Google Ads Attribution: A UK Business Guide to Understanding Which Campaigns Actually Drive Sales in 2026

Every UK business running paid search eventually has the same meeting. Someone opens the Google Ads interface, sorts the campaign list by conversions, points at the campaigns near the bottom and asks why the company is still funding them. It is a reasonable question asked against unreasonable evidence, because the number being sorted on is a product of the Google Ads attribution model in use, not a measurement of which campaigns caused the sale. Change the model and the ranking changes. Nothing about the business changed; only the accounting did.

This is not a marginal effect. On accounts where the purchase decision takes more than a single session — which is nearly every considered B2B or high-value B2C purchase in the UK — the campaign that introduces a customer to the brand and the campaign that catches them at the moment of purchase are almost never the same campaign. Judge both on last-click conversions and the first one looks like waste. Cut it, and three months later the campaign that used to convert reliably starts to starve, because the demand it was harvesting was being created upstream by the thing you switched off.

This guide covers what attribution actually does to your numbers and your budget decisions. It walks through the two models Google Ads still offers in 2026 and why the others disappeared, how data-driven attribution reaches its conclusions, how to read GA4 attribution reporting and the conversion paths report without over-interpreting it, how multi-touch attribution works in practice rather than in theory, and how offline conversion tracking closes the gap between a form fill and the invoice that eventually followed it. It also covers the UK-specific constraints — consent mode, PECR, UK GDPR handling of hashed customer data — that determine how much of your data survives to be attributed at all. Expect realistic costs, realistic timelines, and a clear account of what attribution cannot tell you.

What Google Ads attribution actually is — and what it is not

Attribution is the set of rules that decides which advertising interactions get credited for a conversion when more than one interaction preceded it. That is the whole of it. It is a bookkeeping convention, not a measurement instrument. When a customer clicks a broad-match discovery ad in March, returns through a branded search in April and finally converts through a shopping ad in May, exactly one sale occurred. Attribution decides how that single sale is written down across three campaigns, and every model writes it down differently.

The distinction that matters most, and the one most commonly missed, is between attribution and incrementality. Attribution answers “which of my ads did this converting customer touch?” Incrementality answers “would this customer have converted anyway?” They are different questions with different answers, and no attribution model of any sophistication can answer the second one. Branded search is the standard illustration: it attributes superbly under every model because customers who already intend to buy from you search your name before purchasing. That does not make the spend incremental. A meaningful share of those customers would have found you regardless. Attribution will never tell you which share, because attribution only sees the customers who clicked.

What attribution is genuinely good for is comparative budget allocation between campaigns of similar type, and detection of the assist patterns that last-click hides entirely. Used for that, it is one of the highest-return pieces of measurement work a UK advertiser can commission. Used as a proof of return on investment for the channel as a whole, it will mislead you, usually in a flattering direction. Hold both facts at once and the rest of this guide is straightforward.

The second thing to be clear about is the difference between the attribution model and the conversion window. The model decides how credit is split among the clicks inside the window. The window decides which clicks are eligible in the first place. A 30-day click window on an eight-week sales cycle is not an attribution problem — it is a scope problem, and no model change will fix it. Windows are configurable per conversion action from one to 90 days for clicks, and one to 30 days for engaged views. Most UK accounts have never changed them from the defaults, which is a five-minute fix with a larger effect on reported performance than most model debates.

Attribution by the numbers — what UK advertisers are actually working with

The figures below reflect the pattern that repeats across mid-market UK Google Ads accounts: monthly spend between £3,000 and £60,000, a mixture of search, shopping and Performance Max, a CRM that holds the real revenue data, and a measurement setup that was configured once at launch and has been inherited by three agencies since. They are not universal, but they are the numbers to argue against if you think your account is different.

