What each marketing agencies attribution method means for your reporting

What to take away

Why attribution methods matter more in 2026

Attribution rules decide which channel gets credit for a sale. That single choice moves budget between paid search, organic, email and out-of-home. It also moves the fee an agency can justify.

Historically, many agencies reported last-click because it was easy to explain. Since 2025, Google Analytics has defaulted to data-driven attribution in its reporting, which means a buyer now sees a modelled number unless someone changes the setting. That shift makes the method a commercial question rather than a technical one.

The marketing agencies measurement and reporting guide 2027 sets out how to compare reporting offers across suppliers if you need the wider framework before choosing a model.

Last-click against data-driven, side by side

Last-click gives all credit to the final interaction. It is transparent, cheap to run and easy to audit. It also flatters retargeting and brand search, which rarely start a purchase.

Data-driven attribution uses a model trained on your conversion data to spread credit. It is more accurate in principle, but only where there is enough conversion volume to train on. Below that threshold, the output is unstable.

First-click gives all credit to the first interaction, highlighting discovery but potentially understating channels that close the sale. Linear splits credit evenly across touchpoints, which suits long consideration cycles. Time-decay weights recent touchpoints more heavily, which suits journeys where later interactions matter more. A common positional model gives greater weight to the first and last touchpoints and shares the remainder among the middle interactions; it highlights discovery and closing activity, but depends on the chosen weighting.

Consider a customer who first discovers a service through organic search, returns through email and finally clicks a paid-search ad. Last-click credits paid search, first-click credits organic search, linear divides credit across all three, time-decay favours the final interaction, and a common positional model gives greater weight to the first and final interactions. A data-driven model estimates contributions from observed conversion data, so its allocation can differ from these fixed rules.

A reporting window is the period after an eligible interaction in which a conversion can receive credit. A shorter window may exclude later purchases; a longer one may include more, increasing the reported total. Compare reports only when the model, conversion definition and click-through or view-through windows are clear.

What to ask a marketing agency before you sign

Which model will you report on, and what is the fallback?

Ask for the named model in writing. If conversion volume drops and the model becomes unreliable, the agency should state the fallback in advance, usually a positional or linear model. A written fallback stops a mid-quarter argument.

How will we reconcile the model with platform numbers?

Every platform reports its own conversions, so totals will never match. Ask how the agency reconciles the gap and how it labels modelled figures against observed ones. A clear label is the difference between evidence and decoration.

A marketing agencies reporting dashboard in England is only as good as the definitions behind it, so agree those definitions first.

VAT and contracting points

Ask whether reporting and modelling are included in the retainer or billed separately, and have the scope and fee schedule state what is covered.

Where an agency bills for analytics or consultancy, VAT treatment depends on the service. The VAT rates on different goods and services published by GOV.UK set out the standard, reduced and zero rates, so check the correct rate before you approve a purchase order.

Contract clauses worth adding: the model name, the reporting period, the data sources included, and who owns the modelled output if the relationship ends.

Where industry practice is heading

The Agents of change plan from the IPA describes how agency business models are shifting towards consultancy and specialist advisory work, which often brings measurement and attribution into the fee conversation.

For networking and shared standards, the IAB UK communities and groups are where measurement practice gets discussed between agencies and advertisers.

Decision table

Common questions

Is last-click attribution still acceptable?

Yes, provided it is named in the contract and everyone understands what it overstates. It is cheap and auditable, which suits smaller accounts.

Can two agencies report different numbers for the same campaign?

Yes. Different models, attribution windows and platform definitions all produce different totals. Agree the definitions before comparing suppliers.

Does attribution change the agency fee?

It can, when modelling and reconciliation are separately charged; there is no single figure that applies to every agency and scope. Ask whether the work sits inside the retainer or is billed separately.

How often should the model be reviewed?

At least once a year, and whenever tracking, consent rates or channel mix change materially.