Notes

Working notes on getting the number right.

Each one ends in a model you can run, not a conclusion you have to take on trust.

24 September 2026 Your D7 cleared target and payback still moved D7 ROAS measures the first tenth of a cohort's first-year revenue. Payback is decided by the rest. Here is the arithmetic for why a campaign can hit target and still push payback back by months.
22 September 2026 D7 ROAS is not a verdict on brand Upper-funnel activity gets scored with bottom-funnel metrics because that is what exists. The result looks rigorous, is not, and kills campaigns that worked.
22 September 2026 You are measuring two populations as one Blended ARPDAU averages payers and non-payers together, and the two behave nothing like each other. Splitting conversion from payer retention changes what your model says.
22 September 2026 Two clocks: why mobile loses the budget argument Your payback horizon runs for two years. Your fiscal year runs for one. Portfolio budget gets allocated on the shorter clock, and nothing in the standard toolkit reconciles them.
22 September 2026 Why am I doing this, and did it work? Two questions attach to every initiative. Teams answer the first one carefully and almost never answer the second, because the forecast was never built in a way that could be checked.
17 September 2026 Soft launch to global: how to pick the number Market intelligence says one thing, your soft launch says another, and you have to put a single number in the plan. A method for choosing it that survives the meeting.