Attribution for Shopify stores in the UK: what the apps cannot prove

What to take away

The order journey inside one Shopify store

A UK order leaves a long trail. There is the first session, the referrer, the email click, the discount code typed at checkout, the payment, and any refund weeks later.

Each event carries a timestamp and a source field. Vendors read those fields in different orders, so two apps can report different numbers for the same order.

Neither app is broken. They disagree about which touch earns credit, and both work from the same incomplete log.

The ONS retail industry statistics hub tracks how much UK spending moves through online channels, which is the market these tools sell into.

The Shopify attribution apps UK merchants compare

Triple Whale, Northbeam, Lifetimely, Polar Analytics and Analyzify are the apps most often shortlisted by UK Shopify stores for ad attribution. All of them work from the same recorded events: sessions, referring visits, discount codes, checkout and order records. They differ in the credit model they apply, the window they count and the way they treat consent and refunds, which is why two apps can report different order counts for the same month.

What they have in common is the limit. They can show which channel was credited, and they can flag tracking that has broken. They cannot show the orders a campaign did not get, because every order in their reports already received the ad.

What platform attribution can prove, and what it cannot

Platform models are useful for direction and for spotting broken tracking. They cannot separate correlation from cause, because every order in the report already received the ad.

Example: the customer who clicks an ad before reordering

A customer buys once in March through a search ad. In May she clicks a paid social ad and reorders with a saved card. Paid social takes credit for a sale she would have made anyway. A holdout group would show whether the ad added any orders.

Where consent timing changes the numbers

UK cookies and similar technologies sit under PECR, and the ICO guidance on direct marketing sets out the consent rules for storing information on a device.

If a tag waits for consent, early sessions look direct or unattributed. If it fires first, the reported numbers improve while the compliance position weakens.

Neither setting suits every store, but the choice shifts the baseline an app reports against. Two UK shops with different consent banners cannot compare dashboards directly.

PECR governs the cookie or similar technology itself. The UK GDPR governs the personal data that follows it. A store processing order data for marketing still needs a lawful basis for it, and should be able to show which one it relies on.

What a controlled test adds that an app cannot

The method is a controlled experiment, and the standard description of A/B testing covers the statistics behind it.

Two named routes cover most of this. For paid social, the holdout is usually run as a conversion lift test inside Meta's ads experiments tooling. For search and shopping, Google Ads geo experiments hold a matched set of regions out of the campaign for the length of the test. Both answer the question an attribution dashboard cannot: how many orders did the spend add?

What to check in an app before you buy

Store owners also need to check VAT treatment, consent timing and data exports, and our Shopify attribution apps UK comparison sets out those vendor questions before purchase.

The capability question that follows

Once you know which questions need a test, the next task is deciding who runs it. A shop with one analyst and a part-time agency can run a geo holdout. A shop with a data team can run matched cohorts.

Matching the test to the skills you already have means planning attribution from the work in front of you, not from a hoped-for headcount.

Common questions

These are the questions UK store owners ask most often before they buy an attribution app.

Can Shopify attribution apps show incremental lift?

No. They re-credit orders that already happened. Only a holdout or geo experiment can estimate how many orders a campaign added.

Why do two apps report different order counts?

They apply different models, windows and consent rules, and they treat refunds and cross-device journeys differently. Neither figure measures cause.

Is post-purchase attribution worth setting up?

Yes, for reconciling what a platform charged you against what it claimed. It is a reporting check, not proof that the spend worked.

What should a small UK store test first?

Start with one channel that carries real spend. Split two regions or two audience segments for four weeks and compare new customer orders.