Meta's numbers and my CRM's numbers cover different time periods. How do I compare them fairly?
Meta reports conversions back-dated to the ad impression or click, and its default window credits clicks within seven days and views within one. Your CRM records the booking on the day it happened. So the same job sits in two different weeks. To compare fairly, pull Meta on a date-of-conversion basis, or align both to the lead creation date and accept that the ad platform will keep revising recent days for about a week.
Two clocks, two events
Meta's reporting is anchored to the ad interaction. When someone sees an ad Monday and submits a form Thursday, Meta files that conversion under Monday, because Monday is when the money was spent. Your CRM files it under Thursday, because Thursday is when the record was created. Neither is wrong. They are answering different questions: what did this spend produce, versus what happened this week.
The practical consequence is that any week-over-week comparison built by putting the two exports next to each other is off by the average lag between impression and action, and that lag is not constant across campaigns.
Recent days are not final
Because conversions get back-dated into the window, yesterday's numbers keep growing for several days. Pull a report on Tuesday and the previous Friday will look worse than it will look the following Tuesday. Operators who check daily and react to it are reacting to an artifact.
The rule of thumb: treat the most recent stretch of days covering your attribution window as provisional. Judge performance on periods that have fully settled, and if you must look daily, look at leading indicators like spend, reach and lead volume rather than at ratios that will change under you.
Pick one spine and map everything to it
- Spend belongs to Meta. No other system knows it. Pull it by date and never restate it.
- Lead creation belongs to your CRM. Timestamped, unambiguous, and the same definition every week.
- Revenue belongs to the field service system. Invoiced amounts, not estimates, and dated when the job completed.
- The join is the lead, not the day. Match individual leads across systems, then aggregate by whichever date you chose. Do not aggregate first and subtract.
What this buys you
Once individual Meta leads are matched to individual jobs, the time-window problem stops mattering, because you can report the same lead under any date convention you like. That is the whole argument for doing the join at the record level in a marketing intelligence layer instead of comparing platform totals.
It also lets you honestly report a lag distribution: how long Meta leads typically take to book compared with search leads. That distribution is genuinely useful for planning, and it only exists if you kept the individual records. See how the data plumbing works for the mechanics.
Topics: attribution · reporting · time windows · reconciliation
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.