The first of those four is the one that surprises people. Google Ads used to offer six attribution models: last click, first click, linear, time decay, position-based and data-driven. First click, linear, time decay and position-based were retired across 2023 and 2024, and the interface now offers data-driven and last click only. A great many UK advertisers — and a surprising number of agency reports — still describe a position-based or time-decay setup that has not existed for two years. If a proposal or a monthly report references one of the retired models, it was written from a template rather than from the account.

The 28% figure deserves the most attention, because it is the number that turns attribution from a methodology discussion into a budget decision. On a £25,000 monthly account, a 28% redistribution of credit between campaigns typically moves £4,000 to £7,000 of monthly budget once the optimiser responds — either you or Smart Bidding. That is real money moving on the basis of an accounting rule, which is precisely why the rule deserves scrutiny rather than acceptance.

The median of 3.7 ad interactions per converting path also carries a quiet implication. If the typical converting customer touches your advertising nearly four times, then last-click attribution is discarding roughly two-thirds of the observed interactions as worthless. It is not that they are being weighted lightly. They are being assigned exactly zero, and any campaign whose function is to produce them will read as a failure in every report you run.

Where credit actually sits — conversion paths by touchpoint position

The chart below shows how often each type of Google Ads campaign appears in a given position on a converting path, drawn from the shape of typical UK multi-touch accounts with a sales cycle of two weeks or longer. Read it as a description of role, not of value: a campaign that appears overwhelmingly in the opening position is doing introduction work, and a campaign that appears overwhelmingly in the closing position is doing harvesting work. Both are necessary. Only one of them survives a last-click review.

Branded search sitting at 71% as the final touch is the single most consequential line on this chart, and it explains most of what goes wrong in UK paid search budgeting. Branded campaigns are cheap, they convert at rates that look extraordinary next to everything else, and under last click they collect the credit for demand that other campaigns generated. The predictable result is an account that gradually reallocates budget toward branded search, reports improving efficiency for two or three quarters, and then discovers that branded search volume itself is declining — because nothing upstream is still creating people who know the brand name to search for.

The generic search figure is the mirror image. Generic terms open 58% of converting paths and close comparatively few. Under a last-click model these campaigns will always look expensive relative to branded, always be the first candidates for cuts, and always be the wrong cut to make. This is the specific failure mode the rest of this guide is designed to prevent, and it is worth noting that it compounds silently: the damage from cutting upstream demand generation appears one full sales cycle after the decision, by which point the decision looks unrelated to the symptom.

Performance Max deserves separate comment. Its 41% mid-path assist rate is the honest version of a campaign type that is frequently either over-credited or dismissed entirely. Because Performance Max spans search, shopping, display, video and Gmail inventory inside a single campaign, it collects touchpoints across the whole path and reports them under one line item. That makes it look excellent under any model that rewards presence, and makes genuine evaluation depend on segmenting the channel distribution inside it rather than reading the campaign total. The same discipline that applies to reading a technical SEO audit report applies here: an aggregate score is a starting point for investigation, never a conclusion.

What attribution work costs in the UK

Attribution is not a single purchase. It is a set of separable pieces of work, each of which can be bought independently, and the order you buy them in matters more than the total. The table below gives realistic UK market rates for 2026 across agencies and specialist consultancies. Prices are for a mid-market account — roughly £10,000 to £50,000 monthly media spend, one primary domain, one CRM — and scale roughly with the number of distinct conversion actions and systems involved rather than with media spend itself.

The ordering advice is firm and it saves money. Buy the audit first, always. A meaningful proportion of UK accounts commissioned for advanced attribution work turn out to have a duplicate conversion action double-counting a subset of leads, or a thank-you page tag that also fires on page refresh, or a Performance Max campaign optimising toward a micro-conversion someone set as primary in 2023. None of that is fixed by a better model; all of it invalidates any model you apply on top. Spending £6,000 on server-side tagging to more reliably transmit a wrong number is a common and entirely avoidable mistake.

