Why do GA4, your store and your ads show different results?

1. Make sure you're comparing the same thing.

Say the owner sees 120 orders in the store, the marketer sees 105 purchases in analytics and the ad platform credits campaigns with 112 conversions. These numbers don't necessarily describe the same set. An order is a sales record in the store. A purchase event is a signal sent to analytics. Crediting a purchase to an ad also depends on the rule you've chosen, that is, the attribution model.

Start with a sentence like: "We want to reconcile the number of orders placed this week, excluding canceled ones." That way you're not trying to settle sales accuracy, customer recognition and ad performance all at once. Write down the report name and the exact metric name from each tool. The word "purchases" without a definition isn't precise enough.

2. Close the period and check the dates.

Compare a completed range, e.g. Monday to Sunday. The current day is hard to judge: orders keep coming in, statuses change, and cost and sales imports run on their own schedule. Google Analytics also processes data at different intervals and may update results later. So write down the time you read both reports.

Check which date the purchase was assigned to. Placing an order, paying for it and clicking an ad are different moments. A Sunday-evening transaction may fall on Monday in another time zone. If the gap appears at a month boundary, look at the last and first hours of the period. Don't shift dates by hand just to make the totals match.

3. Break revenue down into its components.

Agree whether the report shows order value before adjustments, after reported returns or only paid transactions. None of these definitions is interchangeable with the others. In Flowpeek, read the description of the report you're using; an amount after adjustments doesn't by itself confirm that the payment has been settled.

Next, check shipping costs, discounts and tax. The word "net" is sometimes used to describe the value after adjustments, so without a description don't assume it means the amount excluding VAT. Compare currencies separately. If one table contains PLN and EUR, you need an explicit conversion rule, not a simple sum of the values. This often explains a gap bigger than a single missing event.

4. Compare specific order numbers.

Pick a sample that covers different situations: a normal purchase, a cancellation, a partial return, a purchase with a recognized source and a purchase without attribution. Compare the same ID, status, time and value. If the event report doesn't let you trace back to the order, note it as a limitation of the comparison.

In a working spreadsheet, mark four states: present in both systems, only in the first, only in the second, present with a different value. Each state points to a different cause. For an extra event, check whether the purchase could have been sent twice; for a missing record, check the integration and the import scope. Don't correct the total with a ballpark factor, because the problem will come back next month.

• The same order number and currency.
• The same status and the same point at which a return is taken into account.
• Checking for duplicate events instead of treating them as extra orders.

5. Only now compare channel attribution.

A correctly imported order may have no link to an observed visit. In that case the sale exists, but the source stays unattributed. A historical order import doesn't recreate old clicks. Don't turn a missing source into Direct: that would be new information added without evidence.

Attribution credits a share of sales to the sources the customer came from before the purchase. The result depends on the crediting rule, the time window considered before the purchase and the visits that were recorded. So it can change even though the number of orders stays the same. Don't add conversions from separate ad platforms into a single store total: the same purchases may be claimed by more than one system. In Flowpeek, look at the sources and the journey, then write down which rule you use in your recurring report.

Flowpeek gives you your own point of reference. It doesn't favor an ad platform because of its name: it credits shares based on the recorded journey and the selected model. That's why you should compare channels under the same rule and the same time window before purchase. This doesn't mean every channel gets the same result. It does let you discuss budget on a common basis.

In the report shown, Google Ads has 47 conversions under First Touch and 39 under Last Touch. Direct changes from 74 to 83. That doesn't mean the store gained or lost orders after switching models. What changed is the rule for crediting sources. Both tables still show 136 unattributed purchases — picking a different model alone doesn't restore the missing link.

Flowpeek also supports Last Non-Direct, a linear model, a 40/20/40 position-based model and Time Decay with a seven-day half-life. Assisted conversions are a separate breakdown of a source's earlier presence on the journey. They aren't extra orders to be added to store sales.

6. Example: 100 orders turn into three different results.

Say a store has 100 orders of PLN 200 each, five of them canceled. The report before adjustments will show PLN 20,000, and the breakdown excluding cancellations PLN 19,000. On top of that, ten of the remaining orders have no recognized source. A channel report that leaves out the "Unattributed" row will show yet another total.

The fix is to agree on statuses and make unattributed sales visible again. Only the rest of the gap goes to an integration check. The figures above are an arithmetic example, not a Flowpeek customer's results. They show why, before judging data quality, it's worth building a simple bridge between reports: starting point, adjustments, end point.