You have tracking installed and verified. Now comes the part that confuses even people with years of experience: the numbers change after you look at them. The week that had 12 conversions yesterday has 15 today. The campaign that «didn't convert» on Friday converted on Tuesday. It is not a bug: they are two mechanisms — attribution and lag — that you need to understand in order to read any report properly.
Which click gets credit for the sale: attribution
Before buying, a person may have clicked on your ads several times (a generic one on Monday, your brand on Thursday). Which one gets credit for the conversion? The conversion action's attribution model decides:
| Model | Splits the conversion... | Status in 2026 |
|---|---|---|
| Data-driven | Between the clicks according to what Google's model estimates each contributed | Default and recommended |
| Last click | All of it to the last click before converting | Available; undervalues whatever opens the path |
With data-driven attribution, a campaign can show 1.4 conversions: they are fractions. And the brand campaign usually «loses» conversions compared with last click (which credited it with all of them) in favour of the generic terms that started the search. That is fairer, not an error.
All of this happens only inside Google Ads: a click on a Facebook ad is not part of the split. That is why the figures in Google Ads and in Analytics (which does look at every channel) never match, and there is no reason why they should.
«Conversions» versus «All conversions»
Two different columns:
- Conversions: only the primary actions (the ones Smart Bidding uses). It is the column that counts.
- All conversions: primary + secondary + conversions from other sources (store visits, non-primary calls...).
If a campaign has 5 conversions and 60 «all conversions», what you are looking at is 5 real contacts and 55 clicks on «show phone number». A report with the wrong column tells a false story.
Lag: conversions arrive late
Google records the conversion on the date of the click, not on the date of the purchase. If someone clicks on the 1st and buys on the 6th, on the evening of the 6th the 1st gains a conversion. Consequences:
- The most recent days always look worse than they will end up being: their conversions have not happened yet. The longer your sales cycle, the more «incomplete» days.
- A report about «yesterday» is never closed. With week-long cycles, don't evaluate the last 7 days; with month-long cycles, wait a month.
- Yesterday's CPA is always higher than the real one. If you use it to touch bids, you will be touching them on false data (one of the ten mistakes in Module 5).
Google offers the «by conversion time» columns (the date the purchase happened) so you can see it from the other side, plus a conversion lag report that tells you how many days it takes for 90% of your conversions to arrive. Look at it once: that is the number of days you should exclude when evaluating.
Reading without fooling yourself: three habits
- Compare closed periods: complete week against complete week, both outside the lag window.
- Use «Conversions» (primary) for performance and «All» only for diagnosis.
- Don't compare with Analytics or with the CRM expecting identical figures: different attribution, different window, different consent, different date. Do compare trends: if Google goes up and the CRM goes down, something is wrong.
When the numbers don't add up
The usual discrepancies and their cause:
| Symptom | Likely cause |
|---|---|
| Google Ads counts more leads than the CRM | Duplicates («every» count), form spam, modelled conversions |
| Google Ads counts fewer than the CRM | Cookie rejection without advanced mode, leads from unmeasured channels, conversions outside the window |
| Conversions drop to zero all at once | Broken tracking (redesign, thank you page deleted, CMP) — not the market |
| Last week's conversions go up on their own | Conversion lag: normal |
💡 Ninja trick: Google's attribution knows which click converted, but it does not know whether the lead was worth anything: a spam form and a €5,000 customer weigh the same. Our Lead Rating closes that gap: it scores each lead by its behaviour on the site and by what happens in the CRM, and sends Google offline conversions with a value — Smart Bidding stops optimising towards «forms» and starts optimising towards «customers».
What you should remember
- Attribution splits every conversion between Google Ads clicks (data-driven by default): fractions and a brand campaign «losing» credit are normal.
- Conversions = primary; All = primary + the rest. Judge with the first one.
- Conversions are recorded on the date of the click and arrive with a lag: the most recent days always look worse. Evaluate closed periods.
- The figures from Google, Analytics and the CRM don't match by design; compare trends, not totals.