You pulled the March report on April 1 and sent it to your CEO. Two weeks later you pulled the same report for a board deck and March had more conversions in it than the version you already presented.
Nothing broke. This is Google Ads working exactly as designed, and if you do not know the design, it makes you look careless in a meeting where you cannot afford to.
Conversions are filed under the click, not the conversion
Google Ads attributes a conversion to the date of the click that led to it, not the date the conversion happened.
Someone clicks your ad on March 28 and fills in your form on April 9. Google Ads adds that conversion to March 28, twelve days after you already reported on March.
So March is not finished on March 31. It keeps filling in for as long as clicks from March remain eligible to convert.
This is the right design. It is the only way to know what a given day of spend actually bought. It just means your reporting cadence has to account for it.
How long March stays open
Your conversion window. Each conversion action has one, and it sets how long after a click a conversion can still be counted.
- Click-through window: default 30 days, settable from 1 to 90.
- View-through window: for display and video impressions that were never clicked, default 1 day, settable up to 30.
- Engaged-view window: for video views that met an engagement threshold.
With a 30-day click window, a click from March 31 can still produce a conversion that lands in March as late as April 30. Your March number is not final until then.
Choosing the window
The instinct is to set it long so you capture everything. Resist that.
Too short and you systematically undercount a long sales cycle. Smart Bidding never learns about the conversions that fell outside the window, so it optimizes for whoever converts quickly rather than whoever converts profitably.
Too long and your recent performance reads as terrible, because the last few weeks have not had time to fill in. You end up comparing a mature month against an immature one and making decisions on a difference that is purely an artifact of time.
The workable approach: look at how long your leads actually take to convert, using the "Days to conversion" report in Google Ads, then set the window to cover the bulk of them rather than all of them. If 90% of conversions land within 21 days, a 30-day window is fine and a 90-day window mostly adds reporting lag.
Changing the window is not retroactive. It applies going forward, which means a change creates a discontinuity in your trend line. Note the date you changed it somewhere you will find it again in six months.
Attribution models move credit around too
Separate from windows, data-driven attribution keeps redistributing credit across the touchpoints in a conversion path as it learns. A conversion that was credited entirely to one keyword last week may be split across three this week.
Your total will be stable. The distribution across campaigns and keywords will not be. This is why a keyword-level report from three weeks ago rarely reproduces exactly.
A reporting cadence that holds up
Do not report a month within the first week after it ends. You are reporting an incomplete number and you will have to revise it.
Pick a lag and stick to it. Report March on April 15 if your window is 30 days, or April 30 if you want it effectively closed. Consistency matters more than the specific choice, because it makes your months comparable to each other.
Label in-flight periods. When you have to show the current month, say plainly that it is still filling in. One sentence in the deck prevents the entire conversation about why the number moved.
Use the same date range and the same attribution settings every time. Most reporting disputes come down to two people pulling slightly different reports.
Compare like periods. A month-to-date comparison against a full previous month is not a comparison. It is a mistake with a chart around it.
When the change is not the window
Windows explain conversions being added to past dates. They do not explain everything.
If a past number goes down, look elsewhere. That points at a conversion action being edited, removed from "Include in conversions," or a reprocessing correction on Google's side. If a past number jumps sharply on a single day with no matching spend, that usually points at a tracking change, not attribution.
Learn the difference between the two and you can tell your CEO which of these you are looking at with confidence, which is most of the job.
Want your settings reviewed?
The Conversion Tracking Audit covers conversion windows, attribution settings, primary and secondary classification, and whether your reporting is reconcilable month to month. You get a written list of what to change and why.
$599, fixed fee, no retainer and no call required first.
