APPLE ADS ENCYCLOPAEDIA

How reconciliation aligns Apple Ads numbers

Apple Ads truth lives across three streams: console, frameworks and partners. This guide gives the weekly table that keeps them honest together.

By Katie Delaney · 2026-09-04 · 4 min read

Quick answer

Reconcile Apple Ads weekly across console installs, framework installs and partner installs with null shares annotated. Hold tolerance bands, investigate steps and fix joins before touching bids.

Section 01

Three streams, three truths#

Apple Ads reporting runs in three streams: console credit inside tap and view windows, device framework postbacks under privacy rules, and partner dashboards joining the two to spend. Each is correct under its own rules and none matches the others exactly.

Console paces spend intraday; frameworks verify quality weekly; partners narrate the middle. The AdServices reference grounds the token side, while the Apple Ads hub maps all three streams for newcomers.

Arguments about which number is real waste the meeting. Reconcile the streams and move on to decisions.

Section 02

The weekly reconciliation table#

Run one fixed table every week: console installs, framework installs by source, partner installs, null share and spend, all for identical tap date cohorts. Same weekday, same definitions, same tolerance bands, published where stakeholders actually read it.

Anchor cohorts on tap date so privacy delays land in the right bucket, per our postback windows guide, and annotate threshold heavy rows with null context. Keep framework concepts straight with AdAttributionKit and SKAdNetwork notes beside the table.

Routine beats brilliance: twelve calm weekly tables outperform any single forensic deep dive.

Section 03

Set tolerance bands before drama#

Agree tolerance bands while numbers look healthy, so later gaps trigger process instead of panic. A steady gap inside band is plumbing noise; a step change or a gap doubling across two cycles is an incident with an owner and a ticket.

Typical bands widen for fresh campaigns, narrow splits and pre-order cohorts under longer windows. Document the bands on the table itself so every reader knows what normal looks like.

Bands turn reconciliation from theatre into instrumentation. Set them once, review them quarterly.

Section 04

Diagnose gaps in order#

Diagnose divergence in fixed order: cohort alignment, window edges, null share, SDK currency, mapping drift, then cost joins. Most gaps resolve in the first three steps, which cost nothing but care.

Confirm tap date anchoring and 30 day tap plus 1 day view boundaries first, then check value mapping per our conversion values guide and partner health via partner integration and the AppsFlyer and Adjust setup guides. The AdAttributionKit timing reference settles delay disputes.

Ordered diagnosis resolves tickets in hours. Random poking resolves them never.

Section 05

Fix joins before touching bids#

When ROAS or CPA looks wrong, verify spend joins and identifier mapping before changing a single bid. Renamed campaigns, restructured ad groups and broken cost pulls corrupt economics silently while traffic stays innocent.

Match identifiers exactly between Apple Ads and the partner after every structural change, as our account structure guide demands. Rebuild the join, rerun the table and only then judge performance.

Bids changed on broken joins optimise fiction. Fix the plumbing, then steer.

Section 06

Report one number per decision#

Publish one agreed number per decision type: console for pacing, reconciled partner blends for ROAS reviews, framework cohorts for quality audits. Three streams do not mean three competing truths in the board pack.

Label every chart with its stream and cohort definition so readers never compare console intraday against framework weekly again. If reporting politics keep reigniting, talk to folkfox and we will referee the measurement story.

One decision, one number, one footnote. That is the whole reporting standard.

Questions

Frequently asked questions#

Why do console and partner installs differ?

Different windows, privacy delays, thresholds and null handling separate the streams. Weekly reconciliation with null annotations and tap date cohorts explains most gaps without escalation.

How often should I reconcile?

Weekly, on a fixed weekday with fixed cohort definitions. Intraday reconciliation chases delay artefacts; monthly reconciliation lets mapping errors compound unnoticed.

What tolerance is normal?

It depends on volume, splits and windows, which is why bands must be set while numbers look healthy. Steady small gaps are noise; steps and doublings across cycles are incidents.

Which stream should pace budgets?

The console, since bidding and pacing read console credit. Framework and partner streams audit quality and ROAS on a weekly rhythm rather than steering intraday spend.

What breaks reconciliation most?

Misaligned cohorts, forgotten window edges, unannotated nulls, stale SDKs and broken cost joins after renames. The ordered diagnosis in this guide clears them fastest.

How do pre-orders affect reconciliation?

Extended 90 day tap and 61 day view windows stretch credit across months, so keep pre-order cohorts separate until launch fulfils them. Blended tables misattribute announcement buzz to release genius.

Keep reading