Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Analytics › Troubleshooting

Channel Conversions Add Up to More Than Sales

Why conversions reported by each channel add up to more than real sales: shared credit, view-through, mixed windows, date logic, duplicates and refunds.
Troubleshootingintermediate8 min readUpdated

First, put the channels beside the orders

The symptom only exists when a total is compared with something real. Choose a closed period long enough to have settled, and put three columns side by side: each platform's or channel report's claimed conversions, their sum, and the orders or closed deals recorded in the back office for the same dates. Subtract the orders from the sum. That difference is the excess this page explains.

Two checks keep the comparison honest. Use net orders as the business counts them, after cancellations and refunds, and note which definition you used. And make sure every claimed figure covers the same dates in the same time zone. Then split each platform's claim by type where the platform allows it: click-based against view-based, and by window. That split usually accounts for much of the excess before any error is found. The step-by-step reconciliation method lives in the guide on why platforms and analytics disagree; this page diagnoses why the cross-channel total is inflated. The reverse symptom, an affiliate partner reporting sales that the platform never recorded, is traced in the tutorial on tracing missing affiliate conversions. For other measurement symptoms, start from the analytics troubleshooting hub.

Diagnosis table

Diagnosis

SymptomCheckLikely causeFix
The excess grows each time a platform is addedOrders that more than one platform claims, sampled from paths or order IDsEvery platform claims the same saleRead channels in one deduplicated view; stop summing platform totals
Upper-funnel platforms report many conversions but few clicksShare of each platform's claim that is view-basedView-through counted alongside clicksReport click and view claims separately
Platforms disagree most on long-consideration productsEach platform's click and view windows, and the analytics lookback windowDifferent conversion windowsAlign windows for comparison and record them
Last week's totals keep rising for days after the week endsWhether each report dates conversions by click or by orderConversions dated to the click in one report and the order in anotherCompare closed periods only, with the date basis noted
One source claims close to twice its share, sometimes on the same orderConversion events or tags firing more than once per orderDuplicate conversion eventsRemove the duplicate; deduplicate on order ID
Claims stay above net sales even after the aboveRefunds, cancellations and test orders in the claimed figuresRefunds, cancellations and test orders never removedSend refunds; exclude tests; report net

Fix, verify and prevent each cause

Every platform claims the same sale

A buyer who clicked a social ad, later searched the brand and clicked a search ad, and also opened a promotional email, will be counted as a conversion by all three, because each system checks only for its own touch. Google's conversion tracking documentation describes a conversion as an action recorded after someone interacts with an ad, and every other platform applies the same logic to its own ads. The platform attribution versus analytics guide explains why the sum of those claims has no meaning.

Fix: stop adding platform totals in reports; read channel performance in one view that gives each order once, such as the analytics property or the CRM, with platform figures kept alongside as labelled claims.

Verify: the deduplicated view totals close to the orders in the back office, with the remaining gap explained by the other causes.

Prevent: design the management report so that no row sums platform-reported conversions, and say on the report which system supplies each figure.

View-through conversions counted alongside clicks

Several platforms credit a conversion to an ad that the buyer saw but did not click, within a view window. Meta's help centre documents attribution settings that combine click-through and view-through windows, and video and display platforms commonly do the same. A view claim from one platform will often sit on an order that another platform claims from a click. The attribution models comparison explains how each model decides which touches count and why none of them measures cause.

Fix: split every platform's figure into click-based and view-based claims, report them in separate columns, and use click-based figures for cross-channel comparison.

Verify: the click-only sum falls much closer to real orders, and the size of the view-based column is known for each platform.

Prevent: record each platform's attribution setting in the measurement plan, and require a note in the report whenever a setting changes.

Different conversion windows across platforms

One platform may credit conversions that happen long after a click while another credits only those that happen soon after; the analytics property applies its own Lookback Window. On products with long consideration, a longer window claims more orders, and the platforms stop being comparable. The attribution settings guide shows where the analytics windows are set and what moving them does.

Fix: list every platform's click and view windows next to the analytics settings, and for comparisons, set each report to the nearest equivalent windows the tools allow.

Verify: with aligned windows, the gap between platforms narrows and the remaining differences are stable month to month.

Prevent: treat any window change as a break in the series, and annotate it in the change log before it happens.

Conversions dated to the click in one report and the order in another

Many platforms report a conversion on the date of the ad interaction, while the order system and the analytics property record it on the date of purchase. A month in the platform therefore includes orders placed the following month, and a recent period goes on growing as late conversions arrive. The platform versus analytics discrepancies guide covers dating and processing delays in detail.

