When advertising platforms and the order book disagree

A monthly trading review often puts two numbers side by side. The commerce platform or ERP shows net sales for the month. The advertising accounts show "conversion value" for the same month, and the sum across Google Ads and Meta is frequently larger than the sales the business booked. The usual reaction is to suspect a broken tag. Before looking for one, check the documented counting rules, which differ from the way a finance team records revenue.

Google Ads reports conversions against the date of the ad interaction, while other reporting tools record them on the date the conversion happened. Google's own troubleshooting guide says conversions can be reported up to 90 days after the click, and it recommends adding the "Conversions (by conv. time)" columns when comparing Google Ads with other systems. A sale made on 3 November from a click on 28 October sits in October in the default Google Ads view and in November in the order book.

The conversion window decides which sales a platform is allowed to claim at all. In Google Ads the default click-through window is 30 days, the default view-through window is 1 day and the default engaged-view window for video is 3 days. A change to the window applies only to conversions recorded from that day forward, so a report that spans a settings change mixes two definitions.

Meta applies a different set of rules. Under its standard attribution model, click-through counts events within 1 or 7 days of a link click, view-through counts events within 1 day of an ad impression, and engage-through counts events within 1 day of a non-link interaction such as a 5-second video play. Meta's help centre warns that results cannot be compared across ad sets that use different attribution models.

Neither rulebook knows about the other. A customer who clicks a Meta ad on Monday, clicks a Google ad on Wednesday and buys on Thursday can satisfy both platforms' rules, and each can record the order in its own conversion value.

The order book records what happened after checkout. Cancellations, refunds, partial returns, fraud reversals and test orders reduce net revenue. A platform only learns about them if someone tells it.

Google Ads accepts that correction through conversion adjustments. An advertiser that passes an order ID with each conversion can later upload a "restate" to change the value or a "retract" to remove the conversion. Google's documentation sets the limits. Adjustments made within 7 days of the conversion are visible to automated bidding, adjustments are accepted for up to 54 days, and a retracted website conversion cannot be added back. Where no adjustment is uploaded, a refunded order keeps its original value in the advertising account indefinitely.

Configuration also inflates totals. Google lists "action optimization settings" as a cause of discrepancies. Where non-purchase actions such as page views or add-to-cart events are set as Primary and carry a value, they enter the main conversion value column and bidding counts them. One mis-set action can add value every day without any visible error.

Some platform figures are estimates. Google Ads models cross-device conversions, and where Consent Mode is in place Google Ads and Google Analytics use cookieless pings from users who declined storage to model conversions. Modelled conversions are a reasonable estimate for steering bids. They cannot be matched to an order ID, so they need to be separated before any order-level reconciliation.

The method below needs a finance analyst, the channel owner and read access to the order book and each advertising account.

Choose a closed period and wait until late conversions have landed. Google Ads can add conversions up to 90 days after the click, and its time-lag reporting shows how long your own customers take to convert, which tells you how long to wait. Pull net revenue from the order book for that period by order date, after cancellations and refunds, with the order ID on every line.

From each advertising platform, pull conversions for the same period by conversion date, with the transaction or order ID where the integration passes one. In Google Ads that means the "by conv. time" columns. Record the attribution model, the window and the counting setting for each conversion action that fed the report, because those settings define what the figure means.

Join the platform rows to the order book on order ID and sort every order into four groups. Some orders are claimed by exactly one platform. Some are claimed by more than one. Some are claimed by a platform but are absent from net sales because they were cancelled, refunded or placed as tests. The rest appear in the order book with no platform claim at all, which covers direct, organic, email and repeat purchase revenue.

For each channel, divide platform-reported value by the net revenue of the orders it claimed, after deduplication and refunds. Keep that ratio as a standing bridge between the advertising dashboard and the management accounts, and recalculate it every month. When the ratio drifts with no change in campaigns, check settings and tagging changes first, starting from the date the drift began.

A reconciled figure tells you how much of the order book each platform has claimed under its rules. It does not tell you how many of those orders the spend caused. A buyer who was already going to purchase may click a branded search ad on the way to the checkout, and the platform records the sale in full.

The platforms acknowledge the difference. Meta now offers an incremental attribution model that optimises delivery towards conversions its models predict were caused by an ad, alongside the standard model that credits conversions by time window. The two produce different counts for the same campaign, which is why Meta warns against comparing them in one table.

For a material budget change, the stronger evidence is a controlled test. Hold spend back in a set of regions or audiences, keep it running in a comparable set, and compare net sales from the order book rather than platform-reported conversions. The test is slower than reading a dashboard. It produces a number that finance can defend to a board, because it is measured in the same ledger as the rest of the P&L.

Finance and marketing can only argue productively about advertising return once they work from the same written definitions. A one-page measurement charter should record which revenue figure is authoritative for budget decisions, the attribution model and windows in force for each platform, and the date of any change to them. It should state whether conversion adjustments for refunds are uploaded, by whom and how often, and which conversion actions are Primary. It should also set the cadence and owner of the monthly reconciliation, and the threshold of drift in the bridge ratio that triggers an investigation.

The charter belongs with the management accounts rather than inside an agency report, and it should be signed off by the CFO and the CMO together. The first budget decision taken against it will show whether the bridge is ready. If the channel owner can explain the gap between platform value and net sales line by line, the reallocation can proceed.