Every Google Ads report template Google publishes in its own gallery, opened signed out against its sample data. The counts are floors: I read the summary and footnote pages of each report, not every page of every one, so each number is what I found rather than what exists.
That is not a bug in your export, and it is not your reporting tool. It is how Google Ads counts. A conversion is credited to the day of the click that produced it, not the day it happened, so a click from the 3rd that turns into a booking on the 24th lands back on the 3rd, three weeks after you reported the 3rd as finished.
Google documents this in at least six help pages, and it goes further than documenting it: inside Google Ads, hovering over the conversions column shows conversion lag data, and the page annotates itself when the range you picked is affected. Google built a feature to warn you.
Then the warning dies at the export. I opened all 35 Google Ads report templates in Google's own public report gallery, and exactly one of them tells you the numbers arrive late.
What Google actually says
On which day a conversion lands:
Google has a reason for this, and it is a good one. From the same page: "This allows you to accurately measure metrics like cost per conversion or return on ad spend, because ad spend is also calculated based on the time of the click." Put the conversion anywhere else and it no longer lines up with the money that bought it.
On why your Google Ads number rarely matches another tool:
That page exists to help you close the gap rather than to excuse it, and it names the fix: add the "Conversions (by conv. time)" columns, which re-cut the same data by the day the conversion actually happened. Google still tells you to expect a residual difference, in its words "discrepancies, often up to 20%, are expected".
On how long a click stays open for conversions: the default click-through window is 30 days, and "you can set it anywhere from 1 to 30, 60, or 90 days for Search and Display campaigns depending on the conversion source." The view-through default is one day.
And, most usefully, on what that does to your ratios:
Read that last one again, because it is Google naming the direction of the error in your report. Cost is charged when the click happens and settles far faster than conversions do, though not perfectly: Google issues invalid activity credits after the fact, and its own report on them notes that "the original click is attributed to the month in which it occurred, while the corresponding credit is applied to the subsequent month." Conversions, meanwhile, keep attaching to that same click day for weeks. So the ratios with cost on top and conversions on the bottom, cost per lead and cost per acquisition, usually start too high and fall. The ratios the other way up, ROAS and conversion value per cost, usually start too low and rise.
And that quote is not an admission wrung out of Google. It comes from the documentation for a Google Ads feature built for this exact problem: conversion lag reporting shows hover cards and forecasts what your CPA and ROAS will look like once the late conversions land. The warning exists. It just does not survive being exported to a spreadsheet.
Usually, not always, because the column moves in both directions. Google removes conversions it later decides were invalid, and says so:
Either way, dates you already reported as closed are still being edited.
One note for anyone checking these pages against each other: Google uses three names for the same behaviour. One page says time of the click, another says ad impression date, a third says query date, and the API calls it impression-to-conversion. They describe the same thing.
The vocabulary exists because the delay does
If you want a sense of how seriously Google takes this internally, look at its API rather than its marketing. The Google Ads API ships an enumeration called ConversionLagBucket with 19 named buckets:
- Day by day for the first two weeks. LESS_THAN_ONE_DAY, then ONE_TO_TWO_DAYS and so on in single-day steps all the way to THIRTEEN_TO_FOURTEEN_DAYS: fourteen buckets before the second week is out.
- Then the steps widen. FOURTEEN_TO_TWENTY_ONE_DAYS, TWENTY_ONE_TO_THIRTY_DAYS.
- And they keep going. THIRTY_TO_FORTY_FIVE_DAYS, FORTY_FIVE_TO_SIXTY_DAYS, SIXTY_TO_NINETY_DAYS.
(There are two more values, UNSPECIFIED and UNKNOWN, which are plumbing rather than buckets. The enum measures impression to conversion, and its sibling for conversions plus later adjustments runs to 41 buckets.) Nobody builds nineteen buckets for something that does not happen.
You can segment your own account by that field, and you should, because the answer is specific to your business and no benchmark can give it to you. A dentist taking same-day bookings and a B2B firm with a six-week evaluation cycle have nothing useful to tell each other about conversion lag.
A separate problem: part of the column may never have been observed
Modeling does not make the number move. It makes part of it an estimate, which is a different complaint, and worth keeping separate so neither one inflates the other. From Google's page on modeled conversions:
Google lists where it applies: conversions that start on one device and finish on another, traffic where cookies are restricted or short-lived, users who did not consent under consent mode, and app journeys where device IDs are unavailable. In its own words, "if conversions can't be measured because cookies aren't allowed by browsers, or have limited time windows then conversions are modeled based on your websites' traffic."
It does not apply to everyone. Google conditions modeling on having enough data to do it confidently, and says that for accounts without enough weekly conversions, "we don't report any modeled conversions." It also says modeled numbers keep settling on their own schedule: they "can take up to 5 days to fully process and stabilize," and conversion values "are subject to retroactive increases for a period of several days as data modeling is finalized."
