Skip to the content
searchboar.

Search results and measurement

Attribution: assigning credit for a measured outcome

Attribution is the rule or model used to assign credit for a measured outcome to marketing touchpoints. A purchase may involve search, email and a later return visit. Assigning credit does not prove that the outcome would not have happened without a particular channel.

Also known as: attribution · conversion attribution

At a glance

  • Attribution distributes credit; it does not create additional conversions.
  • Different models can report different channel performance for the same customer journey.
  • Attribution reporting and causal impact analysis answer different questions.
On this page
A journey through search, email and a return visit ends in a business goal, with credit assigned to intermediate points.
Attribution distributes credit according to a chosen model; it is not causal proof on its own.

What problem does attribution solve?

A customer may read a search-discovered guide, return through a newsletter and order several days later. If every channel claims the full order, combined reports can imply more revenue than the business actually received.

Attribution allocates an outcome among observed touchpoints within a defined system. The conversion is the completed goal; attribution concerns how marketing credit is assigned to it.

What does the model change?

A last-touch model allocates credit differently from a data-driven approach considering multiple interactions. Google Analytics' documentation describes available approaches and the treatment of direct traffic. Read the rule used by the particular report.

Consider a HUF 60,000 order. One model may assign it to one source while another divides the credit. The received amount remains HUF 60,000. If a channel's reported revenue rises after changing models, examine the rule change before announcing better business performance.

Example: a guide, an email and a return visit

Imagine a coffee roaster whose customer searches for grinding advice. An organic result leads to a guide, the customer subscribes and later purchases through a newsletter offer.

The guide and email may serve different purposes. UTM parameters can identify the email click, but cannot explain every motivation. A friend may have recommended the brand earlier, outside analytics coverage.

This fictional journey illustrates partial observation. A guide receiving no credit in one report is not automatically useless, but its contribution should not be exaggerated without evidence either.

Which other settings matter?

The lookback window determines how far back eligible interactions can be considered. Identification, device changes, measurement coverage and the precise outcome definition also affect results. Two systems with similarly named columns can therefore disagree.

Traffic source and channel group provide different levels of detail. One newsletter source is not the same breakdown as the whole Email channel.

Before comparing reports, record the model, time window, outcome and dimension. Check that a reported conversion means the same action in both systems.

Why is attribution not proof of causation?

Attribution assigns credit to observed interactions. A causal question asks what would have happened without the activity. That requires a suitable experiment or another analysis designed for the circumstances.

A returning customer clicking a branded advertisement may already have decided to buy. The click is measurable and can receive credit, but the report alone does not show how many additional purchases the campaign caused.

Acknowledging this limitation is more useful than treating a reporting model as a complete explanation of customer behavior.

How should a business read the report?

First reconcile overall outcomes with actual orders or inquiries, allowing for differences in coverage. Then examine where credit goes and how sensitive the result is to the selected rules. Consider costs and outcome quality alongside trends.

For content decisions, use attribution as one input. Entry questions, subsequent actions and return visits can also help explain the customer's journey. No single credited-revenue column can fully determine which explanation helped someone understand the product.

For SEO ROI, attributed revenue is only an input. The calculation must state which business value is counted, which SEO costs are included and which period is compared; revenue and profit are different amounts. A more favorable attribution model can change credited revenue without changing orders or costs, so it cannot by itself demonstrate a better return.

Sources

  1. Google Analytics: About attribution
  2. Google Analytics: Conversions and key events
  3. Google Analytics: Traffic-source dimensions