Blended CPI and modelling organic uplift
Blended CPI hides a failing paid channel behind organic growth. A simple monthly regression estimates how much of that growth paid spend drove.
Blended cost per install takes total UA spend and divides it by every install a game recorded, paid ones together with organic ones. It gets quoted in leadership meetings more than any other number, because it looks like one clean efficiency figure, and it is also the number most likely to hide a paid channel that is genuinely falling apart. Say organic installs rise for reasons that have nothing to do with paid spend: seasonal search interest, a store feature, a competitor's outage. Blended CPI drops while the paid channel gets more expensive per install, and a team reading only the blended figure misses the failure behind a headline that still looks fine.
The opposite mistake is just as common: crediting all organic growth to paid spend, on the theory that paid installs drive visibility, visibility drives chart position, and chart position drives more organic installs. Some of that is real. Chart position genuinely correlates with organic discovery on both app stores, and a paid-driven spike in daily installs can nudge a title into a chart band where organic pickup increases. Not every organic install traces back to paid spend though, and a team treating the whole organic base as paid-attributable overstates what its UA budget actually buys.
Separating the two effects
The concept worth having is paid-driven organic uplift. Some people shorten it loosely into a k-factor style term. It means the share of organic installs that would not have happened without paid spend in the same period. Isolating it means treating organic installs as a time series that responds to paid spend with a lag, rather than a constant baseline that paid spend sits on top of.
Here is an approach a UA analyst can run monthly without specialist tooling. Regress daily organic installs against daily paid spend and daily paid installs, with a lag, controlling for the obvious seasonal and chart-position variables a team already tracks. Day of week matters, and so do known store features and competitor launches in the same category. If the paid spend coefficient comes out reliably positive and holds steady across a few months of data, that is a reasonable estimate of the uplift. Noise, or a coefficient that flips sign month to month, tells you something too: the organic base is big and steady enough that paid spend is not really moving it, so blended CPI deserves more suspicion as a planning number.
This is a simpler cousin of marketing mix modelling, not a replacement for it. A full MMM wants roughly two years of weekly data before it can pull seasonality apart from price and channel effects with any confidence. The monthly regression is the lighter, faster check on whether the uplift story holds up at all, which a team wants before committing to the heavier exercise.
Reading the result honestly
Once a studio has a working estimate of paid-driven uplift, even a rough one, blended CPI becomes usable again, because now it comes apart into two numbers: true paid CPI, meaning the marginal cost of an install that spend actually caused, plus an uplift-adjusted organic contribution that nobody has to write off as unexplained background growth.
Report both. A dashboard showing the single blended figure flatters a leadership meeting; one showing the decomposition tells a growth lead whether the channel mix is working.
A studio that skips this step tends to make the same mistake in both directions across a title's lifetime. The error just flips sign partway through. Early on, when organic is thin and paid spend genuinely does most of the work, blended CPI understates how expensive growth really is, because a small organic base that paid spend legitimately drives dilutes it. Later, once a title has an established organic audience that would keep arriving with or without a UA budget, the same blended figure starts to overstate paid efficiency, because a shrinking paid contribution now sits on top of a large organic base it no longer meaningfully causes.
Making it a habit rather than a one-off
What makes this durable is treating the regression as a monthly habit rather than a one-off someone runs when finance asks a hard question. Paid-driven uplift is not stable across a studio's lifecycle. It typically shrinks as a title matures and its organic base becomes self-sustaining, and a team that checks once, early in a game's life, will go on crediting paid spend for organic growth the game would now get for free. Automate the re-run once the regression exists. It keeps the uplift estimate current with where the title actually is rather than where it was when somebody first built the model.
