Why Shopify, GA4 and Meta All Report Different Revenue

Shopify, GA4, and Meta Ads are never going to report the same revenue number, and no amount of pixel debugging will change that. Nothing is broken. The three systems measure genuinely different things: Shopify counts orders that exist, GA4 counts purchase events it managed to observe in a browser, and Meta counts conversions it believes its ads deserve credit for. Expecting them to match is expecting a cash register, a security camera, and a salesperson's commission claim to agree on the same figure.

This article walks through every source of divergence with concrete numbers, decomposes one hypothetical day that produces €48,200, €43,100, and €19,400, and then makes the case for the only fix that actually holds: stop reconciling in spreadsheets and define revenue once, in one governed layer, on top of one warehouse.

Three systems, three different questions

A revenue mismatch between platforms is a definition conflict. The first step to resolving it is understanding what each system was actually built to answer.

Shopify is the order of record. An order row is written server side the moment checkout completes, whether or not a single line of JavaScript ran in the buyer's browser. Ad blockers and consent banners cannot touch it. Shopify answers the question: how much did we actually sell?

GA4 is client side behavioral analytics. A purchase in GA4 is an event that fired on the order confirmation page (or arrived via the Measurement Protocol) and survived the journey to Google's servers. GA4 answers a different question: which sessions and channels led to the purchases we could observe?

Meta Ads is an attributed conversion counter. Meta starts from the purchase events your pixel or Conversions API sends, keeps only the ones it can match to an ad click or impression inside its attribution window, and books the full order value against its own campaigns. Meta answers a third question: how much revenue can we plausibly claim our ads influenced?

The differences, side by side:

Every row in that table is a reason the numbers diverge. Most stores are hit by all of them at once.

Attribution windows do most of the damage

The single largest gap is between what Meta claims and what anything else shows, and attribution windows are the reason.

Meta's default window is 7-day click and 1-day view. Someone clicks your ad on Tuesday and buys on Sunday: Meta claims the full order. But by Sunday that buyer may have also clicked a Google Shopping ad, opened two emails, and finally arrived through a branded search. GA4's data driven model will spread the credit or hand it to search. Shopify does not care at all. One order, three stories.

The view side is worse. A 1-day view conversion is someone who scrolled past your ad, never clicked, and bought within 24 hours. GA4 structurally cannot see ad impressions on Meta's properties, so this revenue exists in Ads Manager and nowhere else. On accounts with large retargeting audiences, view-through can be a fifth or more of Meta's claimed conversion value.

And because every ad platform runs the same self-serving logic, the claims overlap. Meta, Google, and TikTok can each attribute the same order to themselves, in full, at the same time. Sum the platform dashboards for a month and you routinely get 150 to 250 percent of the orders that actually happened. That arithmetic alone should end the practice of reading platform ROAS as truth, an argument we make at full length in Platform ROAS Is Fiction.

Timing and definitions: when an order counts and what revenue means

Even between Shopify and GA4, two systems that both try to count all orders, the definitions quietly diverge.

When an order counts

Shopify distinguishes order created, order paid, and order fulfilled, and different Shopify reports key on different timestamps. GA4 fires its purchase event when the confirmation page renders, which is usually payment authorization time. For a bank transfer or invoice order, those moments can be days apart. For cash on delivery, "paid" may trail "created" by a week.

Timezones compound this. A store on Europe/Amsterdam with a GA4 property accidentally left on US Pacific time (the default many properties inherit) shifts every order placed before 09:00 CET onto GA4's previous day. Daily comparisons become noise; month boundaries move real revenue between reporting periods. Over a month it roughly nets out. On any given day it does not.

What counts as revenue

• Tax and shipping. Shopify's total sales includes taxes and shipping. Many GA4 implementations send a tax-exclusive value, others send the full order total, and plenty of themes changed behavior across updates without anyone noticing. In the EU, where displayed prices include VAT, this alone can be a 15 to 21 percent definitional gap.
• Refunds and cancellations. Shopify subtracts refunds when they are processed. GA4 only knows about refunds if you send refund events, and almost nobody does. Six months in, GA4's cumulative revenue drifts above Shopify's net figure.
• Test and staff orders. Sitting in Shopify until someone filters them, absent from GA4 unless the tester used the real storefront.
• Gift cards. In Shopify reporting, gift card sales are tracked separately and redemptions are treated as a payment method. A GA4 setup that fires a full-value purchase event both when the card is bought and when it is redeemed can count the same €50 twice, depending on how the theme handles it.

Each of these is a defensible choice. They were just made independently, by systems that never agreed on a contract.

Tracking loss: the revenue GA4 never sees

Everything so far assumed GA4 sees every order and merely labels it differently. It does not. A material slice of purchases never produces a collectible event at all.

• Consent. In the EU, banner rejection rates of 10 to 30 percent of visitors are common, depending on banner design. Google Consent Mode v2 sends cookieless pings and lets GA4 model some of the loss back, but modeled data is an estimate with thresholds, not observation.
• Ad blockers. Estimates vary, but a quarter to a third of desktop users running some form of blocking is a common industry figure. Many blockers stop GA4 collection outright. The revenue impact is smaller than the user share, since ecommerce skews mobile, but it is not zero.
• iOS and Safari. Intelligent Tracking Prevention caps cookie lifetimes and strips link decoration, which fragments returning buyers into new users and breaks attribution joins even when the purchase event itself fires.
• Redirect breakage. Hosted payment pages (iDEAL, bank redirects, some wallets) send the buyer off-site to pay. A buyer who closes the tab after paying, without returning to the confirmation page, is a real Shopify order and a GA4 ghost.

