Product Metrics Tree
Most metric answers fail before the arithmetic starts, by reaching for DAU on a product nobody wants to use daily. Pick the product shape first and the right north star follows — then find the guardrail your change might break, and see which growth lever is genuinely worth most.
Calibrated to how product teams actually argue about metrics, not how metrics get taught.
Start with the product, not the metric
The right north star depends entirely on how often the product genuinely delivers value. Pick the shape and the tree changes underneath it.
Sessions per day per user — a feature that raises DAU by pestering people shows up here first
The vanity trap for this product
Raw DAU is defensible here, and almost nowhere else. Even so, "opened the app" is not the same as "got value" — a user who opens, scrolls and leaves without interacting is a churn risk counted as a success.
The move that gets you marked up: name the metric your change might break, not just the one it should move. Every answer that only lists what goes up reads as someone who has not shipped.
Inputs
The retention curve is the whole diagnosis
Not the level — the shape. A curve that flattens means some group found lasting value. A curve that keeps falling to zero means nobody did, however good month one looked. This is the clearest read on product-market fit in any dataset a PM sees.
Flattens at 38% — a loyal core exists, and cohorts stack on it.
Which lever is worth most?
A 10% relative improvement to each in turn. The answer is not the one most people expect, and getting it right is the difference between reciting growth folklore and understanding the model.
They come out equal, and that is the finding. Steady-state base is the product of all three, so a 10% gain anywhere gives 10%. The folk wisdom that “retention compounds and acquisition does not” is wrong at the margin — and right about everything that matters afterwards. The acquisition gain has to be re-bought every single month, forever, at rising cost as you exhaust the cheap channels. The retention gain is bought once and pays out for as long as the product lives.
Where they genuinely diverge is the floor. Drag terminal retention toward zero and no amount of acquisition builds anything, because there is nothing for cohorts to stack on. That is the real asymmetry, and it is structural rather than marginal.
Share your result 🌳
Instagram and TikTok have no desktop share link, so copy the caption and paste it into the app.