.png)
Your order system is the truth for how many orders happened. Google Ads is the truth for what the ad platform is optimising toward. They are answering different questions and they should never match — the mistake is trying to reconcile them instead of assigning each one a job.
On this page
- Why does Google Ads report more conversions than my store?
- How do I report only real purchases in Google Ads?
- Which number should I use for the P&L, and which for steering campaigns?
- Why is my reported customer acquisition cost lower than reality?
- How close should Google Ads and my store numbers be?
- Should I average my platform numbers or turn off view-through attribution?
Why does Google Ads report more conversions than my store?
On a typical e-commerce account the raw conversions column rolls in page views, add-to-cart, begin-checkout, view-item, search, and local actions like directions and store visits. It is not a purchase count and was never meant to be one.
Measured on a leather-goods store in a single week: the raw column read about 42, the real purchase count was 5, and the store's own web orders were 24. Three numbers, all internally correct, all answering different questions.
How do I report only real purchases in Google Ads?
Report the purchase conversion action only, by conversion time. Not the aggregate column.
Then segment before believing any count, because the all-conversions column sums every action attached to the account:
SELECT segments.conversion_action_name, metrics.all_conversions, metrics.all_conversions_value FROM campaign WHERE campaign.id = <id> AND segments.date BETWEEN …
That query is where you find the duplicate action, the legacy action, and the Test action still firing in production.
Which number should I use for the P&L, and which for steering campaigns?
| Source | Its job |
|---|---|
| Order system | The P&L. How many orders, at what margin. Counted once. |
| Google Ads, purchase action, by conversion time | Steering the account. Which campaign to scale. |
| GA4 | Journey and behaviour. Not a source of truth for revenue. |
| Blended efficiency (total revenue ÷ total ad spend) | Whether the whole operation is working. Budget against this. |
Trying to make them agree is the error. They cannot agree, because they count different things over different windows with different attribution.
Why is my reported customer acquisition cost lower than reality?
Before quoting any blended acquisition cost, strip the harvesting campaigns out.
On a supplements account a bottom-of-funnel Shopping campaign was credited with 8,232 first-time orders on $16,245 of spend — $1.97 a customer — because a click-based attribution model gives last-click credit to buyers who had already decided. Brand search behaved the same way at $1.42.
Blended in, reported non-brand acquisition cost read about $15. Stripped, the real figure was about $30. Twice the number, and the $15 version is not actionable by anybody, because no decision could produce more customers at $1.97.
When a harvester contaminates one ratio in a bucket, it contaminates all of them. Fix every metric in that bucket, not the one you happened to notice.
How close should Google Ads and my store numbers be?
The right discipline is a tolerance band, not a reconciliation. Pick your band, treat anything inside it as noise, and chase anything outside it as a bug.
The way to find a real bug is to tie to source day by day, not at the week or month total. A mismatch that looks systemic at month level almost always collapses to a single day or a single order when you go daily. Working through one store's books that way surfaced returns booked to the wrong date, exchanges treated as refunds, shipping understated and tax not netting returns — every one invisible in the monthly total and obvious at day grain.
Should I average my platform numbers or turn off view-through attribution?
Do not average the platforms' claims and call it truth. Two over-counts averaged is still an over-count.
Do not switch off view-through attribution and declare the discrepancy solved. You have not made the number true, you have made it differently false.
Related questions
Does Google Ads double-count a customer who clicks more than one ad?
It can. The aggregate conversions column sums every conversion action attached to the account, so one person triggering several tracked actions is counted several times. Segment by conversion action name and report the single purchase action to see distinct outcomes.
Which number should go in my P&L?
Your order system. It is the only place a sale is counted once. Platform numbers steer the platform; they do not describe the business.
What is an acceptable difference between Google Ads and Shopify?
Set a tolerance band and stick to it rather than reconciling to zero. Anything inside the band is noise; anything outside it is worth chasing day by day rather than at month level.
What is a harvester campaign?
A campaign that intercepts demand you already created rather than creating new demand. The signature is an implausibly low cost per new customer, an implausibly high return, and a spend line that barely moves whatever you do.
Do I need GA4 at all if I have Google Ads conversion tracking?
Yes, but for a different job. GA4 tells you about journey and on-site behaviour. Google Ads conversion tracking steers bidding. Neither is your revenue record — that is your order system.
Written by Usama Khan, who runs Google Ads for e-commerce brands and home-services companies at AdventurAds. Figures are from accounts under active management and are anonymised. Last reviewed 2026-08-26.
