Analytics says 120 leads, the CRM 87, the ad platform 140: which is right?

No number can be trusted as is: each system counts something else

So 120 leads in analytics usually means 120 triggers, not 120 people.

Each system answers its own question, so their numbers are not supposed to match. The business lead count comes from the reconciled CRM.

In Google Analytics 4 a key event, one marked as important to the business, counts every time it fires by default: five form submissions in a session give five. So 120 leads in analytics usually means 120 triggers, not 120 people.

The CRM, the system of contacts and deals, counts records from the form integration and those typed in by staff, so it depends on the integration, duplicate merging and discipline.

An ad platform applies its own conversion window (the period after a click or view in which an action is credited) and attribution model (the rule for sharing credit between touchpoints), often dating a conversion by the click. It sees only its own visitors but may count actions you do not treat as leads, so it can exceed analytics for the whole site.

Where extra leads come from and where real ones disappear

Different definitions. A lead is a form submission to analytics, any conversion action to the platform, even a chat button tap, and a person record to the CRM. Even inside Meta, Events Manager shows most events received, Ads Manager only those attributed to ads.

Duplicates and tests. A double click, a thank-you page reload or a resubmission give several events from one person. GA4 once per session counting is legacy, and a method change affects only future data. Form tests stay in reports unless marked.

Spam. Analytics counts bot submissions as normal events. Google Ads removes some conversions it finds invalid. The CRM keeps spam until someone marks it.

Cookies and blockers. Basic Google consent mode loads no tags before consent; advanced mode sends cookieless pings that Google uses to model gaps. Meta names ad blockers stopping the pixel as one cause of differences. These people are in the CRM but missing or modelled in analytics.

Calls and messages. Analytics misses calls without call tracking, the platform may count the button tap, and the CRM gets the inquiry only if someone records it. Our inference from how the systems work: with many calls, this is often the largest line.

Even the week starts at a different moment in each system

Time zones. GA4 assigns a visit to a day by the property time zone, and a zone change only affects future data. Google Ads asks those uploading offline conversions to watch for time zone differences. A CRM on London time and analytics on Dubai time cut the week in different places.

Click date. Google Ads dates conversions by impression or click, other tools usually by conversion; by conversion time columns allow comparison. A Sunday click and a Wednesday form land in different weeks.

Windows. GA4 lookback defaults to 90 days for most key events and 30 for first visits. Google Ads click-through defaults to 30 days (1 to 90), view-through to 1 day. Meta credits 1 or 7 days after a link click and 1 day after an impression. Window changes in GA4 and Google Ads apply only going forward.

Models. GA4 models give no credit to direct visits unless the whole path is direct, and a reporting model change rewrites history too. Meta warns that ad sets with different models cannot be compared in one table.

Delayed uploads. When the CRM sends confirmed leads back, past weeks change: Google Ads usually processes an upload in under 12 hours, up to 72, dates it by click and advises against comparing by upload date.

Illustrative example: 120 and 87 both reach 95

Every reconciliation line becomes a task: protect the form from bots, mark tests, fix the integration and set a rule for recording calls.

Invented numbers, not a company result or benchmark. Week: Monday 5 to Sunday 11 October 2026. A lead is a unique person, by phone or email, first in touch that week by form, call or chat, no tests or spam, on the CRM clock. There were 95: 78 forms and 17 calls and messages.

Lines come from matching records, not from working back to the answer; any leftover is an unexplained line.

Analytics overstated by 25, or 26.3% of 95, through repeats, tests and spam, yet missed 8 people and all 17 calls and messages. The CRM understated by 8, or 8.4%: 4 forms lost to an integration failure, 9 inquiries never entered, 5 extra records never removed.

Every reconciliation line becomes a task: protect the form from bots, mark tests, fix the integration and set a rule for recording calls.

Reconcile the ad platform with its own ad leads, not with all leads

The platform only sees its own ads, so compare it with leads whose CRM source is that platform by UTM tags or click identifier: 61 of 78 form leads here. Of its 140 conversions, 50 matched CRM records; 11 more ad leads it put in last week's row or missed. Reconcile ad calls separately by CRM source.

The largest line is a definition: 38 of 140 conversions were button taps. A platform that learns from taps brings people who tap, not people who inquire. View-through is a different crediting rule: Meta says it may credit a purchase to an ad seen but not clicked. Such conversions cannot be matched to CRM records, so they get their own line.

