Marketing Attribution: What It Can Explain and What Remains Uncertain

Marketing attribution can explain which observed touchpoints a measurement system gives credit for a defined event, such as a submitted enquiry or a purchase. That can help a team compare recorded journeys and frame a test. It cannot, on its own, prove that an ad caused a sale, that one channel deserves all the budget, or that every interaction by a person was observed. Attribution is a way to interpret collected data, not a causal verdict.

The limited Ahrefs volume for this topic is an estimate for Morocco as of 22 October 2026. It identifies a decision-stage measurement question; it does not predict traffic, rankings, leads, or results.

What attribution can explain

A conversion may follow several interactions: a search, an ad, a direct visit, an email and then a form submission. Google Analytics defines attribution as assigning credit across interactions leading to a key event. Its attribution-path report can show recorded paths to key events, touchpoints, revenue where applicable, and time to the event (Google Analytics documentation).

This makes attribution useful for focused questions. Which channels appear early or late in converted journeys? Did a named campaign produce measurable enquiries? Do leads that become qualified retain a recorded acquisition source? Start by selecting one outcome: submitted form, accepted enquiry, attended meeting, paid order, or closed sale. A dashboard that combines all of those as one “conversion” removes the meaning needed for a decision.

Credit is not proof of causation

When a model gives partial credit to a channel, it applies a rule or calculation to available data. In Google Analytics, a data-driven view can distribute fractional credit across interactions. That does not turn the total into a controlled experiment. An ad may have contributed, may merely be present in the path, or may receive credit for demand that already existed. This distinction matters before increasing, reducing or pausing spend.

Begin with a decision, not a report

An attribution view earns its place when it helps a decision that can change. Write the decision in plain terms: “Should we continue this campaign test after checking lead quality?” or “Which decision-stage page should support an enquiry?” Avoid “Which channel is best?” because it does not define the event, standard, or action.

Define the event and its evidence

For each stage, record the technical trigger, the accountable team and the evidence expected. A form submission is a website event. A qualified enquiry needs a documented commercial rule. A sale needs a consistent status and date in the order or CRM system. Define exclusions too: spam, internal tests, duplicate entries, cancellations and out-of-scope requests.

A usable measurement register includes the event name, definition, location, optional value, consent basis, verification method and owner of future changes. This prevents a platform-recorded click from being compared with a CRM-recognised sale as though they were the same thing.

Keep three questions separate

Use separate views for acquisition, the observable journey and the commercial outcome. Acquisition explains how a visit arrived. Journey reporting describes recordable interactions before the event. Commercial reporting describes what happened after the enquiry in the CRM or order system. They can disagree without any one report being inherently wrong. A Tracking & Analytics service engagement should establish those definitions before adding a more elaborate dashboard.

Make acquisition data interpretable

A model cannot resolve inconsistent source data. Google recommends using utm_source, utm_medium and utm_campaign consistently for manually tagged links. Missing required parameters can create “not set” values, and parameter value case affects reporting (Google’s campaign URL guide).

Govern a short naming convention

Agree a stable value for platform, medium and campaign before publishing. Add fields only when they support a decision. For example, utm_source=linkedin, utm_medium=paid_social and utm_campaign=offer-audit-q4, with content used to separate creative variants. Do not place personal data in a URL. Do not let every publisher spell the same source differently or mix upper and lower case without a rule.

Platform integrations and auto-tagging can supplement this scheme. Google also notes that, in some cases, if any UTM parameter is present, Analytics derives all cross-channel traffic-source dimensions exclusively from UTM values. Incomplete tagging can therefore fragment the report (Google Analytics traffic-source documentation). Test a link before launch, then check source, medium and campaign in acquisition reporting.

Test the actual journey

Test browser, mobile, form and confirmation page separately. A click on “submit” should not automatically be treated as a successful enquiry. If a site sends the same event from the browser and the server, test deduplication. Meta states that its recommended Pixel and Conversions API approach uses matching event names and matching event IDs for corresponding events (Meta’s deduplication guide).