The second point on cost is that offline conversion import is almost always the best value item on the list for a lead-generation business, and almost always the last one bought. It costs more than the others and requires cooperation from whoever owns the CRM, which is why it gets deferred. But for any business where a form fill is worth between nothing and £50,000 depending on what it turns into, no amount of front-end attribution sophistication compensates for optimising toward an undifferentiated lead count. Getting that sequencing right is a governance question as much as a marketing one, and it belongs in the same conversation as the rest of your technology strategy and IT roadmap rather than being decided inside a single channel.

Last click versus data-driven attribution — how the two models differ

Since the retirement of the rule-based models, this is the only model choice Google Ads presents. It is set per conversion action, not per campaign or per account, which means a single account can and often does run different models for different conversion actions — a fact that explains a great many confusing reports. The comparison below covers what each model does, what it is good at, and where it fails.

For the large majority of UK accounts with more than one campaign type and a sales cycle longer than a single session, data-driven attribution is the better default. The reasoning is not that the model is clever — it is that fractional credit makes upstream work visible at all, and last click makes it structurally invisible. A model that is approximately right about several touchpoints beats a model that is precisely wrong about one.

Two caveats are worth stating plainly, because vendors rarely volunteer them. First, the opacity is real. You cannot audit a data-driven allocation, cannot reproduce it, and cannot explain to a finance director exactly why a campaign received 0.34 of a conversion. If your organisation requires defensible numbers — regulated sectors, board-level scrutiny, agency accountability disputes — you will need a reconciliation layer of your own alongside it. Second, retrospective restatement genuinely disrupts reporting. Data-driven credit for a given month can and does move after the month closes, which makes month-on-month comparisons noisier and makes anyone comparing a report against a screenshot from three weeks ago suspect something is broken. Neither caveat argues against the model; both argue for explaining it before you switch, not after the first restated report lands on someone’s desk.

Attribution readiness — where UK accounts typically fall short

Before changing a model, it is worth knowing whether the underlying data can support any model at all. The grid below lists the findings that recur most often in UK conversion tracking audits, graded by how much they distort attributed results. Items marked high risk will produce misleading campaign rankings regardless of which model is selected; items marked lower will degrade precision without inverting conclusions.

Two of these are worth expanding because they account for a disproportionate share of genuinely wrong reporting. The first is auto-tagging. If auto-tagging is off — usually switched off years ago by someone who preferred manual UTM parameters — then no GCLID is appended to landing page URLs, no click identifier reaches the CRM, and offline conversion import is impossible until it is switched back on. It is a single toggle in account settings, and turning it on today does nothing for the historic data, which is why it should be checked on day one of any attribution programme rather than in week six.

The second is the counting setting. Google Ads lets a conversion action count “every” conversion or “one” per click. For e-commerce, “every” is correct because two orders are two orders. For lead generation it is usually wrong, because one person filling in three forms while comparing options is one lead, and counting them as three tells Smart Bidding to pursue indecisive researchers. A high proportion of UK lead-generation accounts inherit the wrong setting, and it inflates reported conversions in exactly the campaigns that attract early-stage, high-comparison traffic.

A realistic attribution implementation timeline

The sequence below is what a full attribution programme actually looks like for a UK mid-market advertiser, from first audit through to a reporting layer that finance will accept. The calendar assumes a competent development resource available for roughly one day a week and a CRM owner willing to take part. Compress it if both are dedicated; extend it — realistically by half again — if the CRM sits with a third party who bills for changes.

The week 11 instruction to hold is the one most often ignored and the one that most reliably ruins the programme. Data-driven attribution changes the signal Smart Bidding optimises against, and Smart Bidding responds over a period of days to weeks, not hours. Accounts that switch the model and immediately begin adjusting budgets in reaction to the first week’s numbers are reacting to the transition itself rather than to any steady state, and they typically conclude that the model made things worse.

Measurement maturity benchmarks across UK advertisers

The rows below show roughly what proportion of UK business Google Ads accounts have each capability in place. They are useful as a positioning exercise: the items near the top are table stakes and their absence is a defect, while the items near the bottom are genuine differentiators where being in the minority is a competitive advantage rather than a gap to feel bad about.

