Google Ads looks worse after fixing conversion tracking: how to read the change.
- “Google Ads conversions dropped after fixing tracking”
- “Google Ads ROAS dropped after removing duplicate conversions”
- “Google Ads CPA went up after fixing double counting”
- “Google Ads performance worse after conversion tracking fix”
- “Google Ads learning period after changing conversion actions”
- “did fixing conversion tracking hurt my Google Ads”
Baris Asa, tracking consultant, London · Last checked: November 2026
If Google Ads looks worse after you fixed conversion tracking, the likeliest reason is that the old numbers were too good. Remove a purchase that was counted twice and conversions fall, cost per conversion rises and ROAS falls, on the same spend and the same sales. Two other things stack on top: bidding can take a short adjustment after a conversion change, and the newest days of data are still filling in. A real drop in sales shows up in the shop. A tidier count shows up only in Ads.So the first check isn’t in Ads at all. Put the shop’s orders for the same dates beside the Ads figure. The usual cause is a purchase counted twice, on record as F-104 when two Google conversion actions both count it, or F-101 when two senders fire on one thank-you page.
If the Ads figure and the shop’s orders now agree, nothing got worse. Only the number changed. If they don’t, jump to what a real drop looks like.
Why does a fix make Ads look worse?
Ads reports what the tag tells it. If one order was reported twice, the account showed two conversions for one sale. Bidding, reports and anyone reading the dashboard all worked from that figure.
Fix the tag and the same order is reported once. The ads didn’t change. The tag now reports each order once. Three effects follow, and they arrive together, which is why it feels like a collapse.
- The count falls to the real number. Cost per conversion rises and ROAS falls by the same arithmetic. If every sale was counted twice, the reported CPA was half the true one. If a third were doubled, it was about three-quarters of it.
- The history doesn’t change. Google says changes to conversion count settings only apply to future conversion reporting. Last month stays inflated and this month is honest, so the chart in Ads shows a cliff on the day of the fix. It is a change of definition, not a change of results.
- Bidding may adjust. Smart Bidding (Google’s automatic bidding) learned from the inflated figure. It may re-calibrate if Google counts the change as a goal change.
| One week | Before the fix | After the fix |
|---|---|---|
| Ad spend | £5,000 | £5,000 |
| Conversions reported in Ads | 200 | 98 |
| Cost per conversion in Ads | £25 | £51 |
| Orders in the shop that came from Ads clicks | 100 | 100 |
| Real cost per order | £50 | £50 |
In this example, the Ads account looks twice as expensive after the fix, and the business hasn’t lost a sale. The real cost per order was £50 the whole time. The report caught up. The small gap on the right (98 against 100) is the kind of thing conversion delay and consent leave behind.
Does Google Ads go into a learning period after a conversion change?
It can, and Google puts limits on it. Its page on the Learning status lists three reasons: a new or reactivated strategy, a composition change (campaigns, ad groups or keywords added or removed), and a goal change, which it words as the conversion goals or actions you optimise towards having significantly changed. Google says Learning never lasts more than seven days, and that performance can take one to two conversion cycles to stabilise.
Two things Google doesn’t say, so I won’t either. It doesn’t define “significantly changed”. A switched-off duplicate action is plainly a goal change. A correction to a value or a duplicate sender on a single action may or may not register. And it gives no figure for how much performance moves. Read the status label in the bid strategy report, and hover over it to see the reason.
Google’s advice for the period is plain. Use the account as usual, expect key metrics to vary, and when you evaluate, give it two conversion cycles or a two-week window, whichever is longer, allowing for the gap between a click and its conversion. A “conversion cycle” there is the time from click to conversion, so a business with a 14-day buying cycle should wait longer than one with a same-day one.
For primary and secondary changes, Google gives a similar line: one to two conversion cycles to adapt to a significant change in the conversion behaviour Smart Bidding is optimising towards. If your fix was to demote a duplicate action to secondary, see primary and secondary conversion actions for which actions bidding uses.
How do I tell a real drop from a cleaner count?
Four checks, in this order. The first two need no access to the tag; the third does.
- Date the fix. Find the day the tag change was published (the version history in Google Tag Manager, or the change history in Google Ads for action settings). Write it down. Everything before it used the old definition.
- Compare with the shop, not with Ads. For the same dates, count the shop’s orders that came from Google Ads traffic. If GA4 is linked and tagged correctly, list purchases from Google Ads by Transaction ID. After a clean fix, Ads’ purchase count should sit close to those orders and normally not well above them. Treat it as a sanity check, not proof: attribution windows and click date against order date can move it a little. No clean attribution? Use the shop’s total orders for the week: if Ads claims more than the shop took in total, something is still counting twice. For value rather than count, see Google Ads conversion value vs actual revenue.
