TL;DR
Three platforms reporting three purchase counts for the same Shopify day are usually all correct under their own definitions. The order count in your Shopify admin is the ruler; the three platforms are measurements taken with different rules about which day, which basket and which visitors count. Work the causes in the order below, because most reported three-way gaps die on comparison basis before attribution is ever involved. Delivery failure, the only cause that is a bug, is last on purpose.
Key Takeaways
- Anchor every comparison on Shopify's order count for a fixed date range in the store's time zone. Shopify's sales reports include pending, unpaid and canceled orders, so decide which of those you want in the ruler before you compare anything [2].
- Google Ads credits a conversion to the date of the ad click, while GA4 and Shopify credit the date of the purchase. On a store with a multi-day consideration cycle that alone splits the same sale across two calendars.
- Check whether the numbers have finished arriving before you call them different. A same-day three-way comparison is a comparison of three partial numbers.
- Attribution windows, view-through counting and modelled conversions mean each platform is measuring a different basket of orders, not the same basket badly.
- GA4's per-conversion attribution settings, shipped January 16, 2026, exist to reduce GA4-versus-Google-Ads reporting discrepancies, which means two key events in the same property can now disagree with Ads by different amounts [1].
- Admin, POS, draft and subscription-renewal orders are in Shopify's count and were never attributable clicks. Subtract them from the ruler or accept a permanent gap.
- Shopify renamed its returns metrics to "sales reversals" in a changelog dated March 13, 2026 [3], and the deprecated returns fields were removed after May 1, 2026 [2].
Before you diagnose anything, fix the ruler
Most three-way discrepancies are not attribution problems. They are comparison problems, and they die in the first twenty minutes if you fix four things before you look at a single campaign.
The day boundary. Each platform's day starts at a different hour, in a time zone set on the ad account rather than on your store, and the order's own day is a third thing. An order placed at 23:58 store time can legitimately appear on two different calendar dates across two reports. Revenue-level versions of this are worked through in why Shopify Analytics and GA4 revenue never match.
There is a mechanical wrinkle here worth knowing about server-side setups generally: ad-platform conversions are usually stamped with the order's own timestamp, while a GA4 purchase is stamped when the event is delivered. Delivery takes seconds, so this is immaterial except across a day boundary, where it is exactly the difference between one report and another.
Conversion time versus click time. Google Ads reports a conversion on the date of the click that led to it. GA4 and Shopify report it on the date of the purchase. Nothing is wrong in either. The two calendars simply describe different events, which is the subject of attribution windows and reporting delays on Shopify.
Reporting delay. At least one of your three numbers has not finished arriving. The per-platform ladder of when each number settles is in how long until conversions show up, and the hour-by-hour version for peak days is in Black Friday reporting delays.
Orders no ad platform was ever going to see. Admin orders, POS orders, draft orders converted by staff and subscription renewals are all in the Shopify count and none of them are attributable clicks. See Shopify admin and POS order tracking and tracking subscription orders. Which order states Shopify's own reports include, and what deleting an order does to them, belongs to analytics discrepancies on Shopify: the triage guide.
If your gap survives all four, now the interesting causes begin. If it does not, you have saved yourself a week. For the wider case where the disagreement is between Shopify and one other system rather than among the ad platforms, start from analytics discrepancies on Shopify: the triage guide, which routes by pair rather than by cause.
Is each platform counting a different basket?
Yes, and this is the part most operators expect to be the whole answer. It is one of five groups, and it holds four of the fourteen causes.
Window length. Meta's attribution setting, Google Ads' conversion window and GA4's lookback are different lengths, so each platform's eligible-order basket is a different basket. Sourced window facts and the ROAS consequences are in why ad platforms disagree on ROAS accuracy and GA4 conversion lookback windows.
Click versus view. Meta's setting includes a view-through component that Google's search conversions do not. A platform that counts view-through credits sales nobody clicked for, which is why a platform total can exceed your order count with nobody double-counting a single delivery.
Attribution model, now settable per conversion. On January 16, 2026 Google shipped adjustable attribution settings for individual conversions in GA4, describing the goal as helping customers "eliminate a common cause of conversion reporting discrepancies versus Google Ads" [1]. Read that twice: the vendor names GA4-versus-Ads disagreement as the thing the feature exists to reduce. It also means two key events in one property can now disagree with Ads by different amounts, so "GA4 uses a different model" is no longer a single sentence about the property. Model behaviour itself is covered in GA4 attribution models on Shopify.
Modelled and estimated conversions. Part of Meta's and Google's numbers is not an observed event. It cannot be reconciled to an order, ever, at any level of tracking hygiene. The Meta-specific version, with its own fix list, is Meta Ads Manager shows fewer purchases than Shopify, and the 2026 counting change is how Meta changed conversion counting in 2026.
Which sales are invisible to the browser?
Two causes, and they hit the three platforms unevenly, which is what makes them look like platform bugs.
Consent gaps. Denied visitors are absent from the browser layer everywhere, but not by the same amount, because each platform's browser and server paths carry different shares of your volume. Start with is your consent banner silently erasing your analytics for the stable, store-wide version, the June 2026 GA4 consent change for the signal semantics, and why your tags fire before Shopify answers the consent question for the intermittent, per-page-view version that makes consented-session counts jitter.
