Why your ad conversions do not match your sales in 2026

The two figures are not counting the same thing. An ad platform counts a conversion when a browser reports that a tracked action happened, while your accounts count money that arrived, after refunds, cancellations and failed payments, which is why a gap in the region of 10% to 30% is normal.

Why do ad platforms report more conversions than you had sales?

Your accounts count cash. An ad platform counts events it can plausibly tie to its own advertising, and it has no visibility of what any other channel did.

A worked example: a customer clicks a Meta ad on Monday and does not buy. On Wednesday they search for the product again, click a Google search ad and buy a £300 item. Meta records a conversion worth £300. Google records a conversion worth £300. Your accounts record £300 once. Two platforms, two conversions, one order.

Platforms also count things that are not sales. A quote request, a newsletter sign-up, an add to cart or a phone call over 60 seconds can all be defined as a conversion. In a platform report that looks identical to a purchase.

Different platforms also use different rules for credit. A click-based model credits the last advert clicked. A view-based model credits an advert somebody saw but never clicked. Both are defensible, and neither matches how the customer would describe their own decision.

Refunds and cancellations rarely travel back to the platform. A sale reported in week one that is returned in week three usually stays in the platform’s totals.

How does an ad platform decide that a conversion happened?

A tag, which is a short piece of JavaScript placed on your site, runs when a page loads and sends a message to the platform. To connect that message to an ad click, the platform needs an identifier, normally a cookie set in the browser at the moment the ad was clicked, or a click identifier added to the destination URL.

Those identifiers are visible in your web analytics as parameters such as gclid for Google and fbclid for Meta. Finding them on your landing page reports is a quick way to confirm the chain is intact. Break either link and the conversion is either not recorded or recorded with no click behind it, so it lands in the platform as direct or organic traffic.

Consent banners sit in the middle of this. A tag is supposed to wait until the visitor has agreed to being tracked. If they decline, some platforms fill the gap with modelled conversions, which are statistical estimates built from the behaviour of similar users rather than observations of that person. Modelled conversions are reasonable inputs for automated bidding and poor material for a report you intend to reconcile against your bank.

What usually breaks the tracking chain?

Six causes account for most of the discrepancy:

• A declined consent banner, so the identifier is never stored.
• Browser restrictions. Safari and Firefox block third-party cookies by default and limit how long cookies written by scripts survive, so a Monday click may be unreadable by the following week.
• Ad blockers and privacy extensions, which stop the tag loading at all.
• Cross-device journeys. A click on a phone in the evening and a purchase on a work laptop the next morning look like two unrelated visits.
• Technical faults. A tag missing from one page template, a thank-you page on a separate domain, or the same tag firing twice will all distort the total.
• Definition drift. The platform counts a quote request while your accounts count paid invoices, and nobody notices the two numbers were never meant to match.

How much of a gap is normal?

Typically 10% to 30%, and the direction matters. Platforms usually over-report because of double counting, but under-reporting is common too when consent rates are low or a tag has broken.

A gap wider than roughly a third in either direction usually points to a fault rather than ordinary leakage. At low volumes, percentages mislead: a business taking 40 orders a month only needs two or three unattributed orders for the gap to swing wildly, so judge over at least a full quarter.

Two checks are worth running monthly: whether the platform total sits above or below your sales, and whether the size of the gap has changed suddenly. A stable gap is a reporting convention. A sudden one is usually a broken tag or a consent banner update.

What is an attribution window, and how does it change the numbers?

An attribution window is the period after a click or a view during which a sale is credited to that advert. Common defaults sit around 30 days for clicks and 1 day for views, and most platforms let you shorten the click window to 7 days.

A longer window always reports more conversions, because it catches slower buyers. A 90-day window on a considered purchase will look impressive and will also claim sales you would have made without any advert. Changing the window mid-quarter rewrites the recent past, so any month-on-month comparison across that change is meaningless.

What is server-side tracking and when is it worth paying for?

In normal browser tracking, the visitor’s browser sends the message straight to the ad platform. With server-side tracking, the browser sends it to a container running on your own subdomain, and that container forwards it, sometimes with additional customer data attached. Because the request goes to your own domain, it is less likely to be blocked by browser restrictions or ad blockers.

It does not fix declined consent, cross-device journeys, refunds or definition drift. It improves the reliability of the signal that does get through, which matters most when you spend enough for automated bidding to need accurate feedback.

Costs are real. Setup commonly runs from £1,500 to £4,000 with an agency, or 15 to 40 developer hours done in house. Hosting on a cloud container is typically £40 to £150 a month, plus a few hours of maintenance. A business spending £3,000 a month on ads that recovers visibility on 10% of that spend has improved the signal on £300 of monthly spend, which is not the same as finding £300 of extra revenue.

What should a small business do about the gap, in order?

Write down what counts as a conversion, in one sentence, and make sure the tag and your accounts use the same definition. That single step removes the largest category of confusion and costs nothing.

Then feed real outcomes back to the platforms. Offline conversion imports in Google Ads and CRM-based conversion APIs in Meta let you send actual paid sales rather than form fills, so automated bidding optimises towards revenue. Keep your URL tagging consistent so you can see channel performance in your own analytics.

Judge the account on blended numbers: total ad spend for the month against total revenue from all sources, using your accounting data. Platform conversion counts are a bidding input, not a scoreboard.

Set a review date. If the gap is stable and your cost per sale from blended numbers is acceptable, the tracking work can wait a year without harming the business.

Does a gap mean your ads are not working?

No. If £3,000 of ad spend coincides with £12,000 of revenue across all channels, the ratio is the thing that matters, not whether the platform recorded 30 conversions or 45.

Server-side tracking is worth revisiting once you are spending enough that a 10% to 20% improvement in signal changes bidding decisions, commonly somewhere above £3,000 to £5,000 a month. Below that, written definitions and imported sales data will close more of the gap per pound spent.