The steep fall between “GA4 linked” at 79% and “GCLID stored in the CRM” at 24% is where the attribution problem actually lives. Everything above that line happens inside marketing tools and can be delivered by a marketing team alone. Everything below it requires the CRM, which usually means a different owner, a different budget and a different change-approval process. That organisational boundary — not any technical difficulty — is why the majority of UK lead-generation advertisers still optimise toward form fills they cannot value.

The 48% figure for conversion values is the cheapest gap on the list to close. Assigning values does not require perfect data. If your qualified leads close at 22% and your average deal is £9,000, a qualified lead is worth roughly £1,980 and putting that number in is enormously better than leaving the field empty, because it lets value-based bidding distinguish a £60 enquiry from a £20,000 one. Estimated values are refined later by offline import. Absent values cannot be refined at all.

The conversions that never reach Google Ads at all

Attribution can only distribute credit among the conversions it can see. Before debating how to divide the pie, it is worth knowing how much of the pie never arrives. The figure below represents the share of genuine conversions that go unrecorded in a typical UK account with standard client-side tagging, no enhanced conversions, and no offline import — lost to consent refusal, browser storage limits, ad blocking, cross-device journeys, phone calls and sales that complete outside the website.

Thirty-seven per cent is a representative figure rather than a universal one, and the composition matters more than the total. Roughly a third of the loss is consent-related: users who decline cookies produce no client-side conversion signal, and without consent mode v2 supplying the behavioural signals that Google models from, those conversions are simply absent rather than estimated. Another third is journey shape — a user who researches on a phone and buys on a work laptop breaks the client-side link entirely. The remainder is the offline tail: enquiries that become phone calls, quotes, site visits and eventually invoices, none of which the browser ever hears about.

The practical significance is that this missing share is not distributed evenly across campaigns. Consent refusal correlates with certain audiences; cross-device journeys correlate with considered, expensive purchases; offline completion correlates almost perfectly with the campaigns targeting high-value commercial intent. In other words, the conversions you cannot see are concentrated in exactly the campaigns that matter most, which biases every attribution comparison against your most valuable activity before any model is applied. Fixing capture is therefore prior to fixing attribution, and any programme that reverses that order is optimising a distorted picture with great precision.

It is also the point where measurement becomes a data-governance question rather than a marketing one. Hashing customer email addresses to recover conversions is entirely workable under UK GDPR, but it involves processing personal data for a purpose that must appear in your privacy notice and your record of processing activities, and the hashing must genuinely happen before transmission. Businesses that have already thought carefully about where customer data flows — the same thinking that governs Microsoft 365 and Copilot data security — find this straightforward. Businesses that have not tend to discover the question halfway through implementation, when it is more disruptive to answer.

The 12-point Google Ads attribution checklist

Work through these in order. Each one is verifiable in an afternoon by someone with account access, and the sequence is deliberate — later items depend on earlier ones being true. If you stop after item six you will still have a materially better account than most.

