Why null is not zero in Apple Ads

Null is not zero, full stop#

Null means Apple privacy thresholds withheld the value; zero means the device reported bucket zero. Treating nulls as zeros understates quality, punishes small campaigns and misleads every downstream calculation.

The distinction sits at the heart of honest Apple Ads measurement, so print it on every dashboard that shows framework data. The SKAdNetwork reference describes the threshold mechanics that produce nulls. Start newcomers at the Apple Ads hub before handing them framework reports.

Every other section of this guide is technique for living with that sentence. Memorise it first.

What triggers privacy thresholds#

Thresholds suppress cohorts too small or too distinctive to report safely, hiding installs, values or source detail until volume qualifies. Small budgets, narrow targeting and fresh campaigns trip them most, which is why new tests look unfairly bleak in framework data.

Volume is the only legitimate lever: broader cohorts and longer windows clear thresholds that daily slices cannot. Customer type splits and location splits multiply the problem by slicing cohorts thinner, so consolidate reporting dimensions while diagnosing.

When a cohort flips from null to valued without any product change, volume crossed the line. Celebrate the data, not a phantom optimisation.

Where nulls hide in your reports#

Nulls hide in conversion values, source identifiers and campaign detail, each with different consequences. Missing values flatten quality reads; missing source detail merges campaigns into unattributed pools that tempt wrong budget moves.

Survey every framework fed dashboard and label each field as valued, zero or null, using raw postback views in AppsFlyer or Adjust as ground truth. The AdAttributionKit reference documents which fields suppress under what conditions.

Unlabelled dashboards breed null blindness. Label everything and the fear drains out of the numbers.

Reading rules for null heavy cohorts#

Read null heavy cohorts with wider windows, pooled dimensions and console side evidence, never with intraday framework snapshots. Aggregate up until thresholds clear, then compare the pooled read against console installs for the same span.

Exclude nulls from averages rather than imputing zeros, and show null share beside every KPI so readers discount appropriately. Our reconciliation guide standardises the annotated table, and our postback windows guide sets the waiting times.

Discipline here separates analysts who inform budgets from analysts who merely decorate them.

Budgeting with nulls in mind#

Fund tests above threshold viable volume from the start, because starved cohorts return nulls that prove nothing. Minimum viable test budgets must clear privacy volume as well as learning volume, a double bar small accounts often miss.

Size tests with our learning budget guide and concentrate spend in fewer campaigns per structured splits so each cohort earns reportable scale. Document the volume assumption in the test brief before launch.

Money spent below threshold buys neither learning nor evidence. Consolidate or stay out.

Explaining nulls to stakeholders#

Translate nulls into plain business language: the users may well have converted, but privacy rules prevent confirmation at this granularity. Pair every null heavy chart with console revenue and a one line threshold note, and scepticism turns into patience.

Keep a standing slide that shows null share trends alongside installs so leadership watches suppression fall as volume grows. If the conversation keeps stalling, talk to folkfox and we will present the measurement story directly.

Stakeholders forgive missing data faster than they forgive confident wrong data. Honesty compounds.

Frequently asked questions#