This is not a scandal. Modeling is a reasonable answer to a real measurement problem, and the alternative, reporting only what can be directly observed, would understate your campaigns rather than mislead you less. But it means the conversions column may be partly Google's estimate of what your advertising caused, and the revenue column carries that estimate forward. It is a good number to steer by. It is not your bank statement, and if it disagrees with your CRM, neither of them is necessarily lying.
I opened all 35 of Google's public Google Ads templates
Google publishes report templates for Google Ads in its own gallery, the one most people still call Looker Studio. I have taken this walk before, with seven free Power BI templates and with thirty-eight free Google Sheets ones; the pattern of what a free template quietly leaves for you to do is the same here. I opened every one of them: not the listing pages, the reports themselves, rendered.
The method, and its limits, in one paragraph. The gallery is at datastudio.google.com and calls itself the Data Studio Report Gallery, which is current rather than stale: Google renamed Looker Studio back to Data Studio in April 2026, "reintroducing a beloved and familiar name", and lookerstudio.google.com now redirects there. Every template's /reporting/<id>/preview link bounces a signed-out visitor to the product marketing page, but swapping that for /open/<id> renders the report itself, which is how all 35 were read. Signed out and against sample data, though, so: I could not see parameters, input controls or conditional-formatting rules, and for the multi-page reports I opened the summaries, introductions, glossaries and "how to read" pages rather than every page of every report. I also did not scan the gallery listing descriptions. So every count below is a floor, and every zero means "none seen", not "none exists".
Thirteen of these templates will show you cost per conversion against the previous period. The twenty-two that compare you to yourself can tell you the direction you are moving and never whether you have arrived, because none of them arrives with a target in it.
That last point deserves the other side spoken out loud, because a gallery template is a starting point rather than a finished report, and its author genuinely cannot know your target cost per lead. Hardcoding one would be worse than shipping none. The fair version of the criticism is narrow: none of them ships with a target, and none of them prompts you to add one. You can build it yourself after you copy the report.
And then the one that does warn you. Exactly one Google template says its own numbers arrive late, the Local Campaign Store Report, and this is what it says, typos and all:
You can argue Google wrote it there because the lag is conspicuous: these are people physically walking into shops, weeks later. That is exactly the point. Click-date lag on ordinary online conversions is not conspicuous, which is why it needs the note more, not less.
Two smaller observations from the reading. Three templates carry a "Check in regularly to get updated changes" note, which is the closest anything else comes to saying the past will change, and all three say it about shifting search volumes rather than about conversions landing late. And four of the 35 rendered broken charts for me, showing dataset configuration errors or invalid parameters where a table should be: both Auction Insights templates, Geographic Performance, and Display Targeting. I read those signed out against sample data, so I cannot tell you whether they behave differently once connected to a live account. I can tell you what a prospective user sees today.
Nobody writing about it mentions it either
I also captured the first page of Google results for three queries: google ads dashboard, free google ads dashboard template, and google ads reporting.
Twenty-one organic results came back, but a good half of them are not trying to explain anything: a login page, a sign-in help page, a service status page and product pages have no occasion to discuss attribution, and it would be padding to count them against the charge. Ten of the 21 do name the metrics a Google Ads dashboard should show. I read the six most likely to carry a caveat by hand, rather than trusting one automated pass over them.
None of the six says that the conversions column keeps filling, or that part of it is modeled. That includes Google's own marketing article, "How to analyze Google Ads successfully", which mentions cost per conversion, conversion rate and four other conversion columns in a single passing line, recommends return on investment as "the one main metric we recommend all Google Ads users to monitor consistently", and says nothing anywhere about when any of them settles. I searched that page for lag, delay, settle, conversion window, modeled, click date and 13 month. None of them appears.
How to read a report that is still moving
None of this makes Google Ads reporting useless. It makes a few specific habits necessary.
Know your own conversion window before you argue about a number
It is in the conversion action settings, it defaults to 30 days, and it can be set as high as 90. That is the outer edge of how long new conversions can attach to a past day. It is not the outer edge of how long that day can change: invalid conversions are removed for up to 13 months, and invalid activity credits arrive after the invoice.
Never compare a fresh period against a settled one
Last week against the week before is the most common chart in Google Ads reporting and it is biased by construction: the newer bar has had less time to collect its conversions. If you show both, show the reader which one is not finished.
Mark the unsettled window instead of hiding it
Cutting the last seven days off the chart is not honesty, it is a different distortion, and it throws away the only data anyone wants to talk about. Show it dimmed, or labelled, or both.
Measure your own lag once
In Google Ads go to the Campaigns page, select Segment, then Conversions and "Days to conversion" (the API calls the same field conversion_lag_bucket). Take a quarter of data and find what share of your conversions arrive after day one, after day seven, after day thirty. Then you know how long your own reports need to settle, and you can stop arguing about it.
Add the "by conv. time" columns when you need the other view
Google ships conversion columns cut by the date the conversion happened rather than the date of the click. They answer a different question, and they are the right ones for lining Google Ads up against your CRM.