Put together, GA4 typically lands 5 to 15 percent below Shopify on revenue, and the exact figure depends on your market and your device mix. A stable gap is a healthy instrument. A gap that drifts is your earliest warning that tracking broke.

One day, three numbers

The day below is hypothetical but realistic for a store around €15M in annual revenue. On June 2, three tools report:

• Shopify: €48,200 total sales, 812 orders
• GA4: €43,100 purchase revenue
• Meta Ads: €19,400 claimed conversion value

Here is where every euro of the Shopify versus GA4 gap goes:

And Meta's €19,400 decomposes like this:

• €11,200 from 7-day click conversions. Real orders, but GA4's data driven model agrees with Meta on only €7,100 of them; the rest had a later touch that GA4 credits elsewhere.
• €4,900 from 1-day view conversions. Buyers who saw an ad and never clicked. Invisible to GA4 by construction.
• €3,300 modeled conversions, Meta's statistical fill for consent-blocked and iOS-restricted journeys.

Meanwhile Google Ads claimed €14,600 for the same day, and €5,600 of order value sits in both platforms' claims simultaneously. Total platform claims: €34,000 against 812 real orders worth €48,200, of which a large share came from email, organic, and repeat buyers no ad touched. Every number is internally consistent. No two of them answer the same question.

Define revenue once and retire the reconciliation spreadsheet

The monthly ritual of exporting three CSVs and explaining the deltas to your CFO is archaeology, and next month you get to run the same dig again. The durable fix has four moves.

Pick the order of record

For an ecommerce brand this is Shopify, full stop. Board revenue, finance revenue, and marketing revenue all derive from the same order table. GA4 and the ad platforms are instruments pointed at that reality, not competing realities.

Land everything in one warehouse

Raw Shopify orders and refunds, the GA4 BigQuery export, and every ad platform's spend and claimed conversions belong in one place you own, on a schedule, with schema you can query. That landing job is what managed data extraction pipelines exist to do. The GA4 export in particular rewards study: its nested event schema is covered in our guide to GA4 data in BigQuery.

Define revenue once, in a semantic layer

A semantic layer is a governed set of business definitions that every tool and every person queries through, instead of each writing their own SQL. The definition itself is boring, which is the point:

create or replace view semantic.daily_net_revenue as
select
  date(o.paid_at, 'Europe/Amsterdam') as revenue_date,
  count(distinct o.order_id) as orders,
  sum(o.subtotal_price) as merchandise_revenue,
  sum(coalesce(r.refunded_amount, 0)) as refunds,
  sum(o.subtotal_price) - sum(coalesce(r.refunded_amount, 0)) as net_revenue
from shopify.orders o
left join shopify.refunds r using (order_id)
where o.financial_status = 'paid'
  and o.test = false
group by 1

Payment date, store timezone, merchandise only, refunds netted, test orders out. Whether those are the right choices matters less than the fact that they are made once, written down, and versioned.

Let attribution be a lens

Once revenue is settled, attribution stops being a fight about whose number is real and becomes a set of perspectives on the same real number. Run several models over the same touchpoint data and read their disagreements, an approach we detail in Stop Choosing an Attribution Model. Platform claims become one input among several, never the verdict.

Where Verity fits

Verity packages this pattern as one managed platform: pipelines land your marketing and analytics data in your own BigQuery, SQLMesh models with tests and lineage turn raw exports into reliable tables, and the semantic layer with a business glossary makes net revenue a governed definition instead of a tribal agreement. On top of that foundation, attribution runs as a lens over touchpoints you own, and Data Chat answers revenue questions from the same definitions your dashboards use.

Frequently asked questions

Which number should I report to my board?

Net revenue derived from your order of record, defined in your warehouse, with the definition written down: typically merchandise revenue minus refunds, excluding tax and shipping, on payment date. Pair it with MER (total revenue over total marketing spend) for efficiency. Never report a platform-attributed number as revenue. Those numbers exist to argue about credit, and they always flatter their author.

Should GA4 ever match Shopify exactly?

No, and a perfect match would itself be suspicious. A healthy GA4 sits 5 to 15 percent below Shopify, for the reasons above. What you should watch is stability: baseline the gap, chart it weekly, and investigate when it moves more than a couple of points, because that drift usually means a tag or a consent banner changed.

Is Meta overstating my revenue?

Meta is answering the question it was asked: how much order value can be matched to a Meta ad touch within its window. That number legitimately includes orders other channels also touched and orders that might have happened anyway. It becomes overstatement only when someone reads it as incremental revenue. Use it as a relative signal for comparing campaigns inside Meta, and test big claims with holdouts.

Will server-side tracking close the gap?

Partially. The Conversions API and server-side tagging recover much of the loss from ad blockers and browser restrictions, so the tracking loss slice shrinks. But they change nothing about attribution windows, view-through credit, tax and shipping definitions, refunds, or timezones. The gap gets smaller. It does not close.