Why Google Ads Conversions Don't Match GA4 (and Which Gaps Are Errors)

For example, imagine a hypothetical weekly report: Google Ads says 48 leads, GA4 shows 31 key events from paid search over the same dates, and the sales team counts 22 enquiries that mention an ad. Someone asks which number is wrong, and the honest answer is usually “partly none of them, partly one of them”. Google expects some disagreement: when comparing Google Ads data with third-party analytics such as Google Analytics or Shopify, its help centre says “discrepancies, often up to 20%, are expected due to different attribution models”.

The useful work is separating the gap that comes from counting rules, which you should document and live with, from the gap that comes from a broken setup, which you should fix. This guide covers both, then gives a table you can use to sort one from the other.

Two systems answering different questions

GA4 measures everything that happens on your website or app, from every channel, and assigns each key event to a traffic source under the property’s attribution settings. Google Ads measures conversions that follow an interaction with its own ads, and reports them against the campaign, ad group and keyword that earned the credit. Even with identical tags, these are different questions: “where did this lead come from?” versus “which of my ads helped produce this lead?”

That is also why the figures you compare must come from the same scope. A GA4 total for the whole site will never match a Google Ads total. Compare GA4 key events filtered to Google Ads traffic, for the same conversion, for the same dates, and only then start explaining the difference.

Eight counting differences Google documents

1. Attribution model and who gets the credit

Google Ads now offers two models, last click and data-driven, and Google says data-driven is the default for most conversion actions. GA4 has its own reporting attribution model (data-driven, paid and organic last click, or Google paid channels last click), set in the property’s attribution settings. Changing that GA4 model applies to historical as well as future data, so a colleague’s settings change can move last month’s numbers. Data-driven attribution shares credit between interactions, which is why a campaign can show a conversion figure with decimals.

A common source of gaps: a prospect clicks an ad on Sunday, returns via an organic search on Wednesday and submits the form. Google Ads can still credit its ad. GA4, depending on its model, may give some or all of that credit to organic search.

2. Date of the click versus date of the conversion

Google Ads’ standard conversion columns are “calculated based on the time of the click, not the time of the conversion”. A click last week that converts this week is reported last week. GA4 reports the key event on the day it happened. For comparisons with other systems, use the Google Ads columns marked “by conv. time”; Google suggests “All conv. (by conv. time)” may give a better comparison. In B2B, where a buyer may click, leave and come back days later with a quotation request, this alone can make a single week look badly out of line while the month is close.

3. Conversion windows and lookback windows

A Google Ads conversion window is the period after an ad interaction in which a conversion can be recorded. It can be set from 1 to 90 days, and click-through conversions default to 30 days. GA4’s key event lookback window defaults to 90 days for most key events (30 or 60 are the alternatives) and 30 days for acquisition events such as first visits. If Ads uses 30 days and GA4 uses 90, a conversion 45 days after the click is credited to the ad in GA4 but not in Google Ads. Changes to either window apply going forward only.

4. Counting one versus every

Each Google Ads conversion action has a count setting. “Every” counts each conversion after an ad interaction and suits purchases; “One” counts at most one per click and suits leads. Website actions default to “Every”, so a lead form left on the default can count the same visitor’s two submissions twice. GA4 has its own counting method: “once per event” (the recommended option for new key events) or “once per session” (the legacy default for key events migrated from Universal Analytics goals). Ads on “One” against GA4 on “once per event” will never agree on a site where people submit more than once.

5. Imported GA4 key events versus Google Ads tag conversions

You can send GA4 key events to Google Ads or measure with the Google Ads conversion tag, and Google says you can use both at the same time. They behave differently:

• Imported key events can take up to 24 hours to appear in Google Ads, and history from before the import is not included.
• Conversion actions created through the Google Analytics interface are set as secondary: secondary actions show in “All conversions” but not in “Conversions” and are not used for bidding until you change the action’s optimisation setting in Google Ads.
• View-through conversions are not available for imported Analytics conversions; they need native Google Ads measurement.
• In GA4 you choose which channels can receive credit for conversions shared with Google Ads: Google paid channels or paid and organic channels; Google gives Google paid channels as the default for Google Ads reports. With paid and organic selected, the conversion window and counting method for those conversions are managed in GA4, not Google Ads (source).

The result: an imported GA4 lead event and a Google Ads tag on the same thank-you page are two measurements of one action, with different rules. If both are primary, Google Ads reports and bids on the lead twice.

6. Consent and modelled conversions

When a visitor declines cookies, consent mode changes how Google tags behave so they do not read or write advertising or analytics cookies. Google Ads then uses conversion modelling to estimate conversions it cannot observe, and those modelled conversions appear in the “Conversions” column alongside observed ones. Google notes that modelled conversions can take up to five days to stabilise. GA4 can also model data from consent-mode pings, as its consent management guide explains, but the two platforms model separately, under their own requirements (Google Ads, for example, needs a minimum volume of ad clicks), so their estimates will differ. If your site added a consent banner recently, read our guide to consent mode on GCC websites before treating a drop as a tracking fault. (Consent rules depend on your market; this is not legal advice.)

7. Processing lag and data thresholds

Google Ads warns that reporting lags can take up to 24–48 hours, and newly linked GA4 properties show Google Ads data within 48 hours. Comparing yesterday’s figures is comparing incomplete data. GA4 also applies data thresholds that withhold rows in some reports, for example when demographic data is included, so a filtered GA4 report can show fewer events than the property actually recorded. A longer date range or a report without demographic dimensions helps.

8. Time zones

GA4 files each event under the calendar day in the property’s reporting time zone, and changing that zone affects only future data, sometimes leaving a flat spot or a spike. Google Ads uses the account’s time zone for its time-segmented reports, and Google says that setting can no longer be changed in standard accounts. A property set to UTC and an account set to Riyadh time disagree on which day every evening conversion belongs to. Daily comparisons suffer most; weekly and monthly totals are barely affected.

Expected difference or setup error: a diagnostic table

Use this table when the gap is larger than you can explain, or when it changed suddenly.

A reconciliation routine for one conversion action

1. Pick one action. For example, a B2B distributor might start with “quotation form submitted”. Reconcile one action at a time, never “all conversions”.
2. Use a closed period at least a week old, so lag and modelling have settled.
3. Align the scope. GA4 filtered to Google Ads traffic and the same event; Google Ads showing that one action in the “by conv. time” column.
4. Write down both platforms’ settings: attribution model, windows, count setting, counting method, channels eligible for credit, time zones.
5. Run a test conversion from a real ad click and confirm it fires once, on submission only, in both systems.
6. Check primary and secondary status so only one measurement of each action feeds the “Conversions” column and bidding.
7. Record the remaining gap and the reason for it. A gap you have explained and documented is a known difference; an unexplained one goes back to step 4.

Which number to use for which decision

The platforms will not match exactly, and no setup makes them. What you can have is a clear rule for each use:

• Bidding: Google Ads “Conversions”, built from one primary action per business outcome, because that is what the bidding system learns from.
• Channel comparison: GA4, because it sees every channel under one set of rules.
• Business results: the CRM or sales log. Qualified leads and won deals are the numbers the board cares about, and neither ad platform sees them unless you send them back.

Where the money goes between channels is a different question from whether the counting is right; our guide on measuring marketing when attribution can’t tell the whole story covers that. If the setup itself needs work, start with tracking forms, calls and WhatsApp clicks as GA4 key events.

RX has implemented analytics and tracking setups for multiple client websites, and the accounts stay in the client’s name. If you would rather have the gap audited than chase it yourself, our analytics and tracking service starts with a tracking audit that separates documented differences from setup errors.