1. Confirm auto-tagging is enabled. Account settings, one toggle. Without it there is no GCLID, and roughly half of everything below becomes impossible. Then click a live ad and confirm the parameter actually survives to the landing page rather than being stripped by a redirect.
2. List every conversion action and identify duplicates. Two actions recording the same event is the most common cause of inflated conversion counts. Check the last-recorded date too — dormant actions still counted as primary quietly distort Smart Bidding.
3. Reduce primary conversion actions to one per genuine objective. Everything else becomes secondary. Secondary actions still report; they just stop instructing the bidding algorithm to chase newsletter signups when you sell £40,000 installations.
4. Fix the counting setting. “Every” for transactions, “one” for lead-generation actions. Getting this wrong systematically over-rewards campaigns that attract repeat form-fillers.
5. Set conversion windows to match observed lag. Look at the time-lag report before choosing. If a fifth of conversions arrive after 30 days, a 30-day window is discarding a fifth of your evidence and penalising the campaigns that open long journeys.
6. Assign a value to every primary action. Estimated values calculated from close rate multiplied by average deal size are entirely acceptable. An empty value field makes value-based bidding impossible.
7. Verify consent mode v2 is passing the right parameters. Check that ad_user_data and ad_personalization are being transmitted with correct states, and that the banner is not blocking the tag outright before any signal can be sent. Confirm the implementation aligns with your PECR obligations and ICO guidance on cookie consent.
8. Enable enhanced conversions. Web for online transactions, leads for offline closes. Confirm hashing is SHA-256 and occurs before data leaves the browser, and add the processing to your privacy documentation.
9. Store the GCLID against the lead record in the CRM. Hidden form field, populated from the URL or first-party cookie, persisted through record merges and deal creation. Test end to end with a real ad click.
10. Build the offline conversion import. Start with one deal stage and one conversion action. Wait at least six hours after the click before uploading, and always upload inside the 90-day GCLID validity window.
11. Switch primary actions to data-driven attribution and hold. Change nothing else for three weeks so the effect is isolable and Smart Bidding has a stable signal to learn against.
12. Add tracking verification to your website release checklist. Every template change, form plugin update and consent vendor upgrade is a candidate for silent breakage. Automated or manual, this check is what keeps the other eleven items true in six months’ time.

Score your own attribution setup

The gauge shows where the median UK business account sits when scored against the twelve-point checklist above, weighted toward the items that affect budget decisions most. Score yourself by awarding points for each item genuinely in place — verified this month, not assumed — rather than each item someone once said was done.

A score in the forties or fifties is normal and not a crisis. It typically describes an account where conversion tracking works, GA4 is connected, and everything downstream of the website has never been attempted. The gap between 51 and 80 is almost entirely composed of three items: conversion values, enhanced conversions, and GCLID capture in the CRM. None of them requires a platform migration, a new agency or a large budget. All three require someone to own the work across the marketing and CRM boundary, which is the actual scarce resource.

Scores below 35 usually indicate a structural problem rather than a list of missing features — most often an account being managed by a party who does not have access to the website or the CRM, and therefore cannot implement anything beyond the ads themselves. If that is the situation, the correct first action is not an attribution project. It is fixing the access and accountability arrangement, because no amount of measurement design survives being unable to change the systems that produce the data.

Scores above 80 introduce a different failure mode: over-confidence. An account with full offline import and data-driven attribution produces numbers that look authoritative enough to stop questioning, and the incrementality gap described at the top of this guide has not gone anywhere. Sophisticated attribution tells you more precisely which ads converting customers touched. It still cannot tell you which of them would have bought anyway, and the businesses that get this wrong are usually the ones that got everything else right.

Common attribution mistakes to avoid

These are the recurring errors in UK accounts, ordered roughly by how much money they move. Most are not technical failures; they are reasoning failures made confidently on the basis of correct numbers read the wrong way.

