Repeat rate 1.05leave it alone
$40 reported → $42 real
Under five per cent, which is inside the noise of modelled conversions and attribution windows. Splitting your reporting history to correct that is a worse outcome than the error itself.
Repeat rate 1.25worth fixing
$40 reported → $50 real
A quarter of your recorded leads are the same people twice, and your cost per lead reads twenty per cent under. Try order ID deduplication before the dropdown: it fixes the cause and leaves your history intact.
Five of the nine guides show you how to change the counting setting. None of them shows you how to check whether it is costing you anything, and Google publishes the exact multiplier for your own account.
What I checked
I opened the ten pages Google returns on page one for two phrasings on 6 September 2026. WordStream refuses this location outright, so nine were read. Five questions, fixed before I started, answered by reading each page.
Six of the nine go further than the setting and describe order ID deduplication, which is the other way to solve this. So the guidance on fixing it is not bad. The gap is entirely on the other side: nothing tells you what state you are in before you start changing things.
The number that settles it
Google publishes your multiplier. From its help page on counting options, describing the repeat rate column in the Conversion actions table:
The repeat rate column is, in Google's own words, "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."
That is not an estimate. It is the exact ratio between the two settings, for your account, per conversion action, already computed. Go and read it before you touch the dropdown, because it tells you whether there is anything to fix.
What your repeat rate does to your numbers
Say a hundred conversions are recorded in a month against four thousand dollars of spend, so your reported cost per lead is forty dollars.
- At a repeat rate of 1.25, those hundred came from eighty ad clicks. Your real cost per lead is 50 dollars: the count is 25 per cent high and the cost per lead is 20 per cent low.
- At a repeat rate of 1.05, they came from ninety five. Your real cost per lead is 42 dollars, and the error is under five per cent.
The direction is the same in both cases and the size is not. A five per cent understatement sits inside the noise of modelled conversions and attribution windows, and it is not what is making an account look comfortably under target. A twenty five per cent one is.
So the threshold is practical rather than principled: if your repeat rate is close to one, read this article and do nothing. Splitting your reporting history to correct a rounding error is a worse outcome than the error.
If you do have a gap, you have two fixes
Almost every guide reaches for the dropdown. Google's own page links to the other one in its related topics, and six of the nine describe it.
Order ID deduplication attacks the cause. You send an order or transaction ID with the conversion, and Google discards a second one carrying the same ID. That fixes a tag firing twice and a visitor reloading a slow confirmation page, which is where most inflated counts come from. It changes the data at source, so there is no seam in your history and nothing to remember in twelve months.
The Count setting attacks the aggregation, and it is forward only. Google is explicit: "Any changes to conversion count settings will only apply to future conversion reporting." Change it today and yesterday stays wrong for ever, every year on year comparison straddling the change compares two definitions, and nothing in the interface marks the date.
Try the first before the second.
What One actually does, and what it costs
If you do switch, know the mechanism. One conversion counts a maximum of one per ad click, not per person and not per month. Two of the pages that recommend it describe it as one per user, which is not the same thing: somebody who returns a week later on another device is a different click, and this setting does not merge them.
That cuts both ways. A genuine second enquiry from the same click, weeks later and inside a click through window that can run to ninety days, is suppressed. On long consideration work that is a real lead deleted from your record.
Two more consequences nobody writes down. Switching cuts your recorded conversions overnight, so a target cost per action calibrated against the inflated series is now wrong in the other direction and the campaign may re-enter learning: re-base the target the same day. And if your attribution is set to paid and organic channels, the dropdown is not there at all, because Google moves the setting into Analytics.
Google's own advice is not the flat rule the guides give either: "You may also want to count one conversion for some conversion actions and every conversion for others."
Three things to do, in this order
Read the repeat rate on each conversion action
Goals, then Conversions, then Summary, then look at the column. If it is near one, stop here: you do not have this problem and the rest is not worth your reporting history.
If it is meaningfully above one, find out why before choosing a fix
A tag firing twice and a customer genuinely enquiring twice look identical in the total and want different remedies. Order ID handles the first without touching your history.
If you change the Count setting, date it and re-base your targets
The change applies forward only, so the day you make it becomes a seam in your own reporting. Write it down, and move your target cost per action to match the number of conversions you will now be recording.
Where this sits in our own tool
Google Ads Vitals reads the export your account produces and works out cost per lead as spend divided by the conversions in that file. It has no idea how those conversions were counted, because the campaign export does not carry a count setting or a repeat rate. If your account is inflating the count, the dashboard shows the same understated cost per lead in nicer type.
I want to be exact about that, because there is a version of this sentence that sounds like physics and is not. A dashboard reading a conversion actions export could compute the deduplicated figure, because that export carries the repeat rate. Mine asks for campaign rows instead, so it cannot. That is a decision I made about the template, not a law about dashboards, and it is a reasonable thing to want changed.
What the file does is the layer above: which campaigns your cost per lead is concentrated in, each one held against a target. That target lives in the Settings sheet of the template workbook, and if you load a plain CSV there is no Settings sheet, so it falls back to a generic forty dollars that has nothing to do with your business. Change that before you read anything into the colours.
Related reading
- Google Ads reporting: the number is still moving when you send the report
- The search terms report that does not add up
- Breakeven ROAS: the one field no calculator asks for is what a corrected cost per lead should be measured against
- Customer lifetime value: the formula argument is over
Questions people actually ask
Read the repeat rate column on each conversion action, under Goals then Conversions then Summary. Google defines it as the number of conversions you would record on every conversion divided by the number you would record on one, so it is the exact multiplier for your account. Near one means you do not have this problem.
A dropdown on each conversion action with two options. Every conversion counts each completion after an ad click; One conversion counts a maximum of one per click. Note per click, not per person: somebody returning a week later on another device is a different click and this setting does not merge them.
It depends on the conversion source, and the two commonest lead actions differ. Google's documentation says every conversion is the default for website actions and one conversion is the default for calls from ads, so a form and a phone call in the same account arrive counted differently.
Only after reading your repeat rate, and preferably after trying order ID deduplication, which fixes duplicate firing at source without touching your history. Google itself says you may want one conversion for some actions and every conversion for others rather than one rule for the account.
No. Google states that changes apply only to future conversion reporting. The day you switch becomes a seam in your reporting, and because your recorded conversions drop overnight you should re-base any target cost per action the same day or the campaign can re-enter learning.
If your attribution is set to paid and organic channels, Google moves the count setting into Analytics and it is not editable in Google Ads. It is also not editable for app install actions.
WRITTEN BY
Olha · clinic data analyst
I build the reporting our managers open every morning at a multi-branch medical clinic — and package it so other practices don't have to start from scratch.
Published on 7 September 2026. Four limits worth stating outside the body text. The sample is two search phrasings and one day, collected from outside the United States, and one of the ten refuses that location: rerun the method rather than trust the counts. The repeat rates in the worked example are illustrations; the whole argument is that you should read your own rather than borrow mine. Google’s counting page still describes defaults in Universal Analytics vocabulary, so if you create an action in the newer category-first flow, check what Count is pre-set to rather than assuming. And an earlier draft of this piece told you to change the setting without telling you how to check whether you needed to, which is the exact charge it lays against everybody else; the correction is the second source below.