Google Tag Manager setup checklist for UK lead-gen sites

A Manchester growth lead opens GA4 on a Tuesday, then Google Ads, then HubSpot. The enquiry counts do not match. One dashboard says 41 form submits last week. Ads claims 67 conversions. CRM shows 29 new contacts, and four of those are the same person hitting submit twice on a slow mobile form. Nobody is lying. The container is.

That is the job of a Google Tag Manager setup checklist for UK lead-gen sites: not a catalogue of tags you could theoretically fire, but a scored audit of whether your events survive a redesign, a cookie banner, a SPA handoff, and a Performance Max campaign that is already eating most of the search budget. Failing does not look like a red error in Preview. It looks like booked work you cannot defend in a board pack, CPCs that keep climbing because Smart Bidding is learning from duplicates, and a 20%+ hole in conversion data that privacy tools and ad blockers already punch through.

A tactics list will tell you to “add Enhanced Conversions”. A scored audit tells you which of twelve inspectable items is actually leaking pipeline this week, and which leak you must close before you spend another pound.

Twelve GTM checks that decide whether UK lead-gen numbers can be trusted

Work through these in Preview and in the live container. You are scoring the system that feeds Google Ads, Meta, GA4, and the CRM, not the beauty of the workspace names. Each item is something you can inspect before Friday. If you cannot prove it, it is not a pass.

1. 0 Landing pages are fast enough that tags finish before bounce Fewer than six in ten origins pass all three Core Web Vitals (Google CrUX, 2025). A heavy CMP plus eight marketing pixels on a Glasgow landing page means the conversion tag never fires for the people who convert fastest. Check INP and LCP on the actual enquiry template with a Core Web Vitals checker. If the form is below a slow hero video, you are measuring a different population than the one that booked.

2. 1 Preview, environments, and publish rights are not a free-for-all Look at Versions. If five people publish from Live with no workspace, last week’s “quick pixel” is still live. Good looks like Dev / Staging / Live environments, a named version note (“stop double form fire on HubSpot”), and a two-person rule for anything that hits Google Ads conversion actions. This is boring. It is also how you stop a Friday publish from poisoning bidding over the weekend.

3. 2 You can reconcile GTM, Ads, GA4, and CRM within a known band Pick seven days. Count unique successful form events in GTM (or BigQuery), primary conversions in Google Ads, GA4 key events, and new CRM contacts with a paid source. A healthy UK lead-gen stack disagrees by a modest band because of consent, view-through, and delayed CRM create. A 2x gap is a container problem, not “attribution is dead”. Fix the gap before you scale Performance Max.

If form fields look complete in GTM but people still abandon, the leak is on the page, not the tag. Watch the same enquiry template in HeyLead Insights for scroll depth, rage clicks on the consent bar, and fields they never reach. Tracking a broken form perfectly still books nothing.

When the container itself is the mess, a focused Google Tag Manager and events cleanup is usually cheaper than another month of bidding on duplicate leads.

How to mark pass, weak, or fail without a philosophy debate

Use three marks only. Pass: you demonstrated it in Preview and on a live conversion in the last 14 days. Weak: it works on desktop Chrome and fails on iOS, or it works on one form and not the newsletter modal. Fail: you cannot prove it, or two systems disagree by more than you can explain with consent.

Score the twelve items. Do not average them into a vanity 8/12. Sequence the fails. Tracking integrity before budget. Intent (which event is the primary conversion) before new campaigns. Speed and consent before new creative tags. You do not add a TikTok pixel to a container that already double-fires generate_lead.

A Birmingham B2B team had Google Ads at roughly 4x reported ROAS while CRM said the paid pipeline was thin. The mechanism was a thank-you trigger plus a HubSpot listener both marked primary. They paused the URL trigger, kept the dataLayer event, and turned Enhanced Conversions on with the email field. Reported conversions dropped 37%. Booked demos did not. Bidding calmed within 11 days because the remaining signal was a real person.

Weak items that sit on mobile-only forms belong in week one, not the backlog. Mobile already takes most paid clicks and converts worse when tags and pages lag. If your desktop pass rate is perfect and your iPhone fail list is long, your “account average” is a lie.

DIY tools for this audit

Check the plumbing while you score

• UTM link builder

• Core Web Vitals checker

• Schema markup validator

• Meta description checker

• H1 checker

Thirty days of GTM repairs that keep Google Ads and Meta honest

This is not a new measurement strategy. It is an ordered repair list. You do not rebuild the taxonomy on day one if duplicates are still live. You do not negotiate MMM while gclid never reaches HubSpot.