Identity and match rate. An event that arrives but cannot be matched to a click or a person is not counted as a conversion by that platform. On Meta that is event match quality, covered in Meta event match quality on Shopify. On Google Ads it is the missing click identifier, covered in Google Ads conversions without a gclid. GA4 has no equivalent gate, so an event that reaches GA4 is not filtered out for want of a matched click or person.
Counted twice now, un-counted later
Three causes that move a number after you have already read it.
Deduplication. Two senders, one sale, and the platform's dedup rule decides whether that is one conversion or two. The rules differ per platform, and the mechanics of the shared event identifier are in event deduplication on Shopify. The Klaviyo flavour of the same problem, where a second app sends the same order, is in Klaviyo shows double conversions on Shopify.
Reversals. A refund lands on the day of the adjustment, not the day of the sale, and each platform handles the reversal differently. GA4 can subtract with a refund event, which carries the refund's own date; ad platforms largely cannot. See Shopify server-side GA4 refund tracking. Vocabulary note: Shopify's changelog of March 13, 2026 renamed returns metrics to sales reversals, and the deprecated returns fields displayed alongside them only until May 1, 2026 [2][3].
Currency. A number in the wrong denomination is a units problem, and it changes value without changing count. Diagnosis in multi-currency Shopify and wrong conversion values. If your counts agree and only your values disagree, stop here and read that one.
When is it actually broken?
Last, because it is the rarest of the fourteen and the most expensive to chase first. Delivery failure means events that never arrived: a theme change that removed a script, a checkout change, a domain move, a credential that expired quietly.
Run the pixel audit and, if the gap started on a specific date, what to check after a theme or checkout change. A gap that appeared overnight is a delivery problem. A gap that has always been there, at roughly the same size, is one of the thirteen above.
| Cause | Cheapest check |
|---|---|
| Day boundary | Compare one fixed 24h window in the store time zone across all three |
| Click date vs order date | Pull Google Ads by conversion date, then by click date, and compare the two |
| Reporting not final | Re-pull yesterday's number tomorrow and note the delta |
| Non-ad orders | Filter Shopify orders by channel, then subtract admin, POS and renewals |
| Window length | Read each account's attribution setting and write the three lengths down |
| Click vs view | Split Meta's number into click and view components |
| Attribution model | Check the per-conversion attribution setting on the GA4 key event [1] |
| Modelled conversions | Accept the floor, then stop reconciling that portion |
| Consent | Count consented sessions cold versus warm, and check whether the gap jitters |
| Match rate | Read Meta's event match quality, and the gclid presence rate on Google Ads |
| Dedup | Confirm one purchase source per destination, keyed on the order id |
| Reversals | Re-pull the month at day 15 and compare it to the day-1 pull |
| Currency | Compare one order's value in the platform against the Shopify order total |
| Delivery failure | Place a test order and watch it arrive on each destination |
FAQ
Which of the three numbers should I report to my team?
Report Shopify's order count and revenue as the business number, and treat the platform numbers as inputs to bidding decisions rather than as accounting. Platform totals include modelled conversions and view-through credit that no order-level record can confirm.
Meta is higher than Shopify and Google Ads is lower. Can both be right?
Yes. Meta can exceed the order count because it credits view-through conversions and models part of its total, and Google Ads can sit below it because it reports on the click date and only counts conversions it can tie to a click. Neither is a delivery problem by itself.
How close is close enough?
Pick a tolerance per platform, write it down, and stop investigating inside it. A useful discipline is to compare the same fixed window weekly and act on changes in the gap rather than on the gap itself, because the gap's size is mostly definitional and its movement is mostly informative.
Does server-side tracking make the three agree?
No, and expecting it to is the most common disappointment in this area. Server-side delivery removes the delivery-failure and browser-loss causes. Windows, view-through, modelling and click-date reporting are unaffected because they are counting rules, not collection problems.
Where do I start if I only have an hour?
The ruler. One fixed window, one time zone, non-ad orders subtracted, and yesterday's numbers re-pulled today. That hour resolves more three-way gaps than any campaign-level investigation.
When the causes above are ruled out and you want the delivery side to stop being a variable, WeltPixel Conversion Tracking [4] sends the purchase from the Shopify order record to each destination once, matched to the order, across nine integrations, so the number you are reconciling is the same order everywhere.
One habit worth building before you close this tab: write the three numbers and the ruler into the same sheet every Monday, with the window and the time zone in the column headers. A lot of what gets escalated as a tracking bug is two people comparing two different weeks.
Sources
- Google, "What's new in Analytics" (entry dated January 16, 2026: "Improved web conversion management and reporting capabilities for Google Ads customers (Beta)", conversion attribution settings adjustable independently for every conversion, intended to "eliminate a common cause of conversion reporting discrepancies versus Google Ads"), support.google.com/analytics/answer/9164320, accessed September 10, 2026
- Shopify Help Center, "Sales reports" (pending, unpaid and canceled orders are included; test orders are not; deleted orders are removed from reports; sales reversals definition; deprecated return metrics shown alongside the new ones until May 1, 2026), help.shopify.com/en/manual/reports-and-analytics/shopify-reports/report-types/default-reports/sales-report, accessed September 10, 2026
- Shopify changelog, "Returns metrics renamed to reversals", published March 13, 2026, changelog.shopify.com/posts/returns-metrics-renamed-to-reversals, accessed September 10, 2026
- WeltPixel Conversion Tracking, Shopify App Store listing, apps.shopify.com/weltpixel-conversion-tracking, accessed September 10, 2026