By Tom Kató & László Bali · ppcout.com · 15 minute read
Ask most store owners whether their conversion tracking works and they'll say yes — the dashboard shows purchases, the numbers move, the ROAS column is populated. But "shows purchases" and "shows the right purchases, once each, with the right values" are very different claims. In our audits, genuinely accurate purchase tracking is the exception, not the rule: most e-commerce accounts undercount, double-count, or misvalue a meaningful share of their orders without anyone noticing.
This matters more than almost anything else in the account, because every automated decision runs on this data. Smart bidding funds the campaigns your tracking says perform; if the tracking is wrong, the algorithm confidently funds the wrong things. A 20% tracking gap doesn't just misstate your reports — it systematically misallocates your budget, every day, at machine speed.
This guide covers why purchase tracking breaks so often, what a reliable setup actually consists of, why revenue values matter as much as conversion counts, how bad data corrupts smart bidding, and how to audit your own tracking against your order backend. If you'd rather have this checked by senior specialists, tracking verification is the first thing a professional Google Ads audit does — because nothing else in the account can be judged until the data is trustworthy.
E-commerce purchase tracking has a hard job: it has to fire exactly once, on the right event, with the right value, at the exact moment a customer completes an order — across every browser, device, payment method, and consent choice. Each of those variables is a place it can fail, and most stores have several failing at once.
In European markets, a meaningful share of visitors decline cookies. Without a correctly configured Consent Mode, every one of their purchases is simply invisible to your ad platform — the order happened, the ad drove it, and your account never finds out. Depending on your market and banner design, that alone can hide 15–40% of real conversions. With Consent Mode set up properly, Google can model a portion of these lost conversions back; without it, the gap is silent and permanent.
Purchase tags typically fire on the confirmation page. But customers paying via external gateways — PayPal, bank transfer schemes, buy-now-pay-later providers, local payment methods — sometimes never return to that page. They close the tab after paying, or the redirect fails, or the gateway's return flow skips your confirmation URL. The order is real; the tag never fired. Stores with a high share of redirect-based payments routinely undercount by 5–15% from this alone, and the loss is invisible unless you compare against backend orders.
Tracking doesn't only undercount. A confirmation page that customers revisit — from an email link, a browser restore, a refresh while waiting — fires the purchase tag again unless deduplication is in place. Multiple tags measuring the same event (an old hardcoded tag plus a Tag Manager tag plus a platform integration) is the other classic: one order, two or three reported conversions. Overcounting is arguably worse than undercounting, because it flatters your results and hides the problem behind good news.
The final structural cause is organizational: tracking accumulates in layers as stores change platforms, themes, agencies, and plugins. Each era leaves tags behind, and after a few years nobody can say with confidence what fires where, or why. A migration from one shop system to another is the classic breakage event — the new platform's native integration goes live while the old Tag Manager setup keeps half-working — and the resulting hybrid produces numbers that are wrong in ways too inconsistent to spot from the dashboard. If your store has been through a replatforming and tracking was never formally re-audited afterwards, assume drift until proven otherwise.
Reliable purchase tracking isn't one tag — it's a small system with four load-bearing parts. Miss any one and the numbers drift.
Every purchase event should carry the order's transaction ID, and your Google Ads conversion action should deduplicate on it — so a refreshed confirmation page or a second tag can't double-report the same order. Just as important: audit for competing tags. Years of platform migrations and agency handovers leave most stores with legacy tracking still firing somewhere. One purchase event, defined once, deduplicated by ID, is the goal.
A properly implemented Consent Mode (v2 in European markets) respects the user's choice while sending Google the anonymous signals that allow conversion modeling — recovering a statistically estimated share of consent-lost purchases. Enhanced Conversions complements it by sending hashed first-party data (email, phone) from the order, letting Google match conversions that cookies alone would miss. Together they close a large part of the privacy-era gap. Neither is optional anymore for a store that wants its automated bidding to see reality.
Purchases can reach Google Ads two ways: the native Google Ads tag, or imported GA4 events. Both work; using both at once for bidding double-counts. The native tag is generally the better bidding source — it's faster, supports Enhanced Conversions most directly, and isn't subject to GA4's attribution processing. Keep GA4 as your analytical view, set the imported event to secondary, and let one clean primary conversion action define success for bidding.
