The organic baseline you are not tracking

The single most consequential number in a UA operation is one that almost no studio tracks with any rigour: how many installs and how much revenue the game would produce this week if paid spend were zero. Every other figure the team reports, CPI through to blended ROAS and payback period, sits on top of that counterfactual. Guess the counterfactual and every downstream metric inherits the guess.

Some practitioners will say this is a distinction without a difference, because attribution already separates paid from organic. It doesn't. Attribution assigns credit to touchpoints; it says nothing about what the organic line would have done had the paid activity never run, and those two questions aren't the same question.

Why attribution cannot supply the baseline

An MMP labels an install organic when no eligible touchpoint claimed it. That label is a residual, and it shrinks whenever a network gets better at claiming, whenever a lookback window lengthens, whenever somebody switches on a modelled conversion. So a falling organic share on your dashboard has two possible causes: you bought more paid installs, or more networks claimed installs you already had. The dashboard can't tell you which.

The baseline you need is a different animal. It estimates the installs that would have arrived anyway, through search, store featuring, word of mouth, cross-promotion, press, whatever a network does or doesn't claim. That figure belongs to the game and its market position rather than to your attribution setup.

The mechanism that hides it

There's a structural reason the baseline goes untracked. It's nobody's KPI.

Paid efficiency is what the UA manager answers for. Retention and monetisation of whoever turns up is what the product team answers for. Finance looks at a blended number and stops there. The organic baseline falls into the gaps between those functions, and measuring it properly tends to make the paid number look worse, since a higher baseline means a larger share of blended revenue was always going to happen. Nobody has an incentive to estimate it high. The person holding most of the data, the UA lead, carries a mild incentive to estimate it low.

None of that is a character failing. It's what happens when a measurement has no owner and the answer is politically loaded.

What goes wrong downstream

The first-order error is obvious: blended ROAS overstates paid contribution when organic goes under-counted. The piece Blended ROAS Calculation: Where It Quietly Goes Wrong covered that arithmetic.

Then the portfolio error, which gets discussed least of all. A studio with several titles cross-promotes between them, and that traffic usually lands as organic in the receiving game while going unmeasured in the sending one. A strong new launch inflates the organic line of an older title; the UA team on the older title reads the lift as their own paid campaigns working. Then the new launch stops. The older title's blended numbers deteriorate, and the team goes hunting for a creative or network problem that was never there.

The second-order effect does more damage and gets even less airtime. Untracked, the baseline lets paid spend look like it has a constant marginal return, when paid volume and organic volume actually interact. A large burst of paid installs can lift store ranking and search visibility, raising organic for a period. Sustained paid activity in a saturated geography does the opposite and cannibalises installs that would have arrived on their own, because you're paying to reach people who were already going to search for you. The first effect flatters paid; the second penalises it. Without a baseline you see neither, so the team scales into diminishing returns while the dashboard still calls the marginal install profitable.

A workable way to estimate it

You can't observe the baseline directly while you're spending, but you can bound it. Three approaches, in rising order of cost.

Use pre-spend and low-spend periods as anchors. Most games have weeks of soft launch or holiday freeze or budget gap where paid sat near zero. Record daily organic installs and revenue across those windows, adjusted for seasonality, and you have a floor.

Model organic as a function of paid with a lag. Regress daily organic installs on trailing seven-day paid installs across a long history: the intercept estimates the baseline, the coefficient estimates the organic uplift per paid install. One condition governs whether that intercept means anything. The history has to include genuinely low-spend weeks, because if paid has never dropped below a few thousand installs a day you're asking the model to extrapolate to a zero it has never seen, and the intercept turns into a line drawn across empty space rather than an estimate. A studio that has spent continuously since launch should treat the regression as a check on the anchor method rather than a replacement for it, and should expect the intercept to move materially the first time a real budget freeze enters the data. As a purely illustrative example: with a fitted intercept of 800 installs a day and a coefficient of 0.15, a day carrying 5,000 paid installs should show roughly 1,550 organic installs, and any excess or shortfall says something about featuring or press or cannibalisation rather than about paid performance.

Run a periodic geo holdout. Pause paid in a small set of comparable regions for two or three weeks and watch what the organic line does. It's the only method that reads baseline and cannibalisation directly, and the design sits in Incrementality Testing for Mobile Games You Can Afford. Once a year is enough for most studios.

The decision rule

Whichever method you use, write the baseline down as a named number with an owner and a review date. Then change one thing in the weekly review: report paid ROAS against incremental revenue, meaning blended revenue minus the baseline estimate, instead of against MMP-attributed revenue.

The practical effect is that paid campaigns answer for what they add above the floor rather than for what they manage to claim. In most games the baseline runs higher than the team assumed, and the honest paid return runs lower. Uncomfortable for a quarter. After that it becomes the one number in the review that finance and product both trust, which makes it the number budgets eventually follow.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• ATT opt-in is 39%. Adjust's retention report and the paid-to-organic ratio nobody budgets for
• The economics of buying your own players back