, repeated actions, and workarounds; recordings show what happened but still need customer context to explain why. https://posthog.com/docs/session-replay

• In-depth user interviews — The GOV.UK Service Manual explains planning and conducting interviews about users’ circumstances and experiences. Helps connect support themes and customer language to concrete post-release behavior without substituting leading questions or agent-generated personas for real evidence. https://www.gov.uk/service-manual/user-research/using-in-depth-interviews

• Analyse a research session — GOV.UK separates observations, grouping, findings, and actions. A practical synthesis method for clustering support and interview evidence while keeping contradictions and original observations visible instead of letting an AI summary erase inconvenient cases. https://www.gov.uk/service-manual/user-research/analyse-a-research-session

• Postmortem culture: learning from failure — Google’s SRE book describes blameless incident analysis and actionable follow-up. Connects logs, timelines, contributing causes, and customer impact to product learning, replacing an unsupported single-cause story with evidence and owned corrective work. https://sre.google/sre-book/postmortem-culture/

• Postmortem analysis across incidents — The SRE Workbook extends learning from individual incidents to patterns across a collection of postmortems. Useful for recognizing recurring regressions and systemic causes before deciding whether to patch one feature, simplify a subsystem, or investigate further. https://sre.google/workbook/postmortem-analysis/

• PostHog experiments — Official documentation for controlled experiments and interpreting results. Supports comparing a change with a control before expanding or reverting it, with attention to uncertainty rather than treating a favorable movement in a dashboard as proof. https://posthog.com/docs/experiments

• Build–Measure–Learn — Eric Ries’s Lean Startup principles frame validated learning and the pivot-or-persevere decision. Use the loop to consider whether evidence warrants expanding, revising, simplifying, investigating, or stopping, instead of automatically turning every signal into another feature. https://theleanstartup.com/principles

• Evaluation best practices — OpenAI’s guidance describes task-specific evaluations, production examples, edge cases, and continuous evaluation. Provides the bridge from observed failures and customer reports back into versioned eval cases so the next agent-driven change is checked against what production taught the team. https://developers.openai.com/api/docs/guides/evaluation-best-practices

• Creating actionable GitHub issues — GitHub’s official issue guide provides the mechanics for capturing and organizing follow-up work. Use an issue as the durable handoff from a production signal to a revised specification, linking the evidence, intended behavior, acceptance checks, and unresolved questions rather than forwarding only a generated summary. https://docs.github.com/en/issues/tracking-your-work-with-issues/using-issues/creating-an-issue

• How Claude remembers your project — Claude Code documents project instructions, scoped rules, memory, and how to inspect or edit what is retained. A concrete destination for reviewed product learning: update agent context and decision records without promoting uncertain observations into permanent facts or leaving stale guidance unchallenged. https://code.claude.com/docs/en/memory