• Cutting generic search on last-click evidence. The single most expensive mistake in UK paid search. Generic terms open most converting paths and close few, so they will always look inefficient under last click. The damage from cutting them appears one full sales cycle later, disguised as a branded search decline.
• Comparing campaigns measured under different models. Attribution is set per conversion action. If your lead form uses data-driven and your phone call action still uses last click, campaign totals mix the two and no comparison between them is valid. Audit the model on every action, not on the account.
• Treating the GA4 number and the Google Ads number as a discrepancy to eliminate. They count different things over different windows with different identity resolution. GA4 attributes across all channels; Google Ads attributes within Google Ads. A stable, explainable gap is a healthy setup. Chasing an exact match wastes months and usually ends with someone breaking a working tag.
• Switching the model and restructuring campaigns in the same fortnight. Guarantees you cannot attribute the resulting change to either action, and denies Smart Bidding the stable signal it needs to re-learn. Change one thing, then wait three weeks.
• Reading Performance Max campaign totals as a performance verdict. Performance Max spans the entire path inside one line item, so it collects touchpoints everywhere and looks strong under any model. Evaluation requires segmenting the channel distribution inside it, not reading the aggregate.
• Optimising toward undifferentiated lead counts. If every form fill is worth the same in the account, Smart Bidding will find you the cheapest form fills available, which are reliably the least valuable ones. This is the failure that offline conversion import exists to solve.
• Letting tracking decay after the project ends. Attribution setups do not degrade gracefully; they break at a website release and stay broken until someone notices a step-change in a chart. Without a verification step in the release process, a six-week implementation has a working life of about nine months.
• Assuming attribution proves incrementality. It does not, at any level of sophistication. Only a holdout test, geographic experiment or budget-split experiment can tell you whether spend caused sales that would not otherwise have happened. Attribution allocates credit among observed clicks; incrementality asks whether the clicks mattered.

A real-world example — a Leeds equipment supplier

A 45-person industrial equipment supplier near Leeds was spending roughly £18,000 a month across search, shopping and Performance Max, selling machinery with an average order value of about £14,000 and a sales cycle running six to ten weeks. Reporting was last-click, conversions were undifferentiated form fills and phone calls, and the quarterly review had reached a familiar conclusion: generic search was costing £190 per conversion against £24 for branded, and the generic budget should move to branded and shopping.

The audit found three things before any model was touched. Auto-tagging had been disabled by a previous agency in favour of manual UTM tagging, so no GCLID had ever reached the CRM. The phone call conversion action and the form action were both counting “every”, so a customer who called twice and filled in a form registered three conversions. And a brochure download had been set as a primary action in 2024, meaning Smart Bidding had been optimising toward brochure downloads for over a year.

Fixing the counting layer alone reduced reported conversions by 31% — a number that had to be explained carefully to the board, because nothing had got worse; the previous figure had simply been wrong. Auto-tagging was re-enabled, GCLID capture was added to the CRM, and eleven weeks later offline conversion import was feeding qualified-opportunity and closed-won stages back to Google Ads with real deal values attached. Primary actions moved to data-driven attribution at that point.

The reallocation that followed inverted the original conclusion. Under data-driven attribution with revenue values, generic search carried 41% of attributed revenue against 19% of attributed conversions — it was opening the expensive journeys and closing few of them. Branded search, which had looked like the most efficient line in the account, carried 12% of revenue on 34% of conversions. The proposed budget move would have redirected spend from the campaigns generating the pipeline to the campaigns collecting it.

We were four weeks from cutting the campaigns that were actually feeding the business. The uncomfortable part was that nobody had done anything careless — we made a rational decision on the numbers we had. The numbers were just answering a different question from the one we were asking.

Two details from that engagement generalise. First, the largest single gain came from correcting counting settings, not from the attribution model — the model change refined a picture that basic hygiene had already transformed. Second, the work stalled for three weeks at the CRM boundary, waiting on a third party to add a hidden form field, which is the same organisational bottleneck visible in the 24% GCLID capture figure earlier in this guide. Neither of those is a technology problem, and both are predictable enough to plan around.

Google Ads attribution at a glance

The reference table below collects the operational facts scattered through this guide — the constraints, defaults and figures worth having to hand when you are inside the account rather than reading about it.

How Cloudswitched approaches attribution work

Cloudswitched works on Google Ads attribution as a measurement engineering problem rather than a reporting preference. That means the audit comes first and is delivered as a specific list of defects with the evidence attached, the counting layer is corrected before any model is changed, and the offline conversion import is treated as the objective rather than an optional extra — because for most UK lead-generation businesses it is the piece that turns campaign reporting into something a finance director can use. Because Cloudswitched also delivers web development, CRM integration and infrastructure work, the changes that live outside the ads platform — form fields, tag deployment, consent implementation, CRM API work — can be specified and built by the same team, which closes the organisational gap where most attribution projects stall. Where landing page changes form part of the work, the same standards applied in a WCAG accessibility compliance review apply to the pages being measured, since a form that is difficult to complete is a measurement problem as well as a usability one.

