Method

Simple math. Assumed labels. No fake precision.

Soft-mode metrics, in plain English

Each Soft pack checks four gates. Left side is your measured number; right side is the bar you must clear. Pass three or four → SCALE. Pass two → ITERATE. Pass zero or one → KILL.

  • D1Day-1 retention

    Share of players who come back the day after install.

    Pass when your % is at or above the minimum.

  • D7Day-7 retention

    Share of players still coming back one week after install.

    Pass when your % is at or above the minimum.

  • CPICost per install

    Average dollars spent to get one install.

    Pass when your $ is at or below the maximum.

  • ROASEarly return on ad spend

    Early revenue ÷ ad spend. 0.40 means you earned $0.40 for every $1 spent so far.

    Pass when your multiple is at or above the minimum.

Spend and burn

  • SpendUA spend to date

    How much you have already spent on user acquisition for this soft launch.

    Used only to size the Assumed burn envelope — not a pass/fail gate.

  • BurnAssumed burn envelope

    Spend × burn multiplier (default 1.5). Rough dollars left if you keep the soft launch open.

    Labeled Assumed — a planning envelope, not a forecast.

How the verdict is scored

  • SCALE3 or 4 of 4 gates pass — Strong enough to widen spend.
  • ITERATE2 of 4 gates pass — Fix weak gates before you scale.
  • KILL0 or 1 of 4 gates pass — Stop — more weeks will not save it.

Default Soft bars (hybridcasual template)

Count how many of those four pass. 3 or 4 → SCALE. Exactly 2 → ITERATE. 0 or 1 → KILL. Assumed burn envelope = UA spend to date × burn multiplier (default 1.5).

Live release

Baseline = mean revenue of all rows except the latest. Cliff if latest is ≥15% below that mean (tunable). Assumed $ at risk = daily revenue gap × horizon days (default 7).

What we refuse to pretend

Soft mode does not forecast lifetime ROAS from day-3 data. Live mode does not replace a crash reporter. If you need full LTV or RUM, keep the tools you already pay for.