Which attribution model should a UK content team choose for reporting?

What to take away

Agree the attribution method before the campaign starts, not after the invoice arrives: the choice decides which channel gets credit.
Last click suits short, low-value sales cycles; first click suits long research phases; position-based suits teams that must justify discovery and closing content.
Data-driven attribution needs enough conversion volume to be stable, so small accounts should treat it as a target rather than a starting point.
Record the model, the lookback window and the exclusions in the contract so reporting stays comparable quarter to quarter.

Compare the main attribution methods

Last click and first click

Last click gives all credit to the final touch before conversion. It is easy to explain and reconcile with platform reporting, which is why agencies default to it, but it rewards bottom-funnel pages that would convert anyway. First click credits the entry point instead, flattering discovery content and overstating one blog post's value.

Attribution models compared

Last click

Credit rule: Final touch
Main bias: Bottom funnel
Best when: Short cycle

First click

Credit rule: Entry point
Main bias: Discovery content
Best when: Discovery focus

Linear

Credit rule: Even split
Main bias: None dominant
Best when: No step matters

Position-based

Credit rule: 40/40/rest
Main bias: First and last
Best when: Clear ends

Linear and position-based

Linear splits credit evenly across every touch in the path. It is defensible when no single step clearly matters, and it stays stable because no interaction dominates. Position-based rules usually give 40 per cent to the first touch, 40 per cent to the last and share the rest. For more on choosing metrics, see the measurement and reporting guide.

Data-driven attribution

Data-driven models allocate credit from observed conversion paths, not a fixed rule, so they need volume. Thin conversion data makes weights swing month to month. If your account converts a few dozen times a quarter, run a rules-based model and revisit later.

The UK's Digital Markets, Competition and Consumers Act 2024 also sharpens expectations around transparent commercial claims, including how performance is described to buyers.

Set the rules before you brief

Choose a model in five steps

List every touchpoint you can measure, including organic search, paid social, email and direct.
Count conversions in a typical quarter and decide whether the volume supports a data-driven model.
Match the model to the sales cycle: short cycles tolerate last click, long ones rarely do.
Set a lookback window, for example 30 days for a considered purchase, and apply it everywhere.
Write the model, window and exclusions into the contract, then freeze them for two quarters.

Handle consent and data properly

Attribution depends on tracking, and tracking depends on lawful consent. The ICO's guidance for organisations sets out the principles for collecting and using personal data in marketing measurement.

That means banners that block tags until consent, records of what was collected, and a fallback model for users who decline. Never present modelled conversions as observed ones in a client report.

Report the numbers without inflating them

Show contribution, not just credit

Attribution assigns credit; it does not prove causation. A channel can look strong because it sits near the end of a path. Report assisted conversions alongside attributed ones, and separate branded from non-branded traffic. Where a client wants a return-on-investment figure, follow the appraisal logic in the Green Book: define the counterfactual, state your assumptions and be explicit about uncertainty.

Build a dashboard both sides can read

A dashboard should answer one question per view. Put spend, attributed conversions and cost per acquisition on one screen, then add a channel view showing assisted conversions. If you are starting from scratch, the reporting dashboard in England walks through the layout and the fields worth tracking.

Common questions

Which attribution model should a small business start with?

Start with position-based or linear. Both spread credit across the path, which protects upper-funnel content, and neither needs the conversion volume data-driven models require.

How long should the lookback window be?

Match it to the sales cycle. A 30-day window suits considered purchases; fast transactional sales can use 7 days, and long B2B cycles may need 90 days or more.

Can an agency change the model mid-contract?

Yes, but the change must be agreed in writing and the old data kept. Switching models mid-quarter makes month-on-month comparisons meaningless, so re-baseline before you report again.

Does attribution prove content caused the sale?

No. It distributes credit within a measured path. Proving causation needs a control group or holdout test, which is what the Green Book approach expects.