Shopping feeds attribution methods decide which channel earns the sale

What to take away

Pick the method from the decision in front of youbudget shifts need discovery-friendly credit, variant tests can survive last click.
Write the model, lookback window and exclusions down before the first figure reaches anyone.
Data-driven models need volume and clean conversion tracking, and inherit every gap in that tracking.
Reconcile platform totals against your own orders, then publish the size of the difference.

Why the method changes the number

Attribution is a credit rule, not a truth machine. Last click shows which touchpoint closed the order; first click shows which started it. Linear and position models split credit between the two.

The five common methods apply different credit rules:

Method	Credit rule
Last click	All credit to the final touchpoint before the order
First click	All credit to the first touchpoint in the path
Linear	Equal credit to every touchpoint in the path
Position	Most credit to the first and last touchpoints, the rest shared
Data-driven	Credit weighted by the conversion paths in your own account

Linear suits short paths where each step carries similar weight. Position suits longer paths where discovery and closing matter more than the middle. Both are fixed rules, so neither adapts to new channels the way a data-driven model can.

All can be right about the same orders while pointing budget in different directions. Data and Marketing Association research publishes guidance on data-driven marketing and consent, so use it to test a vendor's favoured split.

What the main methods do with a feed journey

A shopper sees a feed listing on a comparison site in March. She returns to the catalogue in April and buys through a branded search in May. Last click credits the branded search ad. First click credits the comparison listing.

One order, three touchpoints

March
Sees feed listing on comparison site
April
Returns to the catalogue
May
Buys via branded search ad
Last click
Credits branded search ad
First click
Credits comparison listing

That difference alone can move a five-figure budget between channels. Data-driven attribution weights every touchpoint from your own conversion paths instead of applying a fixed rule. Google's bidding systems learn from the same conversion data, so Google Ads best practices treat conversion tracking as part of campaign setup rather than a reporting extra.

The route from raw feed data to board reporting is mapped in the shopping feeds measurement and reporting guide 2027. Read it before you rebuild a model around one channel.

A worked example with labelled pound figures

Figures are illustrative, including the £4,000 and £2,000 shifts in the next paragraph: a multichannel shop takes £30,000 a month from feed-related orders. Last click attributes £18,000 to paid shopping, £7,000 to organic feed listings and £5,000 to email.

Credit split: last click vs data-driven

Paid shopping last click£18,000
Paid shopping data-driven£14,000
Organic listings last click£7,000
Organic listings data-driven£11,000
Email last click£5,000
Email data-driven£7,000

A data-driven model for that month moves £4,000 from paid shopping to organic listings and £2,000 to email, while revenue stays at £30,000, so only the credit moves. Cutting paid budget on the strength of the last-click column would remove spend that created the discovery.

Where data-driven attribution works and where it stalls

Data-driven attribution needs volume. Each path has to carry enough conversions for the platform to find a pattern. A shop taking 40 orders a month from feed traffic may find the option unavailable, or watch the weights swing week to week.

Platforms typically need hundreds of conversions a month before the model will run. Google Ads has cited a threshold of around 300 conversions and 3,000 ad interactions per conversion action in 30 days, though the figure changes and other platforms set their own. Check the current figure in your own account rather than trusting a blog post.

Consent narrows the picture further. If shoppers decline tracking, the modelled paths cover only those who agreed. Name that gap so nobody treats the figure as complete.

What the platform needs before it will model

Conversion tracking has to fire once per order, with enough tagged conversions inside the lookback window. The feed and the landing page also need a consistent match, so a click can be tied to an item.

How to read the modelled conversions

Treat modelled conversions as a second opinion, not a ledger. Compare the modelled split with last click for the same month and note which channels move. A model that shifts credit towards upper-funnel placements is telling you the last-click report undervalued discovery.

Recording the method so it survives a review

Put the model name, lookback window, channels included and known gaps beside every figure.

The layout matters as much as the maths. The shopping feeds reporting dashboard in England shows how to place feed traffic, conversion paths and reconciliation on one view.

Consent and cross-border data

Attribution runs on customer-level event data, and some of it leaves the UK. ICO guidance on international transfers sets out what has to be in place before personal data goes to a vendor abroad. Put that check in the attribution plan, not in a legal review after launch.

Common questions

Which method should a small shop start with?

Last click, because it needs no modelling and everyone understands it. Move to a data-driven model once conversion volume makes the weights stable.

Can I report two models side by side?

Yes, and the comparison is often the most useful output.

What if platform totals do not match my orders?

Expect a gap. Consent, cancellations and returns all separate the two. Reconcile monthly and record the size of the gap.