Marketing analytics for Scottish and Welsh campaign attribution

What to take away

Scottish and Welsh public bodies run their campaigns on the same marketing analytics platforms as any other UK organisation. What differs is the splits they must keep — by nation, by language and by partner — and the reporting duties behind those splits.

Pick the platform before you fix the denominator

Choose the platform fields first, because they decide which splits your denominators can survive.

For devolved campaigns, Google Analytics 4 with event-scoped custom dimensions covers most needs. Create campaign_language, partner_org, jurisdiction, response_channel and consent_regime as dimensions. Register language_delivered, language_responded, partner_org and lawful_basis as event parameters.

GA4's data-driven attribution compares converting with non-converting paths, and Google's Analytics Help documents how to select the model in its 'About attribution models' page. Data-driven sits alongside the rules-based models — last-click, first-click, linear, position-based and time-decay — and the model you choose changes the campaign figure.

Where browsers block client-side tags, collect server-side through Google Tag Manager so partner and jurisdiction values survive. How to set up GA4 data-driven attribution for UK sites covers that build.

GA4 is not the only option. Adobe Analytics holds the same campaign, partner and language splits in its own custom dimensions, and Matomo can be self-hosted, so the data never leaves the body's own estate.

Building a Scottish multi-partner campaign figure

A Scottish campaign is rarely one thing. A national push from the Scottish Government or Public Health Scotland runs alongside NHS Scotland health board variants and a partner strand delivered by a charity or agency.

Build it in four layers:

Example: a Scottish three-partner campaign

Illustrative arithmetic. A national push, two NHS health board variants and a charity partner reach 1.8 million adults and produce 26,000 responses.

In a matched control area the change is 0.4 percentage points; in campaign areas it is 1.1 points. That defended lift of 0.7 points on 620,000 adults gives about 4,300 attributed responses, roughly 17 per cent of the total. Anything above 4,300 is a modelling choice.

If your team separates populations by hand, attribution audience research that keeps six populations apart transfers to a multi-partner Scottish campaign.

Welsh-language strands need three recorded fields

Record the language of delivery, the language of response and the reach in each language as separate fields.

Welsh public bodies — the Welsh Government and Public Health Wales among them — also work under the Welsh Language (Wales) Measure 2011 and standards set by the Welsh Language Commissioner, which carry their own reporting duty. Each body the standards apply to reports annually on how it has met them, and those reports sit alongside the campaign figures.

Example: bilingual registration campaign

Illustrative figures for a bilingual registration campaign run by the Welsh Government or Public Health Wales. English delivery reaches 400,000 adults and produces 3,200 registrations, or 8 per 1,000. Welsh delivery reaches 42,000 and produces 520 registrations, or 12.4 per 1,000. The Welsh strand wins per head of reach, and the combined 3,720 responses hides that.

Publish a range, not a point. With 520 responses, roughly 10.5 to 14.5 per 1,000 is more honest than one decimal place.

Joining campaign data across devolved systems

Join on campaign_id and jurisdiction to keep Scottish, Welsh and cross-border records apart. Reconcile everything to one reporting date, or the totals drift.

Cross-border audiences between Northern Ireland and the Republic

Contacts collected in Northern Ireland sit under UK GDPR and the Data Protection Act 2018. Contacts collected in the Republic sit under EU GDPR. One consent flag cannot carry both.

Tag each record with the regime and the date, keep UK and EU consent trails separate, and report reach by jurisdiction of residence. The Information Commissioner's Office (ICO) is the UK authority here, and its guidance on direct marketing and the public sector (uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/direct-marketing-and-the-public-sector/) sets what may be sent and on what basis.

Label every claim before publication

Say what kind of number it is, in the sentence that carries it.

The Office for Statistics Regulation reviews official statistics and their presentation against the Code of Practice for Statistics, which the UK Statistics Authority owns. Its publications archive shows how comparable claims have been judged.

Rollback plans for a published figure

Keep the rollback block separate from the claim label. Name who owns the figure, the correction trigger and the restatement wording. If partner returns widen the Scottish range, the published figure changes and the change is logged.

Working with agency partners

An agency working for a Scottish or Welsh public body inherits the same duties. Contracts should name who holds the lawful basis record, who owns the joined dataset and who can produce the audit trail on request.

Common questions

Can we publish one all-Wales figure?

Yes, but only beside the per-language split. A combined total alone cannot show whether the Welsh-language strand worked.

Do we need a control group?

For a return claim, yes. Without a matched area or period, the figure is a before-and-after comparison and must be labelled as one.

What fields must a Welsh-language report carry?

Language of delivery, language of response, reach per language and the date the language preference was recorded.

How do we treat contacts collected in the Republic of Ireland?

Keep them in a separate audit trail under EU GDPR, tagged with the collection regime and date, and out of UK denominators.