AppLovin Creative Reporting: From Assets to ROAS

An AppLovin creative report becomes decision-ready only when its asset identity and delivery data are reconciled with the installs, purchases, and revenue used by the advertiser. The network report is an essential source, but it does not automatically provide every downstream outcome at every creative breakdown a UA team wants.

This is a structural reporting problem, not a cosmetic dashboard problem. A team can export two individually correct reports and still produce a wrong creative ROAS table if their keys, scopes, currencies, or attribution windows do not align.

A note on the name

If your team says AXON, that is the old name. AppLovin rebranded the advertising platform to AppLovin Ads on June 30, 2026. Verified on August 15, 2026: the advertiser site uses AppLovin Ads and calls the optimiser “AppLovin’s AI engine”, and the documentation host support.axon.ai returns HTTP 301 to support.applovin.com. The engine and the reporting surfaces below are unchanged. Only the label moved.

What does AppLovin report directly?

AppLovin exposes separate reporting surfaces for different jobs. As documented on August 15, 2026, its Asset Reporting API returns exactly these columns:

asset_id, asset_name, asset_url, campaign, campaign_id, campaign_package_name, clicks, cost, creative_set, creative_set_id, ctr, impressions

Those documented asset fields do not include installs, sales, revenue, CPI, CPA, or ROAS. Nor do they include anything describing how the engine allocated budget between assets. AppLovin’s own guidance is to read revenue at creative set level instead, which raises a separate question: whether the creative-set revenue ranking is stable enough to act on. We measured it, and it is not.

Two endpoints serve those columns, and their date handling differs in a way that shapes what analysis is possible:

Endpoint	Date handling	Ceiling
/assetReport	Presets only: yesterday, last_7d, last_month	No custom range
/assetAnalyticsReport	Custom range, YYYY-MM-DD	45 day maximum window

The 45 day ceiling matters more than it looks. Any comparison longer than that has to be stitched from several pulls, and any cohort horizon beyond it cannot be read from a single request at all. Plan the pull around the decision, not the other way round.

The advertiser Reporting API exposes broader campaign and creative-set dimensions alongside conversion, sales, purchaser, revenue, and ROAS fields. Those sources should be treated according to their documented grain. A value available at campaign or creative-set grain cannot be assumed to exist at asset grain merely because the final dashboard has an asset row.

For user-acquisition creative analysis, retain a source contract for every field:

Field	Required provenance question
Asset ID and name	Which AppLovin object identifies this uploaded creative?
Spend	At what account, campaign, ad, asset, country, and day grain is it reported?
Impressions and clicks	Are they directly reported for the asset and requested breakdown?
Installs and purchases	Which advertiser report, MMP, or event source supplies them?
Revenue	Is it advertiser revenue, ad revenue, predicted revenue, or a combination?
ROAS	Were revenue and spend aligned before division?

The answer must be explicit. “Available in AppLovin” is not a data contract.

Why creative outcomes require reconciliation

Advertiser outcomes often live outside the asset report. An MMP may attribute installs and purchases. A warehouse may contain validated subscription or in-app revenue. A prediction system may estimate revenue that has not matured yet. Each source can be legitimate while still disagreeing at a given moment.

Before joining them, normalize:

• Account and app identifiers.
• Asset and ad identifiers.
• UTC versus account-local dates.
• Currency and exchange-rate policy.
• Click-through and view-through attribution windows. AppLovin publishes no platform-side window for app campaigns and instead sets minimums your MMP must use, so the window lives on the tracking link. See how the supported networks differ on attribution windows.
• Event names and deduplication rules.
• Late-arriving conversions and restatements.

Do not force unmatched outcomes onto a creative. Preserve an unallocated bucket, report coverage, and make the gap visible. Silent allocation gives every row a complete appearance while destroying auditability.

How should creative-level outcomes be estimated?

When AppLovin does not report an outcome at asset grain, Lemon uses a proprietary estimation and reconciliation system to create a decision-ready asset view. The calculation method remains proprietary.

The output still has to answer the questions a UA team needs before acting:

1. Which values came directly from AppLovin?
2. Which values were reconstructed or derived?
3. Which attribution and cohort horizon applies?
4. Is the value complete, partial, or unavailable?
5. Can the displayed asset rows be aggregated safely?

Ratio metrics such as ROAS, CTR, CPI, and CPA must be recomputed from aligned components rather than averaged. Use the asset-level creative ROAS audit to check provenance, scope, maturity, aggregation, and decision strength before scaling an asset.

A reconciliation checklist before scaling an asset

• Confirm the asset has enough delivery for the selected decision.
• Compare network totals with the ingested source for the same dates.
• Verify that country and platform filters did not change the denominator.
• Confirm that purchase and revenue events use the intended attribution rules.
• Separate actual revenue from predicted revenue.
• Recompute ROAS from aligned revenue and spend.
• Review unmatched coverage and late-arriving data.
• Inspect whether the result holds outside one campaign or country.

An asset should not be promoted merely because it is first in a table. The team should understand whether the ranking reflects a robust business outcome or a narrow slice with incomplete data.

Be especially careful with the spend column. AppLovin’s engine decides how much each asset spends, so ranking by cost tells you what the engine backed, not what earned. Across $32.4 million of AppLovin spend on 742 assets we measured, the highest-spending asset landed at the median of D7 ROAS among the assets running for the same app. Creative testing on AppLovin sets out the method, the counter-evidence, and what a comparison can still prove when you did not control delivery.

What Lemon AI shows

Lemon AI’s AppLovin view places network-reported delivery beside reconstructed creative outcomes. Recovered values use a distinct visual treatment so the provenance is visible rather than hidden. Teams can then break results down by country, platform, and campaign and use the same definitions through the read-only reporting interfaces.

The Creative Analytics page shows the product surface. Reports and API explains programmatic access. The complete rules for reconciliation, ratio aggregation, forecast labeling, and limitations are documented in the methodology.

Primary sources

• AppLovin Asset Reporting API
• AppLovin Reporting API
• AppLovin Ads for consumer brands