30-day fix plan

1. Days 1-3: inventory every tag that fires on a successful enquiry. Screenshot Ads conversion actions, Meta events, GA4 key events, and CRM create. Circle duplicates. Unpublish anything you cannot name.

2. Days 4-8: write the dataLayer contract (event, form_id, lead_type, value GBP, email hash source). Implement generate_lead once. Add firing limits. Kill thank-you URL triggers on pages that also push the event.

3. Days 9-12: fix Consent Mode v2 sequencing. Test accept, reject, and partial. Confirm ads tags stay blocked on reject. Confirm they fire on accept without a second page load.

4. Days 13-16: persist gclid, UTMs, and lead type into hidden fields. Send a test enquiry from a tagged Google Ads URL on iPhone Safari. Confirm the CRM record.

5. Days 17-22: turn on Enhanced Conversions from GTM, not a second snippet in Ads. Deduplicate Meta if you use a browser pixel plus CAPI. Mark one primary per lead type.

6. Days 23-30: reconcile seven days of unique events versus Ads versus CRM. If the gap is still wild, do not raise budget. Recheck SPA page_view, call clicks, and the CMP. Only then re-enable Smart Bidding on the cleaned primary.

Edinburgh teams that skip days 1-8 and jump to “server-side GTM” usually clone the same duplicate into a more expensive hole. Server-side is a later hardening step, not a substitute for a single browser event you can defend.

If paid search still lands on a homepage with a generic H1, fix the page after the container, not instead of it. Run the H1 checker on the enquiry URL your ads actually use. Clean conversion setup on a mismatched landing page still trains bidding on the wrong intent.

Patterns we see in UK containers

Google Ads celebrating 67 conversions in a week while HubSpot creates 29 contacts is a pattern we keep seeing: the extra fires are often the same form on validation plus a leftover thank-you tag from a theme update.

Consent reject still letting the Ads tag through is another: Smart Bidding looks brilliant until the tag is blocked properly and CPA goes up, then settles, because the remaining leads are real.

FAQs

Do we need a separate GTM container per UK domain?

Use one container per distinct site or app. A marketing site and an app subdomain can share a container if the dataLayer contract is identical. Do not share one container across unrelated brands just to save a login. Cross-domain tracking is a linker and cookie conversation, not a reason to dump every pixel into one workspace.

Should Google Ads tags sit in GTM or in the Ads UI?

Put them in GTM so consent, sequencing, and form logic stay in one place. Keep a single primary conversion per lead type in the Ads account. The moment someone pastes a second global site tag in the header “just in case”, you are back to double counting.

What is a reasonable mismatch between Ads and the CRM?

Expect some gap from consent, delayed CRM creates, and people who submit then never become a contact. A small, stable band you can explain is fine. A 2x gap week after week is almost always duplicate tags, thank-you triggers, or call clicks without a throttle. Do not raise spend until that band is boring.

Is server-side GTM worth it for UK lead gen right now?

No. First get one client-side generate_lead, Consent Mode v2, Enhanced Conversions, and click IDs into the CRM. Server-side helps with resilience and some signal quality after that. Starting server-side on a messy browser container copies the mess into Cloud Run and a larger invoice.

How does this relate to Performance Max?

PMax will happily optimise to whatever you mark primary. If that primary is a duplicate or a newsletter modal, it will scale the wrong thing. Clean the conversion, then let PMax work. “Performance Max drives the bulk now” is only useful when the bulk is a qualified enquiry.

Putting it to work

This week

• Export live tags and conversion actions; highlight every form-related fire.

• Submit one real test lead on mobile with a tagged URL; trace GTM, Ads, GA4, CRM.

• Turn off any thank-you URL trigger that also has a dataLayer success event.

• Confirm consent reject blocks ads storage in Preview.

• Write the four-field dataLayer contract and share it with whoever owns the site.

Next 30 days

• Ship Enhanced Conversions from GTM only.

• Persist gclid and UTMs into the CRM.

• Reconcile a full week of unique events versus booked contacts.

• Only then change bids or rebuild PMax.

Pull the last 14 days of primary Google Ads conversions and unique CRM contacts from paid, and do not leave the sheet until you can name the GTM tag responsible for every extra conversion. If that reconciliation is the work you do not want on the marketing team’s plate, HeyLead can own the container, consent sequencing, and event contract so bidding stops learning from duplicates. Chat with us on WhatsApp