Read the model without asking it to do too much

In GA4, the reporting model, eligible channels and key-event lookback window are attribution settings. Changing the reporting model can change attribution in the relevant reports, while the lookback window sets how far back a touchpoint may be eligible for credit (Google’s attribution settings documentation). Record the settings, event and review date before comparing periods.

A lookback window is a visibility rule

A shorter window makes older interactions ineligible; a longer one admits more. Neither is universally truthful. Choose a window that fits the decision cycle under review and keep it stable long enough to make a comparison meaningful. If it changes, annotate the report instead of presenting the movement as a demand trend.

Compare scenarios, not rival truths

Review the same period through two readings that are relevant to the decision, such as last click and the configured GA4 view. If the conclusion only holds under one setting, treat it as a fragile signal that needs more evidence. Meta specifically says campaign-overview results cannot be compared across ad sets using different attribution models, because their counting mechanisms differ (Meta attribution models and settings).

Connect website activity to the commercial outcome

An online enquiry is not automatically revenue. Where sales take time, create a non-sensitive lead identifier, pass source and campaign into the CRM, and agree clear stages: received, contacted, qualified, proposal, won and lost. Review discrepancies with sales. A status that was never updated can make a useful channel appear weak.

Google Ads explains that offline conversion imports can measure activity after an ad click or call, such as a sale completed by phone. This depends on the identifiers, available data, selected configuration and applicable consent; it does not reconstruct every customer journey (Google Ads offline conversion documentation). The purpose of CRM reconciliation is more modest and more useful: among attributed enquiries, which progressed under the commercial definition the team agreed?

Make changes governable

Decide who can create a conversion, change an attribution model, rename a campaign, or import a CRM stage. Keep a change log. Investigate platform, Analytics and CRM differences before merging numbers into one score. A digital strategy service can then prioritise hypotheses, pages and investment choices from documented evidence rather than treat a dashboard as a performance promise.

What will still be uncertain

Attribution can only use interactions that were collected and are eligible under its rules. It may not see an offline conversation, a private recommendation, another device, an unclicked impression, or a visit where measurement was not permitted. Browser behaviour, user choices and cross-device matching limits also reduce what can be connected. The W3C notes that even privacy-respecting conversion-attribution APIs reveal only limited information about a person and frames data minimisation as a web principle (W3C Privacy Principles).

Do not answer that uncertainty by collecting everything. Document the scope instead: included properties, consent state, dates, model, lookback window, events, CRM system and unmatched lead rate. Use results as signals alongside lead quality, commercial context, creative changes and limited tests. When the decision is material, seek a test design suited to the question rather than retrospective certainty.

A practical working sequence

Choose the decision and final event, then write their definitions.
Map the touchpoints and data that must pass from the website into the CRM.
Publish a UTM convention and nominate campaign owners.
Test events, confirmation pages and duplication before launch.
Record model, window and date; review path reporting and commercial outcomes separately.
Explain discrepancies, select a bounded next test, and preserve the rationale for the decision.

This sequence does not promise a perfect number. It makes measurement boundaries visible and prevents a polished report from being mistaken for complete evidence.

FAQ: marketing attribution

Does marketing attribution tell me which channel caused a sale?

No. It allocates credit using observed interactions and a calculation rule. It can support a contribution hypothesis, but does not establish causation on its own.

Why do Meta, Google Ads and GA4 show different figures?

They may use different models, windows, observed data, time zones and conversion definitions. Check the setting and event being measured before interpreting the difference.

Should I use last click or a data-driven model?

Choose for the decision and document the setting. Last click can clarify the final observed interaction; a data-driven model distributes credit differently. Neither makes uncollected interactions visible.

Are UTMs enough for reliable attribution?

They provide important campaign identification for many links, but they do not replace event quality, deduplication or reconciliation with CRM outcomes.

Can I measure sales completed after a phone call?

A reconciliation process or, where configured, certain offline conversion imports can help. Quality depends on retaining a link between click, lead and commercial outcome, as well as meeting applicable requirements.