Server-side tagging moves the measurement from the customer's fragile browser to your controlled environment, making it resilient to ad blockers, browser tracking prevention, and redirect losses. It's more setup and maintenance, so it's not step one — but for stores spending serious budgets, where each percentage point of data accuracy steers real money, it's the logical endpoint of the tracking roadmap.
Counting purchases correctly is half the job. For an e-commerce account running value-based bidding — target ROAS, maximize conversion value — the value attached to each purchase is what the algorithm actually optimizes. Wrong values mean the machine pursues the wrong orders with perfect confidence.
There's no single correct answer to whether conversion values should include VAT and shipping — but there is a correct principle: consistency between the value you track and the target you set. If your ROAS target was calculated on net product revenue, but your tags send gross values including VAT and shipping, your real economics are quietly worse than your dashboard claims. Multi-currency stores add another layer: values must arrive in a consistent currency, correctly converted, or cross-market comparisons become fiction. Pick a definition, document it, and align tags and targets to it.
Discounts and vouchers deserve the same scrutiny. Many platform integrations send the pre-discount value, quietly inflating every reported order from a promotion — which means precisely during your sales, when spend is highest, your data is at its most optimistic. Verify that the value your tag sends matches what the customer actually paid, and test it with a discounted order, not just a full-price one.
Your tracking reports the order the moment it's placed — it never hears about the cancellation two hours later or the return two weeks later. For most stores this is a tolerable rounding error; for categories with high return rates (fashion above all), it's a systematic overstatement of 10–30% that makes every ROAS figure optimistic. Google Ads supports conversion adjustments for retractions and value corrections; at minimum, know your return rate by category and mentally discount reported revenue accordingly when setting targets.
Revenue-based values treat a €100 low-margin order and a €100 high-margin order as identical wins. They aren't. Stores that send margin-adjusted values — or approximate it with value rules and margin-banded campaign splits — point the algorithm at profit instead of turnover. It's the difference between an account that maximizes revenue and one that maximizes the reason you run the business. This is advanced work, but it's where value tracking is heading, and the accounts that get there first compound the advantage.
A note on secondary events: add-to-cart, begin-checkout, and similar micro-conversions belong in your measurement as secondary conversion actions — useful diagnostics for funnel analysis, and occasionally a bidding aid for very low-volume accounts. What they must never be is primary alongside purchases: an algorithm told that a cart addition and a completed order are both "conversions" will cheerfully optimize toward the cheap, hollow one. Purchases with accurate values are the objective; everything else is instrumentation.
The insidious thing about measurement problems is how normal they look. The dashboard still shows conversions — just fewer, more, or differently valued than reality — and it attributes them unevenly across campaigns. The algorithm, optimizing against this partial picture, then makes systematically wrong decisions that no amount of bid or creative work can fix.
Consider a store running two campaigns. Campaign A's conversions track cleanly; Campaign B sells more via redirect-heavy payment methods, so a slice of its purchases never report. The algorithm sees A outperforming B, and shifts budget accordingly — even if B was actually the more profitable campaign. Weeks of spend get misallocated, and the "optimization" the system performed made the account worse. Nothing in the interface would ever surface this; only comparing against backend orders per campaign would.
This misallocation isn't a one-time cost. The algorithm learns from the traffic it buys, so distorted data buys distorted traffic, which generates more distorted data. The loop runs the other way as well: accounts that fix their measurement watch every downstream layer improve over the following months — structure decisions get sharper, targets become meaningful, tests become believable. This is why tracking is the first thing to fix and the last thing to skimp on: it's upstream of every euro the algorithm moves.
Even with perfect tags, Google Ads and GA4 will report different numbers — by design. Google Ads records a conversion against the date of the ad click and attributes on its own model; GA4 records it on the purchase date and attributes across all channels. Neither is wrong; they answer different questions. The practical implications: never judge the last few days of Google Ads data, because conversions are still trickling in against earlier clicks; expect GA4 to credit Google Ads with fewer purchases than the platform claims itself; and pick one system as your decision-making lens per question instead of averaging two different truths into confusion.
You don't need to trust your tracking — you can test it. The method is straightforward, takes an afternoon, and most stores have never done it.
Pick the last 30 days. Pull real completed orders and revenue from your shop backend — the source of truth. Compare against GA4 purchase events and Google Ads conversions for the same window, adjusting for attribution differences (Google Ads counts conversions on the click date, GA4 on the event date). You're looking at two gaps: the count gap and the value gap. A discrepancy within roughly ±5% is healthy. Beyond ±10%, your bidding is running on meaningfully wrong data and the gap needs diagnosis.