Frequently Asked Questions

Which Google Ads attribution model should I use in 2026?

For most UK accounts with more than one campaign type and a sales cycle longer than a single session, data-driven attribution is the better default. It assigns fractional credit across the touchpoints on a converting path, which makes upstream campaigns visible and fundable rather than reading as zero-value. Last click remains defensible for genuinely single-session purchases, for accounts running one campaign type, and for organisations that need a fully reproducible number for governance reasons. Note that these are the only two options remaining — first click, linear, time decay and position-based were retired across 2023 and 2024. The model is set per conversion action, so check every action rather than assuming the account has one setting.

Why do my Google Ads and GA4 conversion numbers not match?

Because they are counting different things. Google Ads attributes conversions to the date of the ad click and only considers Google Ads interactions; GA4 attributes to the date of the conversion and considers all channels. They use different identity resolution, different default windows, and different modelling for unconsented users. A stable, explainable gap between the two is the expected state of a healthy setup, not a defect. The productive goal is to understand the size and composition of the gap and confirm it stays consistent month to month. Attempting to force an exact match typically consumes months and ends with someone disabling a tag that was working correctly.

What is offline conversion tracking and do I need it?

Offline conversion tracking, usually implemented as offline conversion import, feeds outcomes that happen away from your website — a qualified opportunity, a signed contract, an invoiced amount — back to the specific ad click that produced the enquiry, using the GCLID stored on the lead record. You need it if the value of a lead varies materially and the sale completes off-site, which describes most B2B, professional services, healthcare, construction and high-value retail businesses in the UK. Without it, Google Ads optimises toward form fills that all look identical to it, and Smart Bidding will reliably find you the cheapest ones. With it, the platform can bid toward revenue rather than volume.

How long does it take to see the effect of switching attribution model?

Allow three to four weeks before drawing conclusions, and change nothing else during that period. Switching to data-driven attribution alters the signal Smart Bidding optimises against, and the algorithm adjusts over days to weeks rather than hours. The first week after a switch reflects the transition itself rather than any steady state. Reported historic figures will also be restated as the model recalculates, which makes comparisons against screenshots taken before the change misleading. The most common way to waste an attribution project is to switch the model and simultaneously restructure campaigns or adjust budgets, making the effect of either impossible to isolate.

Does data-driven attribution work on low-volume accounts?

Google removed its published minimum thresholds in 2021, so data-driven attribution is technically available regardless of volume. In practice, allocations on very low-volume conversion actions are unstable and can move noticeably month to month, which makes them a poor basis for budget decisions even when the model is doing its job correctly. If a conversion action generates fewer than roughly thirty conversions a month, consider consolidating related actions, or accept that the model output is directional rather than precise. Low volume is also a reason to prioritise conversion values and offline import first — adding value information to a thin dataset improves decisions more than refining how credit is split within it.

What is enhanced conversions and is it compliant with UK GDPR?

Enhanced conversions recovers conversions that client-side tracking loses by sending hashed first-party data — typically an email address, hashed with SHA-256 in the browser before transmission — so that Google can match a conversion to a click it could not otherwise link. It is workable under UK GDPR, but it is not automatically compliant simply because the data is hashed. The processing needs a lawful basis, must appear in your privacy notice and record of processing activities, and the hashing must genuinely occur before the data leaves the page. Treat it as a data-processing decision requiring the same review as any other customer-data flow, and document it accordingly rather than switching it on as a tagging change.

How does consent mode v2 affect attribution in the UK?

