AppLovin’s 84% margin puts merchant measurement in the spotlight

AppLovin’s November 4 results put a striking number in front of advertisers: an 84% adjusted EBITDA margin for the third quarter of 2026. Revenue reached approximately $2.185 billion, up 55% year over year. These are company-reported financial measures; adjusted EBITDA is non-GAAP. Company release

For a merchant deciding where to put the next advertising dollar, the most useful response is to open a second ledger. The platform’s ledger records its own business economics. The merchant’s records the contribution from customers who would not otherwise have bought. A strong result in the first ledger creates a reason to investigate the channel, but the second determines how far a particular advertiser can expand.

Our reading is that the important commercial question after this quarter is the quality of the next budget increase. A platform can be highly profitable while delivering excellent merchant returns. It can also serve advertisers whose marginal returns differ widely. The earnings figures alone cannot distinguish those situations.

Two ledgers, two denominators

A platform profit margin uses platform revenue as its denominator. A merchant’s advertising return usually compares attributed sales with advertising spend. Even before debating attribution, these ratios describe different entities and different flows of money. Dividing one by the other would produce a number with no useful economic interpretation.

That matters because “84%” invites a shortcut: either the technology must create enormous value for every buyer, or the platform must be taking too much. Neither conclusion follows from this margin. Answering the value-sharing question would require evidence about advertiser outcomes and the economic flows across the advertising chain, not merely the platform’s income statement.

The earnings release is therefore a reason to sharpen merchant measurement, not a substitute for it. The merchant must ask whether an order is additional, whether it is profitable after variable costs, and whether the same economics survive a larger budget. Those are separate tests.

A profitable-looking campaign can still lose money

Consider a deliberately hypothetical merchant spending $10,000 on a campaign. Its reporting dashboard attributes $40,000 in sales to that spend, producing a four-times revenue ROAS. Assume the contribution margin before advertising is 40%, after product costs, fulfillment, payment fees and expected returns. These are illustrative assumptions, not AppLovin customer data.

If every attributed dollar were incremental, the campaign would produce $16,000 of contribution before advertising and $6,000 after the campaign cost. If only half the attributed sales were incremental, the corresponding figures would be $8,000 and a $2,000 loss. The dashboard could display the same four-times return in both cases.

Under these simplified assumptions, break-even requires 62.5% of the attributed revenue to be incremental: $10,000 divided by the $16,000 contribution implied by ($40,000 × 40%). This calculation assumes that incremental and attributed orders have the same margin and that material incremental sales outside the attribution window are absent. A real merchant should estimate total incremental contribution directly wherever possible.

Repeat purchases could improve the result, but projected lifetime value deserves its own evidence. A cohort still inside its return window or first purchase cycle cannot establish a mature payback pattern. Conversely, a short measurement window can miss genuine future value. The answer is a defined observation period and a cohort record, rather than choosing whichever horizon makes the campaign look best.

Put the counterfactual into the budget decision

Attribution assigns credit to observed purchases. Incrementality asks how many purchases would have occurred without the tested advertising. Google’s explanation of conversion lift illustrates the distinction through exposed and control groups. It is a methodological reference, not evidence about AppLovin’s performance or a claim that Google’s product can test AppLovin campaigns. Conversion-lift methodology

A merchant should establish the experimental design before increasing spend. Where feasible, randomize eligible audiences; otherwise consider a properly designed geographic experiment with comparable markets and a documented baseline. Decide in advance which campaigns are in scope, what other spending remains constant, what outcome will be measured and how uncertainty will affect the decision.

The primary outcome should sit close to the cash register: additional contribution from net sales, after returns and variable costs, less additional advertising spend. New-customer share is useful context, but a first-time buyer might have arrived organically. A purchase count alone also misses changes in order value, product mix and refund rates.

Experiments have limits. Customers moving between geographic groups can contaminate a comparison, while small samples make noisy results easy to overinterpret. Google’s geographic guidance describes the contamination problem, and its user-based guidance discusses the trade-off between holdout size and test duration. These are reasons to assess feasibility before starting, not reasons to treat every inconclusive result as zero effect. Geographic design · Study design

Test the next dollar, not yesterday’s average

A successful initial test does not establish the return on a much larger budget. The next group reached may be less responsive, impressions may become more expensive, and the merchant may need different creative. These are possible mechanisms to test, not findings reported in AppLovin’s release.

A practical operating sequence is to reconcile campaign reporting with order and refund records, measure a baseline budget, increase spend within a predefined band, then compare the additional contribution with the additional cost. Keep a record of changes to prices, promotions, inventory and other channels. Otherwise a seasonal lift can be mistaken for a campaign effect.

The expansion rule should also specify what happens when evidence is weak. If the plausible profit range includes a material loss, keeping the budget bounded while improving measurement can be more useful than scaling on a point estimate. If profitable lift repeats across budget levels and mature cohorts, the merchant has a stronger basis for expansion.

AppLovin’s quarter provides evidence about the platform’s financial performance. The next commercial proof belongs in the merchant’s ledger: incremental contribution that persists as spending grows. That is the bridge between an impressive supplier result and a durable advertising relationship.

Sources and method

• AppLovin Q3 2026 financial results, November 4: reported platform financial figures.
• Google Ads: conversion lift: causal measurement framework.
• Google Ads: geographic studies and user-based study design: experiment limitations.

The merchant example and operating framework are Open Market Notes analysis. They do not report AppLovin advertiser results.