Understanding Paid Media Attribution

Separate observation from credit

A customer may see an ad, search the brand, read reviews, click an email, and purchase on another device. No system necessarily observes the complete path. Attribution begins with the interactions a system can connect, then applies rules to credit one or more touches. Unobserved exposure, consent limits, deleted identifiers, and offline activity constrain the picture.

Define the conversion precisely. A platform may count leads by click date while a CRM reports accepted opportunities by creation date. Repeat conversions, view-through interactions, calls, offline imports, and refunds may be handled differently. Reconcile definitions before interpreting a gap as an implementation error.

Understand model choices

Rule-based models can assign credit to the last interaction, first interaction, or several positions. Data-driven approaches use observed data and provider methods to distribute credit. Availability and behavior change, so consult current Google Ads and GA4 documentation rather than relying on an old menu or universal model list.

The lookback window controls which eligible interactions can receive credit. Longer windows can capture slow decisions while increasing overlap among campaigns. Changing a model or window can reassign historical-looking credit without changing actual sales. Record configuration changes and avoid comparing periods as if the measurement basis stayed constant.

Reconcile systems deliberately

Create a field-level comparison for advertising platforms, GA4, CRM, and finance: conversion trigger, timestamp, timezone, attribution window, interaction types, identity method, deduplication, currency, value, cancellation, and reporting latency. Differences become explainable once each system's purpose is visible.

Use a stable business identifier for offline conversion imports where lawful and supported, while protecting personal data. Test duplicates and late updates. Do not force every platform total to equal finance; platform reporting may optimize media, analytics may describe site journeys, CRM may qualify pipeline, and finance records recognized revenue.

Estimate incrementality separately

Attribution asks which observed touch receives credit. Incrementality asks what would have happened without the media. Brand search may receive last-click credit for demand created elsewhere or demand that already existed. To estimate causal lift, use randomized holdouts, geo experiments, matched designs, or other methods appropriate to the spend and decision.

Report attributed outcomes with configuration and caveats, then add experiment evidence where available. Use attribution for operational optimization only within its limits. Avoid summing conversions reported by multiple platforms, because the same customer can be credited several times. Finance-level planning should reconcile to deduplicated business outcomes.

Explain a platform and GA4 mismatch

Google Ads reports 180 purchases while GA4 attributes 125 purchases to paid search for the same calendar month.

1. Confirm purchase event definitions, values, duplicate rules, refunds, timezones, and reporting dates in both systems.
2. Compare attribution windows, eligible ad interactions, model settings, consent behavior, and cross-device identity.
3. Reconcile a sample of transaction IDs with backend orders without exposing personal customer data.
4. Document which differences are expected and correct any verified tagging or import defect.
5. Use deduplicated orders for company totals and a controlled test for questions about incremental paid-search impact.

Attribution reconciliation table

Complete one column for each platform, analytics tool, CRM, and financial source.

• Outcome: event trigger, qualification, value, currency, repeats, refunds, and authoritative business record.
• Time: event date, click date, timezone, lookback window, processing delay, and late-update policy.
• Credit: model, eligible touch types, view-through handling, campaign scope, and direct-traffic rules.
• Identity: cookies, consent, user ID, cross-device method, offline import, deduplication, and privacy boundary.
• Use: optimization purpose, reporting caveat, reconciliation owner, experiment evidence, and decision limit.

Common mistakes

• Adding each advertising platform's attributed conversions together and calling the sum total company sales.
• Changing attribution settings without annotating the date, then treating the resulting shift as channel growth.
• Using last-click or data-driven credit as proof that the credited campaign caused every reported conversion.

Try one

Sources

• Google Ads attribution models guideOfficial explanation of how Google Ads attribution models assign conversion credit.
• Google Analytics attribution guideOfficial GA4 guide to attribution models, reports, and settings.