Exact arithmetic. Unproven forecasts.
The arithmetic is exact and checkable. Any forecast built on it is only as good as your inputs, and those inputs are assumptions, not measured facts.
Arithmetic Accuracy: Checkable
The same inputs always give the same outputs. You can verify the core by hand with one cohort:
| Day | Retention | Active users | Revenue that day |
|---|---|---|---|
| 0 | 100% | 2,500 | $212.50 |
| 1 | 46% | 1,150 | $97.75 |
| 7 | 33% | 825 | $70.13 |
| 30 | 25% | 625 | $53.13 |
One cohort: 2,500 free launch installs, no paid UA, $0.085 ARPDAU, the default retention curve, every realism setting switched off. Active users × $0.085 = revenue.
With the starting realism settings on, day 0 becomes 2,500 × $0.085 × 70% warm-up × 95% (after the 5% revenue cut) = $141.31.
These figures match the app's own calculation, checked during development, along with automated checks of every feature: warm-up, drift, seasonality, CPI scaling, channels, purchases, overhead, both tax methods, cross-promo weights and caps, chaining, the range analysis and currency conversion. No independent audit has been done.
Forecast Accuracy: Unproven
The app hasn't been compared with the real results of live games. The starting values for retention, ARPDAU, CPI, revenue cut, running cost, warm-up, CPI scaling, paid-install retention, store fee and the range spreads are rough, deliberately modest estimates, not measured data and not benchmarks. The built-in holiday seasonality shape is illustrative only, and the currency rates are placeholders until you set or fetch real ones. Never treat output from the starting values as expected results.
What Moves the Answer Most
| Input | Why it matters | What the app does about it |
|---|---|---|
| Retention curve | The biggest driver; small differences in week one compound across every cohort | You set eight points per game, a second set for paid installs, and a scale per channel; no single typical curve is claimed |
| ARPDAU | Varies by country, network, ad format and season | One value per game, adjustable with warm-up, drift and seasonality |
| CPI | Differs by country and platform and tends to rise as you spend more | Market presets you can edit, up to three channels, and a CPI-rise-per-doubling setting; no real market data is bundled |
| Free installs | Streamer and organic traffic is hard to predict | You enter a number; the app doesn't estimate it |
| Everything at once | Assumptions are wrong together, not one at a time | The range analysis moves four inputs at once and shows the spread |
What the Model Leaves Out
- Per-player purchase behavior: purchases are an average per daily user, not a payer funnel
- Revenue per user that differs by country or by channel (channels differ in cost and retention, not in revenue)
- More than three paid channels in a game, or retention that differs by individual campaign or creative
- Taxes beyond a flat rate on profit or a monthly carry-forward; no deductions, credits, or studio-level tax beyond the overhead you enter
- Live exchange rates while you work: rates are fixed until you edit or fetch them, and there's no currency risk modeling
- Pairwise control of cross-promo (who sends how much to whom): each game has a receiving weight and a cap
- Random day-to-day noise. The range analysis varies four inputs by a uniform amount you choose; its percentages describe the model's runs, not the real world
- Saving, loading and exporting results
How to Use It Responsibly
Run three cases
Low, base and high: duplicate the game tab, then run the range analysis to see how fragile the base case is.
Use your own numbers
Replace every starting value with your soft-launch numbers as soon as you have them.
Set honest spreads
Set the range spreads to how unsure you really are; the starting spreads are guesses.
Read ranges
Compare scenarios with each other. Never quote a single run as a forecast.