The 90‑Day Post‑Launch ROI Retrospective: Exact Metrics, Dashboards & Exportable Report

Start with three outcome tiers: Acquisition quality, Activation & Early Retention, and Unit Economics. Each tier maps to 2–3 KPIs you must report at D30/D60/D90. Acquisition quality = new users by campaign + new user activation rate. Activation & Early Retention = Day‑1, Day‑7, Day‑30 retention and core activation funnel completion. Unit Economics = CAC (by channel), LTV (cohort-based), and CAC payback period.

Operationalize single definitions for each KPI and lock them before launch. For example, define activation as “completed core event X within 7 days of sign-up,” and retention as “at least one core event in the 7‑day window ending on measurement day.” These consistent definitions prevent metric drift when teams chase improvements.

Use calendar-based cohort starts (month/day) or first-event cohorts with a fixed lookback window. For a post‑launch retrospective we recommend these windows: D0–D1 (activation check), D1–D7 (early retention window), D8–D30 (habit formation), and D31–D90 (monetization/expansion signals). These windows reveal whether early churn is product fit or acquisition noise.

Prefer strict calendar cohorts for comparability (e.g., cohort = signups on March 4) and 24‑hour rolling windows only when you need live dashboards. Document how you treat timezones and late-attribution to avoid false churn (a common pitfall when product behavior crosses midnight boundaries).

Build five canonical charts and expose them on a single 30/60/90 retrospective dashboard: (A) New users by acquisition channel (daily + cumulative), (B) Activation funnel conversion (signup → core event within 7 days), (C) Cohort retention table (D1/D7/D30 with cohort size), (D) Revenue by cohort and 90‑day LTV curve, (E) CAC & payback curve by channel. For SQL, join user table to event table and compute cohort_start_date = DATE(first_event_ts AT TIME ZONE user_tz).

Include confidence columns in the dashboard: cohort size, sample variance for conversion rates, and a flag when cohort n < 50 (treat small cohorts as noisy). Make these charts exportable (CSV or PDF) so the executive one‑page can be generated automatically after each update.

When multiple experiments or campaigns run during the 90‑day window, pick one primary attribution rule and enforce it across all KPIs. Common choices: first-touch, last-touch, or probabilistic multi-touch. For product experiments (A/B tests) use randomized assignment and analyze effect on activation and D7 retention within the cohort that experienced the variant.

Watch for confounders: marketing pushes that inflate D0 installs but don’t change activation, instrumentation changes that alter event names, and seasonality. Always run a ‘sanity check’ — compare matched cohorts (same acquisition channel, same week) to isolate experiment impact. If a metric moves only in small cohorts or during a promotional spike, tag it as provisional and require a holdout validation before scaling spend.

The one‑page must answer three investor-level questions: (1) Are users finding value? (activation + D30 retention), (2) Can we acquire users at an economical rate? (channel CAC, payback), and (3) What is the recommended action? Use a top strip with the decision (Invest / Iterate / Kill), followed by three columns: Key numbers (headline KPIs), Evidence (two charts: retention curve and LTV vs CAC), and Action plan (3 next experiments, owner, timeline).

Provide an automatic export: a PDF or CSV generated from your dashboard that contains the headline KPIs, cohort table, and a short‑form narrative. AppWispr customers and product teams should add an ‘owner annotation’ to each metric so responsibility and context travel with the report — this reduces debate and accelerates investment decisions.