Consent mode v2 sends signals about the user’s consent state — including ad_user_data and ad_personalization — alongside your tags, allowing Google to model conversions for users who declined cookies rather than losing them entirely. In the UK, cookie consent falls under PECR with ICO guidance on how consent must be obtained, and the underlying personal data processing falls under UK GDPR. Practically, a correct implementation recovers a meaningful proportion of otherwise-invisible conversions; an incorrect one — most commonly a banner that blocks tags outright before any signal is sent — loses both the conversion and the modelled estimate. It is worth verifying rather than assuming, because the failure is silent.

Should I use last click for branded search and data-driven elsewhere?

You cannot, because the model attaches to the conversion action rather than the campaign, and branded and generic campaigns typically share the same conversion actions. The instinct behind the question is sound, though: branded search does behave differently and is systematically over-credited under last click. The right response is to segment branded from generic in your reporting and evaluate them against different expectations, rather than trying to apply different accounting rules to the same conversion. Many UK advertisers also run a periodic branded holdout test to estimate how much branded conversion volume is genuinely incremental, which answers the underlying question far better than any model choice could.

What does a Google Ads attribution audit typically cost in the UK?

A conversion tracking and attribution audit generally runs £1,200 to £2,500 for a mid-market account with one domain and one CRM, delivered in one to two weeks. Implementation is priced separately: enhanced conversions around £800 to £1,800, consent mode work £1,500 to £3,500, and offline conversion import £2,500 to £7,500 depending on the CRM and how accessible its API is. Buy the audit first regardless of what else you intend to do. A significant proportion of accounts turn out to have duplicate conversion actions or misconfigured counting settings, and no attribution model applied on top of miscounted data produces a trustworthy answer.

Can attribution tell me whether my Google Ads spend is profitable?

Not on its own. Attribution allocates credit among the ad interactions that converting customers actually touched, which is a different question from whether those customers would have bought anyway. Branded search illustrates the limit clearly: it attributes superbly under every model because people who already intend to buy search your name first, and a share of them would have found you regardless. Only an experiment — a holdout, a geographic split, or a structured budget test — can estimate incrementality. Use attribution to allocate budget between comparable campaigns, and use experiments to decide whether the channel or a major segment of it is earning its place at all.

How often should attribution and conversion tracking be reviewed?

Verify conversion tracking at every website release, and run a full review quarterly. Attribution setups rarely fail loudly — they break when a template changes, a form plugin is replaced or a consent vendor updates, and the loss usually surfaces weeks later as an unexplained step-change in a chart. A quarterly review should re-check counting settings, window configuration, model assignment on each conversion action, GCLID capture end to end, and whether the offline import is still reconciling against CRM revenue. Without a verification step in the release process, a well-built implementation has a practical working life of roughly nine months before something quietly stops reporting.

What should I fix first if I can only do one thing?

Audit and correct your conversion actions — deduplicate them, set one primary action per objective, fix the counting setting, and assign a value to each primary action even if the value is estimated from close rate multiplied by average deal size. This costs the least, requires no development work in most cases, and improves both your reporting and the signal Smart Bidding optimises against. It also frequently changes the conclusion of the budget debate on its own, before any attribution model is touched. Model selection matters, but it refines a picture that basic counting hygiene has to get approximately right first.

Related reading

These guides cover the adjacent measurement, planning and data-governance questions that attribution work tends to raise once it is under way.

• Technical SEO Audit: A UK Business Guide — how organic visibility is diagnosed, and why aggregate tool scores mislead in the same way aggregate campaign totals do.
• IT Roadmap and Technology Strategy for UK Businesses — sequencing measurement and systems work so the CRM boundary does not stall the project.
• Microsoft 365 Copilot Data Security — the UK GDPR thinking that applies equally to hashed customer data used for enhanced conversions.
• Website Accessibility Compliance and WCAG — landing page and form standards that affect both conversion rate and what you are able to measure.
• Proactive Network Monitoring for UK Businesses — the same principle applied to infrastructure: silent failures are found by verification, not by waiting for a complaint.