Marketing AttributionAssigning credit without fooling yourself

The question attribution tries to answer#

A customer sees a video ad, later searches for the brand, reads a review, receives an email and finally buys. Which piece of marketing deserves the credit? Attribution is the set of methods that answer that question with a rule or a model, so that budgets can be shifted towards what appears to work.

The appeal is obvious. Attribution produces numbers daily, at the level of campaign, ad and keyword, which is exactly the granularity buyers of media need. The danger is equally obvious once you see it: credit is not the same as cause. A touchpoint can sit on the path to a purchase that would have happened anyway.

Attribution tells you who was in the room. It does not tell you who closed the deal.

How attribution works#

Every attribution system needs three things: a record of touchpoints, a way to link them to the same person or device, and a rule for dividing credit. Each of those is harder than it looks.

Touchpoint capture is limited by consent choices, ad blockers and browser privacy features. Identity stitching is limited by cookie lifetimes, multiple devices and logged-out behaviour. Joining to outcomes is limited by how well your CRM records lead source. By the time a model divides the credit, the journey it sees is a partial reconstruction.

Two kinds of attribution you will meet#

Platform attribution is what each ad platform reports about itself: conversions that followed a click or view of its own ads, within its own attribution window. It is detailed and fast, and it systematically favours the platform reporting it. Each platform sees only its own touchpoints, so each can claim the same sale.

Cross-channel attribution is done in a neutral tool such as GA4, a CRM or a dedicated attribution product, which sees multiple channels and shares credit among them. It removes double counting but is limited to what it can track, which usually means clicks rather than views, and digital rather than offline channels.

Models: rules versus data-driven#

Rule-based models assign credit by position: all to the last click, all to the first, evenly across touches, or weighted towards the start and end. They are transparent and easy to explain, and each embeds a bias. Data-driven models use the patterns in converting and non-converting paths to estimate each touchpoint's contribution. They are more adaptive and less transparent.

Platforms have changed which models they offer over time; GA4, for example, has narrowed its menu of rule-based options in favour of data-driven attribution and last click. Check your tool's current documentation. We compare the models in detail in attribution models compared.

What attribution cannot see#

Attribution is structurally blind to several things that matter commercially. It cannot see sales that would have happened without any marketing. It struggles with views, especially of video and display, that never produce a click. It cannot see offline media, word of mouth, or the slow build of brand memory that makes someone search for you months later.

The consequence is predictable. Click-based attribution over-credits channels close to the purchase (branded search, retargeting, email to existing customers) and under-credits channels that create demand. Budgets guided only by attribution tend to drift towards harvesting demand that already exists, while the activities that create future demand are cut.

Using attribution well: calibrate, do not worship#

The mature approach treats attribution as one instrument in a set. Use it daily to compare campaigns, ads and keywords within a channel, where its biases are broadly constant. Use incrementality testing to measure what major channels actually add, and apply the results as correction factors. Use marketing mix modelling for the annual shape of the budget, including offline.

Attribution windows and why they move the numbers#

Every attribution system has a lookback window: the period before a conversion during which a touchpoint can earn credit. Platforms usually let you choose separate windows for clicks and for views. A longer window lets a channel claim more conversions; a shorter one claims fewer. Neither is wrong, but they are not comparable.

This is why two reports about the same campaign can disagree honestly. A social platform counting conversions within a period after a view, and an analytics tool counting only the last non-direct click, are measuring different things. Before comparing channels, write down each tool's window and counting rule, and set them as consistently as the platforms allow.

Windows should reflect your buying cycle. A low-cost impulse purchase is usually decided within days; a considered B2B service may take months, far longer than most windows. For long cycles, the CRM rather than the ad platform should carry the story, with the original and most recent source captured on each record.

Getting the foundations right#

Whatever model you choose, its output can be no better than the data feeding it. Consistent UTM parameters, consent-aware tracking, lead source captured in the CRM and offline conversions sent back to platforms all matter more than the choice of model. In B2B, where sales cycles are long, connecting CRM stages to marketing touches is usually the biggest single improvement available.

Finally, report attribution with honesty about its nature. 'Last-click attribution credits paid search with this share of revenue' is accurate. 'Paid search generated this revenue' usually is not. The difference in wording protects the decisions that follow.

Published by Fabulous.Media, a network of specialist marketing agencies. Updated 9 October 2026. Platform features change often; check current official documentation before acting on platform-specific detail.