Running ads
Why does Google Ads show more conversions than my CRM?
The short answer
Both numbers are probably right, and Google Ads usually shows more because it counts different things. It dates a conversion to the day of the ad click while your CRM dates it to the day the inquiry arrived, so the same leads land in different months. Counting settings and conversion windows widen the gap.
This is the argument that ends a lot of agency relationships, and most of the time nobody is lying. The report says 47. The office says they spoke to 31 people. Both are counting honestly, and neither number is the one you actually want.
Before working through the causes, it helps to know that Google publishes a page listing them. Its own troubleshooting article names conversion delay, tag setup, customer conversion time, lookback windows, count settings, action optimization settings, cross-device conversions, view-through conversions, invalid traffic and attribution settings as factors that cause data discrepancies. That is ten mechanisms, and an account can be affected by several at once.
It also helps to know what the word means to a platform. Microsoft defines a conversion as the completion of an action by a customer after viewing your ad, which could be a purchase, a registration, or whatever you consider your goal. A conversion is whatever somebody configured it to be. It is not a synonym for a customer, and it is not a synonym for a sale.
Google dates a conversion to the click, your CRM dates it to the inquiry
This is the largest cause and the one almost nobody explains.
Google states it directly: "Google Ads reports conversions on the ad impression date. Other reporting tools attribute them to the conversion date." Google calls the gap a conversion delay.
Work through what it does to a monthly report. Somebody clicks your ad on 28 March, thinks about it, and calls on 3 April. Your CRM records a lead in April. Google Ads records a conversion in March, because that is when the click happened. Neither is wrong. Pull a March report from each system and they disagree, and pull April and they disagree in the opposite direction.
For a business with a long decision cycle this is not a rounding error. A trade where people compare three quotes over a fortnight will have a meaningful share of every month's conversions back-dated into the previous month. If you have ever noticed that last month's numbers quietly went up after you looked at them, this is usually why.
Your account may be counting every conversion rather than every customer
Google lets you choose, for each conversion action, whether to count every conversion after an interaction or only one.
Its own example is unusually clear. Somebody traveling to Chicago, Denver and New York books a hotel in each city and rental cars in two of them. Under the every setting, that single person produces 5 conversions. Under the one setting, Google Ads counts only one conversion per ad click.
Google describes the one setting as the right choice when you care whether a lead was generated rather than how many sales happened, because usually only one unique lead per ad click adds value. That is exactly the situation most service businesses are in, and it is frequently set the other way by default or by whoever built the account.
The practical version: if somebody fills in your contact form, then calls, then requests a quote, an account set to every can record three conversions for one person your office knows as a single name. Your CRM has one record. Google has three. Both are accurate descriptions of different things.
There is a number in the account that tells you exactly how much this is inflating your total, and almost nobody looks at it. Google reports a repeat rate in the Conversion actions table, which it describes as the average number of conversions you receive based on interactions that lead to at least one conversion. In Google's words it is "effectively the number of conversions you would have recorded using the 'every conversion' setting divided by the number you would record using the 'one conversion' setting". Google's own example: if the repeat rate is 1.5 and you recorded 10 conversions on the one setting, switching to every would give 1.5 times as many. So a repeat rate of 1.5 means roughly a third of the conversions on the every setting are repeats of somebody already counted.
The conversion window decides how long a click keeps earning credit
A conversion window is the period after an ad interaction during which a conversion is recorded in Google Ads.
Google says you can set it anywhere from 1 to 90 days depending on the source, and that for Search and Display campaigns, if you do not customize it, a 30-day default window applies. So by default a click from 29 days ago can still produce a conversion in this month's report.
A second setting gets confused with it constantly. Google notes that the lookback window in attribution reports is not the same as the conversion window. The conversion window says how long after a click a conversion still counts. The lookback window says how far back from a conversion a click can still earn credit. It can be set to 30, 60 or 90 days. Two windows, two jobs, similar names. If somebody explaining your numbers uses them interchangeably, ask which one they mean.
Some of those conversions were estimated rather than observed
This one surprises people, and Google is open about it.
Google describes modeled conversions as using data that does not identify individual users to estimate conversions it is unable to observe directly, in order to offer a more complete report. It models to recover slices of data where attribution cannot be observed, because of privacy protections or technical limits.
Two consequences matter to anybody reconciling a report. Google states that modeled conversions can take up to 5 days to fully process and stabilize, and that conversion values are subject to retroactive increases for a period of several days while modeling is finalized. So a number you screenshot on Monday can legitimately be higher on Friday, for the same period.
This is not a trick, and the alternative is worse, since without modeling the report would only show the observable portion. But it does mean a modeled conversion has no matching row in your CRM, by definition. Nobody filled anything in. It is an estimate of somebody the system could not see.
How to reconcile the two numbers
You will not get them to match, and matching is the wrong goal. Getting them to agree on direction is achievable, and that is what you need.
It helps to see the gap taken apart. Say the report shows 47 and the office counted 31. Nine of the 47 came from clicks in the previous month, back-dated. Six were second and third actions by people who had already been counted, because the setting was every rather than one. Four were modeled rather than observed, so no form was ever filled in. That leaves 28 real, newly arrived inquiries against the 31 the office logged, and the remaining three called without ever clicking an ad. Nothing was wrong. Two honest systems described one month in two vocabularies. Treat the figures as an illustration rather than a benchmark, but the decomposition is the exercise worth doing on your own numbers.
- Compare by click date on both sides, or accept that month boundaries will always disagree. Most CRMs can report by lead source date if somebody asks them to.
- Check the counting setting on every conversion action. If you sell services and it is set to every, that alone can explain a large gap.
- Ask what the conversion window is set to. A 90-day window on a business with a two-day decision cycle inflates everything.
- Ask which conversion actions are counted in the headline Conversions column. A newsletter signup and a booked job should not sit in the same total.
- Wait a week before reconciling a period, because modeled figures are still settling for several days.
- Reconcile one quarter, not one month. The back-dating that scrambles a month largely washes out over ninety days.
The number that actually answers your question is not on either report. It is how many jobs you won and what they were worth, set against what you spent. A platform counting 47 conversions and an office counting 31 conversations can both be true while the only figure that matters, the eleven jobs you invoiced, sits in a third system nobody has connected.
Setting up tracking so those three systems tell one story is one of the things we do, so if the reports have been arguing for a while, that is work we take on.
Terms used on this page
Sources
- Data discrepancies: Factors and troubleshooting
- About conversion counting options
- About conversion windows
- About modeled online conversions
- Report Attributes and Performance Statistics (Microsoft Advertising)
Last reviewed 2026-09-11.