Or run the cheapest version of that test
Export the same date range today. Export it again next week. Put the two conversions columns side by side. Whatever moved is the size of the error in every report you have sent from this account.
Treat the revenue column as direction, not accounting
Say Google Ads reports the campaign returned 4.2 times its spend and your books say 3.1. That is not a broken dashboard. Those are two different definitions, one of which includes modeled conversions and cross-device journeys, and one of which includes only money that arrived.
Lag is one of two reasons a Google Ads report is thinner than it looks. The other is that one report cannot be automated at all: Google removed the auction insights fields from its Data Studio connector in September 2024, and the API keeps the same metrics behind a closed allowlist. I tested both, and wrote down what came back, in the one Google Ads report you still have to open yourself.
What I did not measure, and could not verify
Four limits, stated before the part where I sell you something.
I did not measure how big this effect is. This article is about whether the caveat is disclosed, not about how many percent your numbers move. That number is specific to your account, which is why the section above tells you how to measure it rather than quoting someone else's figure at you.
I read the public gallery, which is not the only surface. Google also serves templates inside the signed-in product, and I could not enumerate that. If a template there carries the warning, I did not see it.
I read signed out, against sample data. Parameters, input controls and conditional formatting are invisible in view mode. A target field could exist in a template and not show itself to me.
I did not open every page of every multi-page report. I opened the pages where a footnote would live. A caveat could be sitting on a page I did not open.
One counting discrepancy I could not resolve. The gallery shows 35 Google Ads cards, and I opened 35 reports, but the catalogue file behind the gallery lists 34 unique report IDs for them. Two cards may point at one report. The two candidates rendered for me with different titles and different metrics, so I have left the count at 35 and am telling you the disagreement exists rather than picking the number I prefer.
What we build, and what it cannot fix
I sell a Google Ads dashboard, so here is the honest version of where it sits in all this.
Google Ads Vitals reads an export, which means it inherits everything above. It cannot make a moving number stand still, and it does not warn you about conversion lag. Nothing that reads an export can: the file does not carry your conversion window, so no tool downstream of it knows when a closed month has finished settling. That part stays with you.
What it does do is refuse to draw an unfinished period as if it were finished. A partial month is trimmed out of the totals rather than plotted as a cliff, an incomplete period is dimmed and labelled instead of quietly being compared at full width against a complete one, and the thresholds that colour anything good or bad come from the target cost per lead and target ROAS you type in yourself, which is the thing none of the 35 public gallery templates arrives with.
If you want the longer argument about what any downloaded dashboard is really asking of you, that is what you actually get when you download a template.
It is a 39 dollar file that runs in your browser. It is not a substitute for measuring your own conversion lag, and it will not stop your numbers from moving. It stops you from reading the newest bar on the chart as if it were done.
Questions people actually ask
Because Google Ads credits a conversion to the day of the click that produced it, not the day the conversion happened. Its documentation says the primary conversion columns are "calculated based on the time of the click, not the time of the conversion", and the default click-through conversion window is 30 days, settable up to 90. So a booking today can land on a click day from weeks ago, after you have already reported that day.
Worse, usually. Google's own conversion lag documentation says late conversions "can sometimes cause CPA to look inflated and ROAS to look deflated", because cost is charged at click time and settles quickly while conversions keep arriving. The word to hold on to is sometimes: Google also removes conversions it later judges invalid, which pushes the numbers the other way.
In Google Ads, open the Campaigns page, select Segment, then Conversions and "Days to conversion". The same field is called conversion_lag_bucket in the API, where it has 19 buckets running from under one day to 60-90 days. Take a quarter of data and read off what share of your conversions arrive after day one, day seven and day thirty. That is how long your own reports need to settle.
Both. Google states that in the Conversions column it "reports both modeled and observed conversions", and that modeled conversions "use data that doesn't identify individual users to estimate conversions that Google is unable to observe directly". Modeling is conditional: Google says that for accounts without enough weekly conversions it does not report modeled conversions at all.
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 18 August 2026 and corrected before publication after four independent fact-checks, three of which found something that mattered. The first draft said Google warns you about conversion lag nowhere. That was wrong: Google Ads shows lag data in hover cards on the conversions column and annotates affected date ranges, and the help page I had quoted as an admission is documentation for that feature. The draft also had the Data Studio rename backwards, calling the gallery's name stale when Google renamed Looker Studio back to Data Studio in April 2026; described the template preview links as bouncing to a sign-in page when they go to a marketing page; said two templates carry the "check in regularly" note when three do; claimed the conversion window was the outer edge of how long the past can change, when removals run to 13 months; and stated that cost is final almost immediately, which Google's own invalid activity credits contradict. One sentence claimed 22 templates show cost per conversion against the previous period; the correct count from my own table is 13. The counts in the table are floors, not censuses, for the reasons given above them.