Fix: compare closed periods only, after each system has finished processing, and note the date basis of each figure. Where a platform offers a conversion-time view, use it for comparisons with the order system.

Verify: totals for a closed period stop changing between readings.

Prevent: set a fixed wait before any period is reported as final, and print the date basis in the report footer.

Duplicate conversion events from double tagging

A purchase tag installed both directly and through a tag manager, a confirmation page that fires again when reloaded, or two integrations sending the same order all multiply conversions without any extra sale. The common analytics mistakes guide covers double installations and repeated purchase events.

Fix: inspect the confirmation page with the tag tools, remove the duplicate, and pass an order ID with every conversion so each platform and the property can apply Deduplication on it.

Verify: for a sample of orders, each order ID appears once in each system.

Prevent: check conversion firing on the confirmation page after every site, theme or plug-in release.

Refunds, cancellations and test orders never removed

The order system counts net sales; most platforms and many analytics setups keep counting the original conversion after the refund, the cancellation or the test order the team placed to check the checkout. Over time the gap becomes a steady overstatement. The e-commerce tracking tutorial includes the refund step most implementations skip.

Fix: send refund events from the back office, keep test orders out of the production property or filter them by a marker, and report net conversions where a platform supports adjustments.

Verify: reported revenue for a closed period matches net revenue in the back office within a small, explained margin.

Prevent: add the refund flow and the test-order rule to the tracking specification, and recheck them whenever the checkout changes.

A prevention routine for cross-channel totals

An inflated total is only a problem when someone treats it as sales, so prevention is mostly about how numbers are reported. Build the monthly channel report around the order system: real orders and revenue at the top, then a single deduplicated Attribution view that splits them across channels, then each platform's own claims in a separate, clearly labelled section. Never place a sum of platform conversions on the page. Keep a settings register listing each platform's attribution model, click and view windows, date basis and conversion count setting, and update it whenever someone changes a setting. Each month, sample a handful of orders and trace them through each system to catch duplicates and missing refunds early. Each quarter, recompute the excess of summed claims over real orders and compare it with previous quarters: a stable excess means the system is behaving as designed; a jump means a setting changed or a tag broke. When budget decisions depend on which channel is really adding sales, use a holdout or other incrementality test rather than any of the attributed figures.

Illustrative example — not a real result

An outdoor clothing shop's monthly report

An outdoor clothing shop closed a month with 2,140 net orders. Its search platform claimed 1,310 conversions, its social platform 1,180, a video platform 460 and the email tool 720, a total of 4,670, more than double real sales. Splitting the social and video claims showed 1,020 of them were view-based. The social platform's click window was longer than the search platform's, and its figures were dated by click. Tracing 50 orders found 6 counted twice in the search platform, because a theme update had added a second purchase tag, and none of the 112 refunded orders had been removed anywhere. The shop removed the duplicate tag, set up refund events and rebuilt the report around back-office orders with a deduplicated analytics view beneath them. The click-only platform sum, read on aligned windows, now sat at about 1.4 times real orders and stayed there month to month.

The comparison of view-through and click-through conversions explains why view credit is usually the first part of the excess to separate out.

Common mistakes

Adding platform conversions to get a total. Computing return on spend per platform and summing it. Comparing a recent week before late conversions arrive. Changing a window without noting it. Treating refunds as someone else's problem.

Frequently asked questions

Which platform's number is the right one?

None of them is wrong and none of them is the business's sales figure. Each platform honestly reports the conversions it can connect to its own ads under its own rules. The business's sales figure comes from the order system, and channel performance is best read against it, using one consistent attribution view for comparisons between channels and incrementality tests where a decision depends on what a channel really caused.

Can we just turn off view-through conversions everywhere to make the numbers add up?

You can exclude them from the reports you compare, and that removes one source of overlap, but the totals will still exceed sales because click-based claims from several platforms overlap too. Changing what a platform optimises toward is a separate decision with consequences for delivery, so separate the reporting view from the bidding setting and change the reporting first.

How big an excess is normal?

There is no standard figure. The excess grows with the number of platforms, the width of their windows and the share of view-based claims, so two businesses with clean tracking can see very different totals. What matters is that you can explain your own excess cause by cause, and that it stays stable from month to month unless you change a setting.

Terms used on this page

Where this page is used

Also related

Sources

  1. About attribution settings — Meta Business Help Centre
  2. About conversion tracking — Google Ads Help

Related