Cohort Profitability Engine
Understand when your cohorts break even and how to get there faster — live chart, sensitivity, and scenario-linked KPIs (this page). Stable calculator: index2.html.
Cohort matrix: each row = install cohort (optional Week 0 paying subs only). One row → CAC, LTV, break-even are for that cohort only. Two+ rows → portfolio (weighted): total CAC = sum of rows; subs/affiliate use total matrix users. Advertising assumptions are defined in Advertising revenue. Fallback users/CPI apply only if a row is incomplete.
Inputs are saved automatically in this browser (local storage) and restored when you reopen this file.
Advertising revenue
Plan escalators (% / calendar year)
Year 1 matches inputs above; later plan years multiply each calendar year by (1 + %⁄100).
Retention
Retention (share of cohort still "active" for ads): week 0 — 60%, week 1 — 30%, weeks 2–7 — 20%, weeks 8–10 — 15%, weeks 11–51 — 10%, week 52+ — 5%.
Business model (optional)
One row per year (auto-generated from Business horizon).
Annual cost inflation (% / calendar year)
Applies each calendar plan year on top of base yearly amounts above (Year 1 unchanged).
LTV/CAC keeps using acquisition CAC logic. Business P&L simulates continuous weekly acquisition from the annual plan (users/year), so years 2–5 include ongoing users plus continuing revenue from prior cohorts.
Affiliate
Simple is the original single conversion % and revenue shape. Categories models four affiliate types and blends them into one affiliate cash stream (same chart column and KPIs as today).
e.g. 2 means 2% of cohort users generate affiliate revenue.
Recurring: total = converts × $ × (projection weeks ÷ 52), flat each week. One-time: total = converts × $ once, spread evenly across the first 52 life-weeks (or fewer if the projection is shorter); no affiliate revenue after that.
Subscriptions
Cash mode: full price when a bill is due. Paid conversion timing (of eventual subscribers): week 0 — 40%, week 1 — 20%, weeks 2–4 — 10% each, weeks 5–6 — 5% each (renormalized if the projection ends before week 7). Within each arrival week, the monthly/yearly split applies. Billing: monthly every 4 life-weeks from first payment; yearly every 52 life-weeks. Renewal rates apply per billing cycle.
Pricing
Conversion
Single cohort row only: if set, this count is treated as 40% of eventual paying subs (same timing as the note above); total paid = count ÷ 0.4 (capped at row users). By default the model uses Total paid conversion % and ignores matrix week-0 subs unless you enable the toggle below.
Renewal (per billing cycle)
Ads — cohort impressions matrix
Week 0 only: enter Week 0 impressions (ad baseline for the cohort). Optional Week 0 paying subs counts paying subscribers in life-week 0; the model treats that count as 40% of eventual paying subs and fills weeks 1–6 using the same curve as in Subscriptions (renormalized if the projection is shorter). Later life-weeks use retention on the Week 0 impression baseline for ads. Add/remove cohort rows. CPI per row: leave blank to inherit the first row’s CPI, or the default CPI above if the first row is also blank.
What-if scenarios (optional)
Users stay fixed for all columns. Week 1+ ad impressions still follow the retention curve from the Week 0 baseline (matrix or override). Override Week 0 ad impressions, CPI, projection length, subscriptions (including optional Week 0 paying subs count), affiliate, and ad CPM below. Leave a field empty to use the same value as the baseline run (main form above).
Scenario A
Subscription overrides (empty = baseline). Shares are scaled to 100% when needed.
Scenario B
Subscription overrides (empty = baseline). Shares scaled to 100% when needed.
Target mode
Set goals for LTV/CAC and payback timing. After Calculate, a status card next to the summary shows whether the baseline run meets both targets. Payback uses the same life-week index as the model; months follow ~4 life-weeks per month (same as Payback month in the summary).
Sensitivity
Stress-test before scenario overrides: these adjustments apply to the shared model state, then Baseline / A / B are recomputed. Use for quick “what moves payback fastest?” exploration.