Data Discrepancy

A data discrepancy is a gap between the numbers two tools report for the same thing, such as orders in Shopify and purchases in Meta Ads Manager. Most gaps come from different definitions, dates, time zones and attribution rules, not from broken tracking.

Daniel Busch
Written by Daniel Busch · Chief of Staff · Updated

In short

  • Is: two tools reporting different numbers for the same orders, revenue, spend or leads. Most gaps come from how each tool counts, not from errors
  • Causes: attribution model and window, click date vs. conversion date, time zone, gross vs. net revenue, definitions, tracking loss, double counting and late data
  • Frequency: about half of the questions customers send us are about numbers that don't match another source
  • First rule: compare like with like, orders with orders on the same finished days, time zone and currency, before you suspect the tracking
  • Tracking loss: 30% of internet users run an ad blocker (GWI), and browsers under ITP carry 24% of web traffic

Why two tools never report the same number

A data discrepancy is rarely a bug. Every number answers questions nobody writes down: which events count, on which day and in which time zone, at what value, credited to which channel. Each tool answers them its own way:

ToolCountsCredits and dates
ShopifyOrders; gross sales count before VAT, shipping, discounts and returns, with cancelled orders, without test ordersReturns on the day they’re processed
GA4Key events its tag recordedPaid and organic channels, within 90 days by default (30 for first visits)
Meta Ads ManagerConversions within 7 days of a link click or 1 day of a view or engagement, by defaultWebsite purchases on the day of the purchase
Google AdsPrimary conversion actions, modeled ones includedData-driven among Google’s own ads, 30-day click window by default, on the click date
KlaviyoOrders within 5 days of an email open or click, by default in new accountsLast touch among its messages, on the send date
adtributeYour shop’s orders, each once across all channelsYour attribution model, on conversion date or click date

Sources: Shopify, GA4, Meta, Meta reporting, Google Ads, Klaviyo, adtribute.

About half of the questions customers send us are about numbers that don’t match another source: spend against the ad account, revenue against Shopify, leads against the CRM, Meta or Google conversions against adtribute.

Causes in how tools count

Attribution model and window

Every platform claims the orders its own ads touched within its own window, views and email opens included, so one order can be a conversion in Meta, Google Ads and Klaviyo at once, as in the platform-reported conversions entry’s example. Independent models disagree too, as the Meta Ads ROAS benchmark shows on first and last touch for the D2C brands in our panel.

Click date vs. conversion date

Google Ads’ standard columns book a conversion on the day of the click (Google Ads Help) and your shop books the order on the day it is placed, so the longer customers take to buy, the more orders fall into a different week or month in each tool, as the click date vs. conversion date entry’s example shows.

Time zone and currency

An order placed at 00:30 in Berlin on 1 March is 23:30 on 28 February in UTC, so a tool that counts in UTC puts it in February. Meta’s own troubleshooting asks you to check that both tools use the same time zone (Meta). Currencies do the same: a Swiss franc order converted at its day’s rate differs from one converted at today’s.

Gross vs. net revenue, refunds and cancelled orders

Take an €89.25 order including 19% VAT, refunded the following month. If your pixel sends the order total, Meta keeps €89.25 of conversion value. Shopify shows €75.00 of gross sales in the order month and a €75.00 return in the month the refund is processed, and adtribute’s net revenue loses it on the order’s own day (docs: Why your numbers differ from other tools). Which deductions each revenue figure makes, and on which day, is worked through in gross vs. net revenue.

Definitions: conversion, lead, new customer

The same word counts different things. Shopify calls a customer new by their customer record, GA4 by a browser cookie, adtribute by the person behind the orders (new vs. returning customers). A lead conversion set to Every counts each form submission, one set to Once per each person once.

Causes in what tools receive

Your shop receives every order, tracking tools only the visits they may and can record. Ad blockers, Safari’s ITP and visitors who decline consent take visits out of their data, as the attribution entry’s section on what breaks attribution today explains. Meta names ad blockers as one reason Ads Manager can show fewer conversions than your own data (Meta).

The order still arrives, but without its visit, click ID and source. It lands in direct traffic, and its channel loses the credit.

Deduplication and double counting

Meta merges a browser event and a Conversions API event only when both carry the same event name and event ID and the second arrives within 48 hours of the first (Meta, CAPI entry’s answer on double counting). Two primary Google Ads conversion actions for the same sale count it twice (docs: Google Ads Conversion Upload), and an ad account in two adtribute connections doubles its spend (docs: Why your numbers differ from other tools).

Late data

Recent days are still moving. Ad platforms deliver some spend late, refunds arrive after the order, and in Google Ads past days keep gaining conversions as late buyers convert. In adtribute’s marketing attribution software, today’s data updates every 30 minutes, refunds, cancellations and older spend the next morning (docs: When your data updates), while Meta and Google show live numbers. More in data freshness.

How to reconcile two numbers, step by step

  1. Pick a pair that can match: orders against orders, spend against spend. A platform’s conversions never match your shop’s orders.
  2. Use finished days in one time zone, without today and yesterday.
  3. Date both sides by the conversion: Google Ads’ columns “by conv. time”, Meta by time of conversion, the model’s date on Conversion in adtribute, not Click.
  4. Use one currency, at the same rates.
  5. Deduct the same things: VAT, shipping, discounts, refunds and cancellations, with test orders out on both sides.
  6. Compare the same scope: shops, markets, sales channels, ad accounts and filters.
  7. Align the model last: against an ad platform, put an attribution model like the platform’s next to its click-only numbers.
  8. Then look for losses and duplicates: pages without your tracking code, consent, ad blockers, an event that fires twice, an ad account connected twice.

What size of gap is normal, and when to dig

No percentage tells you a gap is fine. Our view sorts gaps into three kinds:

  • Same records, same number. Orders and revenue against your shop, spend against your ad accounts: after steps 1 to 6 these should match. What remains is a filter, a connection or a bug, worth a message to support.
  • Different methods, different numbers. Platform conversions against attributed orders, GA4 against Shopify: these never match, and the size of the gap depends on views, attribution windows, models, consent and blockers. On its own, it tells you nothing.
  • The change is the signal. Track the ratio between two numbers that differ by design, such as the purchases Meta reports per order your attribution gives Meta, week by week. When a ratio that held for months jumps overnight, your tracking changed: a theme update that dropped the tracking code, a new cookie banner, a redirect that strips parameters, a replaced tracking template.

How adtribute handles data discrepancies

  • Every definition is readable. Admins and Editors see what a metric includes under Definitions › Attributes or Fields, and which orders or leads count under Definitions › Events: Trigger, Filters and Deduplication.
  • One time zone, day rates. Orders and leads count in your organization’s time zone, spend usually in each ad account’s, other currencies at their day’s exchange rate. Admins set both under App Settings.
  • Conversion date or click date is your pick when you read, in boards, analyses and Looker Studio; what each does to past days is in the click date vs. conversion date entry’s section on how adtribute dates conversions.
  • Every night, your whole history is recalculated with your current settings, so a changed definition applies to the past too.

FAQ about Data Discrepancy

Our numbers don’t match. Which one is right?

Usually both, for different questions: Shopify counts the orders you took, an ad platform the ones it credits to its own ads, an attribution model which channel gets each order once. Compare a pair that can match before you trust either.

Can we match Meta’s purchases to our orders one by one?

Not from Ads Manager. It reports how many conversions each campaign, ad set and ad got, partly estimated by modeling, not which orders they were (Meta). Compare totals over the same finished days instead.