Marketing attribution without false precision

A marketing team follows several green and gold customer paths converging on one business outcome.

By Patricia Jarmek Marketing Strategy & Analytics 26 October 2026

4 min read

A buyer may discover a brand through organic search, return from social, open an email and submit an enquiry only after a direct visit. Every channel participates, but an analytics system still needs a rule for distributing credit.

Google defines attribution as assigning credit to ads, clicks and other factors along the path to a meaningful action [1]. That is not proof of causality. Attribution describes an observed path; an experiment or incrementality test asks whether the result would have happened without the investment.

One outcome, several valid views

GA4, Google Ads, a CRM and WooCommerce do not have to report the same number. They can differ in identity, lookback window, time zone, consent, model, eligible channels and the moment at which an outcome is considered complete.

Define the business source of truth first. Revenue normally belongs to the order or accounting system; lead quality belongs to the CRM. Analytics platforms then explain the path to that known result.

What the GA4 model actually changes

GA4 currently provides data-driven and paid-and-organic last-click models, plus Google paid channels last click. Data-driven attribution distributes credit using data for each key event; last click gives credit to the final eligible interaction [2].

Changing the reporting model affects historical and future reporting that uses event-scoped Source, Medium, Campaign and channel dimensions. It does not change First user or Session dimensions. Fractional key-event and revenue values are therefore normal under data-driven attribution [2].

Inspect paths before moving budget

The GA4 Advertising area includes attribution paths and model comparison by channel [3]. Compare channels that start, assist and close paths, the number of touchpoints, time to outcome and how credit moves between models.

Google Ads Path metrics reports time and interactions before conversion, but Google notes that the path only reflects keywords and ads inside that Google Ads account [4]. A platform path is not the whole customer journey.

The attribution window is a business choice

A window determines how long a touchpoint remains eligible for credit. A short window undervalues early touches in a long sales cycle; a long one may overstate an old interaction. GA4 allows lookback-window settings for acquisition and other key events [5]. Document the window, date and reason whenever it changes.

Do not confuse measurement with optimisation

The choice of channels eligible for credit can also affect conversions shared with Google Ads, bidding and reporting [6]. Google Ads data-driven attribution uses account data to estimate the contribution of keywords, ads and campaigns [7].

Separate the signal used for automated optimisation from the cross-channel view used to allocate the total budget.

A practical framework without false precision

For a monthly decision, combine four layers:

Business outcome: actual revenue, margin, qualified leads and won deals.
GA4 cross-channel view: paths plus data-driven and last-click comparison.
Platform view: ad-network reporting with its scope and window stated.
Causal test: a holdout, geo-test, lift study or controlled budget change when the decision is material.

Do not force systems to match exactly. Set an acceptable variance, monitor its trend and investigate sudden changes. The best attribution report shows whether a decision survives a change of model.

Check outcome definitions, the active model, eligible channels, lookback windows, UTMs, cross-domain measurement and whether a budget conclusion needs an experiment. Attribution is useful when it reduces bad decisions; it becomes dangerous when fractional credit is presented as physical proof that a channel caused a sale.

Frequently asked questions

What is the best attribution model?

There is no universal best model. Use the model that fits the question, compare material decisions under more than one view and test incrementality where possible.

Why do GA4 and Google Ads report different conversions?

Because their scope, model, windows, identity, consent and processing can differ.

Does data-driven attribution prove causality?

No. It distributes credit within available observations. A controlled test is stronger evidence of incremental impact.

Should we change the lookback window?

Only when it does not match the buying cycle. Record the change so comparisons remain interpretable.

Is direct traffic useless for attribution?

No. It can represent returning users or an unrecognised source and should prompt a tagging and journey review.