If the gap is real, segment it. Compare by payment method (redirect losses show up immediately), by device, by market (consent behavior differs), and check for duplicate transaction IDs in GA4 (overcounting). Run a test purchase through each payment flow with a tag debugger open and watch whether the purchase event fires exactly once with the right value. Each segment that deviates points at its own fix — gateway return URLs, Consent Mode configuration, deduplication, or a legacy tag to remove.
Don't skip the value gap while chasing the count gap. Sum the revenue your tracking reported for the window and compare it to backend revenue for the same matched orders: counts can align while values drift, thanks to currency handling, discount codes, or VAT inconsistencies. A store that counts every order but overstates each one by 12% is still steering its bidding with a 12% lie.
Tracking decays like everything else: shop platform updates, theme changes, new payment methods, consent banner tweaks — any of them can silently break a tag that worked for years. Put the backend comparison on a monthly calendar and treat any drift beyond your tolerance band as an incident, not a curiosity. Stores that monitor this catch breakage in days; stores that don't discover it in quarterly reviews, after the algorithm has spent months optimizing against fiction.
Finally, write the setup down. A one-page measurement plan — which events exist, where each fires, what value definition they carry, which conversion action is primary for bidding, and who owns the monthly check — turns tracking from tribal knowledge into an asset that survives agency changes and staff turnover. Half the tracking chaos we find in audits exists simply because no document ever said what "correct" was supposed to look like.
If you'd rather have senior specialists run this verification — and map what the gaps have been costing you — that's precisely the first phase of a professional Google Ads audit. Or start with a free Google Ads audit for a first read on whether your numbers can be trusted.
Within roughly ±5% is healthy — attribution windows, consent losses, and modeling make perfect parity impossible. Between 5–10% is worth investigating. Beyond ±10%, your smart bidding is optimizing against meaningfully wrong data, and the gap should be diagnosed by payment method, device, and market to find the leak.
Pick one, and the native Google Ads tag is usually the better choice: it's faster, supports Enhanced Conversions most directly, and avoids GA4's attribution processing. Keep GA4 purchases imported as a secondary conversion for analysis. Using both as primary double-counts every order and inflates your reported results.
Enhanced Conversions sends hashed first-party data from your order — like email address — alongside the conversion, letting Google match purchases that cookies alone would miss. In a privacy-first landscape with consent banners and tracking prevention, it recovers a real share of otherwise-lost conversions. For any store spending meaningful budget, it should be considered standard equipment.
Depending on market and banner design, 15–40% of visitors typically decline. Without Consent Mode, all of their purchases are invisible to your ad platform. With Consent Mode v2 correctly implemented, Google models a statistically estimated share of those conversions back — imperfect, but far closer to reality than a silent gap.
The defensible answer is: whichever definition your ROAS target was calculated on. The failure mode isn't a wrong choice, it's inconsistency — gross values measured against a net-based target quietly overstates your economics. Most stores are best served tracking net product revenue and setting targets on the same basis. Document the definition and align everything to it.
Almost always duplication: a confirmation page that re-fires on refresh or revisit without transaction-ID deduplication, or multiple tags measuring the same purchase — a legacy hardcoded tag plus a Tag Manager tag plus a platform integration. Audit what fires on your confirmation page and enforce one purchase event, deduplicated by order ID.
Not automatically — tracking reports orders when placed and never hears about what happens after. In high-return categories like fashion, that overstates real revenue by 10–30%. Google Ads conversion adjustments can retract or revalue conversions; at minimum, know your return rate by category and account for it when setting targets.
Run the backend comparison monthly — it takes minutes once set up — and additionally after any platform update, theme change, new payment method, or consent banner modification, since these are the events that silently break tags. Treat drift beyond your tolerance band as an incident to fix, not a curiosity to note.
Online marketing and PPC specialist focused on Google Ads and advertising strategy — the kind that builds not just clicks, but brands. With 10+ years in digital marketing and e-commerce, Tom leads on strategy and measurement, turning strategic scaling and zero-click trends into measurable business results.
in
Performance marketing specialist with deep hands-on Google Ads and e-commerce experience. László leads on campaign execution and growth, building and scaling accounts for e-commerce brands and small businesses — the same senior specialist on your account from day one, not a junior and a dashboard.
in