How to reconcile one week: seven steps

For a marketer or analyst with access to all three systems; the output is two bridges and a task list. The order is our proposal, not platform documentation.

1. Take a closed week from two weeks back, aligned to the CRM time zone.
2. Export CRM records with date, source and phone or email, timed form submissions from analytics, platform conversions by conversion time.
3. Mark tests, spam and repeats with one rule everywhere.
4. Match records: a form missing from the CRM means a broken integration; a CRM record without an analytics event means cookies, blockers or manual entry.
5. Check call and chat logs against the CRM.
6. In the platform, separate non-lead actions and conversions with no record; check window, model and counting.
7. Build the bridges with an unexplained remainder, give each line an owner and a date, repeat next week.

Which system to trust for which question

Ask not which number is right but which system answers the question (an editorial rule).

• Leads, sales, customer cost: the reconciled CRM (see the article on CPL, CAC and ROAS).
• Ad spend: the platform, which issues the invoice.
• Site drop-off: analytics, as shares and trends.
• Best campaign within a platform: that platform, with one window, model and primary conversion.
• Platform bringing suitable customers: the CRM source field. Meta itself advises judging results in external analytics tools if you use them.
• Leads becoming deals: the CRM by monthly cohorts, as in the article on cohort conversion.

Write the lead definition once, for everyone

One document for marketing, sales and the agency, with an owner and a version date; mark changes in reports, or comparisons with earlier weeks lose meaning. The repeat period and minimum call length are your choice: platform help pages set no standard.

• Channels: form, a call with a real conversation, a chat message with a question.
• Uniqueness: phone or email; a repeat within the chosen period is not new.
• Exclusions: flagged tests, spam by written criteria, staff and agency.
• Date: CRM create date, CRM time zone.
• Source: UTM tags or a click identifier at first contact.
• Mapping: the analytics event and platform conversion that match a lead; a confirmed lead is primary, taps secondary.

What reconciliation will not tell you

The numbers are invented. Platform help pages set no standard for an acceptable gap, and we quote no industry figures without a method and date: agree what counts as explained and record the remainder.

Modelled, cross-device and view-through conversions are platform estimates that cannot be matched to CRM records. Facts were checked on 31 October 2026. Reconciliation counts leads but does not judge their quality, which is covered separately.

The publisher of this journal is connected to WEBNOSTRA, a company that provides analytics setup services. You can run this reconciliation yourself; the service link is optional.

Key decisions

1. take lead and sales counts from the reconciled CRM, not from analytics or the ad platform
2. reconcile an ad platform's number only against leads from that platform's ads
3. write one lead definition and make a confirmed lead the platform's primary conversion

What to discuss with your team

• Each system has a note of what it counts and its time zone
• Tests, spam and repeats are marked by one rule
• Calls and messages are counted with forms
• The platform number is reconciled with its own ad leads
• The primary conversion is a confirmed lead
• The lead definition is written down with an owner and a version date

Sources

1. Google Analytics: About eventssupport.google.com · checked 31/10/2026
2. Google Analytics: About key eventssupport.google.com · checked 31/10/2026
3. Google Analytics: About key event counting methodsupport.google.com · checked 31/10/2026
4. Google Analytics: Select attribution settingssupport.google.com · checked 31/10/2026
5. Google Analytics: Get started with attributionsupport.google.com · checked 31/10/2026
6. Google: About consent modesupport.google.com · checked 31/10/2026
7. Google Analytics: Set up Analytics for a website, reporting time zonesupport.google.com · checked 31/10/2026
8. Google Ads: About conversion windowssupport.google.com · checked 31/10/2026
9. Google Ads: Data discrepancies, factors and troubleshootingsupport.google.com · checked 31/10/2026
10. Google Ads: Fix discrepancies and errors in offline conversion importssupport.google.com · checked 31/10/2026
11. Yandex Metrica: Goalsyandex.com · checked 31/10/2026
12. Yandex Metrica: Offline conversion data transfer and processingyandex.com · checked 31/10/2026
13. Yandex Direct: Statistics questions and differences with Metricayandex.com · checked 31/10/2026
14. Meta: About attribution models and attribution settingsfacebook.com · checked 31/10/2026
15. Meta: About conversion count differences between Meta Ads Reporting and third-party reporting toolsfacebook.com · checked 31/10/2026
16. Meta: Differences between event counts in Ads Manager, Ads Reporting and Events Managerfacebook.com · checked 31/10/2026