- Test one order. Place a test order and watch it in Tag Assistant: one Google Ads purchase hit, the right value, the order number sent as
transaction_id. One hit is the fix working. - Look at the newest days last. Recent days can read low because conversions take time to arrive after the click, which is the click-to-conversion gap Google’s learning page describes. Reading last week as if it were final makes any fix look like a drop. GA4 has a similar wait, covered in why GA4 numbers change after a few days.
What a real problem looks like: the shop’s orders from Ads traffic fell too, or Ads now reports far below the orders it should be claiming. That points the other way: the fix removed the wrong sender, a purchase now misses its value (F-302 ), or the tag fires on only some checkouts. Test order first, then the tag.
What do I do about targets and budgets?
This part is a decision, and it belongs to whoever runs the account.
A target CPA or target ROAS was set while conversions were inflated. My reading, not Google’s: the same number now asks more of the account than it used to, so bidding can pull back. The target and the new count need to be read together.
Google’s help on data exclusions is the nearest official guidance. It describes exclusions for conversion data outages such as broken tracking or a misplaced tag, says they cover clicks that could have been attributed to the affected conversions, and says that after applying one you should adjust CPA or ROAS targets to reach the performance you want. Exclusions don’t change reporting, and Google warns against using them often or for long periods. A stretch of double counting isn’t named among its examples, so whether it fits is a judgement. It isn’t a switch I’d flip without reading the page.
Don’t put the duplicate back to restore the old numbers. It restores the inflated figure Smart Bidding was trained on.
How do I know it’s fine?
- One test order gives one Google Ads purchase, with the right value, currency and
transaction_id. - For a closed period after the fix, Ads’ purchase count sits at or below the shop’s orders from Ads traffic, and any gap has a named cause (attribution window, delay, consent).
- The bid strategy shows no Learning status (Google says it never lasts more than seven days). On a manual strategy there is no label to check.
- You read the result after two conversion cycles or two weeks, whichever is longer, against shop orders.
- Nobody is comparing last month’s inflated chart with this month’s honest one as a like-for-like.
I do this check as Google Ads conversion tracking work: the tag, the data layer, the conversion actions and your shop’s orders side by side, so you know whether Ads got worse or only more accurate.
Common questions
Why did Google Ads conversions drop after I fixed my tracking?
Most often because the old figure counted some sales twice. Remove the duplicate and the count falls to the real number, cost per conversion rises and ROAS falls, on the same spend and the same orders. Check it against the shop: if Ads now sits at or below the orders it claimed, the fix worked and the old number was the problem.
Does fixing conversion tracking reset Smart Bidding?
It can put bidding into a short adjustment. Google lists a goal change (the conversion goals or actions you optimise to have significantly changed) as a reason for the Learning status, says Learning never lasts more than seven days, and says performance can take one to two conversion cycles to stabilise. Whether a given tag fix counts as significant isn’t defined, so read the status in the account.
Do old Google Ads conversions change when I fix the tag?
No. Google says changes to conversion count settings only apply to future conversion reporting. Past days keep the figure they had, so a before and after chart in Ads compares two different definitions. Compare against the shop’s orders instead.
How long should I wait before judging Google Ads after a tracking fix?
Google’s advice for evaluating a bid strategy is two conversion cycles or a two-week window, whichever is longer, allowing for the delay between click and conversion. Judge against shop orders, not against the old Ads figure.
Should I lower my target ROAS or raise my target CPA after removing double counting?
Look at it with whoever runs the account. A target set while conversions were inflated may now be tighter than it looks. Google’s own advice after a conversion data exclusion is to adjust CPA or ROAS targets and budgets. I treat it as a decision for the account owner, with the shop figures beside it.
Sources
Read on 10 November 2026:
- Google Ads Help, duration of the learning period for campaigns and what affects it (three reasons for Learning: new strategy, composition change, goal change; never lasts more than seven days; one to two conversion cycles to stabilise; evaluate over two conversion cycles or two weeks, whichever is longer; conversion cycle is the time between click and conversion)
- Google Ads Help, about bid strategy statuses (Learning status; key metrics may vary; continue using the account as usual)
- Google Ads Help, about primary and secondary conversion actions (one to two conversion cycles to adapt to a significant change in the conversion behaviour Smart Bidding optimises towards; Learning for up to seven days after structural changes)
- Google Ads Help, about conversion counting options (changes to conversion count settings only apply to future conversion reporting)
- Google Ads Help, use data exclusions for conversion data outages and about data exclusions (outages such as a broken or misplaced tag; exclusions apply to clicks; adjust CPA or ROAS targets afterwards; excluded conversions still appear in reports; not for frequent or prolonged use)