The ad platform said three people bought the book.
The business had four orders.
That is a small disagreement until you have to decide what to do with next month's money.
I was on a call with a client, moving between the ad account and the funnel. The ad platform had recorded 130 clicks. The funnel showed 62 page views. One screen said the campaign had done one thing. The next screen said it had done another.
The dashboard wanted credit.
Dashboards tend to.
They are built to describe the part of the trip they can see. Facebook can tell you that someone touched an ad. A funnel can tell you that a page loaded. A customer system can tell you that a record appeared. The bank can tell you that money arrived.
None of them can see the whole person.
This creates a strange habit in marketing. We treat the screen with the brightest colors as the authority, then argue about its version of events. The platform reports a conversion, so the ad worked. The platform misses a conversion, so the ad failed. The click count rises, so demand must have risen with it.
On that call, the useful question was smaller: which number could we defend?
The four orders were real. Someone had paid. The contacts in the system were real enough to inspect. The spend was real because it had left the account. The gap between 130 clicks and 62 page views was not a result. It was a warning that two systems defined the trip in different ways.
That distinction changes decisions.
If you accept the platform's story, you optimize the ad. If the first material loss happens after the click, the ad may be the one part doing its job. More clicks would send more people into the same leak.
If you accept the funnel's story without checking the customer record, you may miss orders that entered through another path. A buyer can see an ad, leave, return through an email, and purchase three days later. The ad account and the email platform may both stand up and claim paternity. The buyer does not care. She bought when she was ready.
This is why I start with the denominator now.
Before I ask whether a campaign improved, I ask what population the number describes. Impressions. Clicks. Page views. Contacts. Qualified leads. Calls. Orders. Cash collected. They are not interchangeable stages with nicer labels. Each one answers a different question.
A 20 percent conversion rate can be wonderful or meaningless. Twenty percent of what? Over what window? Who was excluded? What broke between systems? Did the business collect money, or did a pixel fire because somebody reached a thank-you page?
The order matters too.
Start with the business result that can be inspected. Work backward through the systems. Match orders to contacts, contacts to visits, visits to clicks, and clicks to spend. Stop at the first place the counts stop agreeing.
That is often where the work is.
The lesson from that call was not that Facebook lies. A ruler is not lying because it cannot weigh a stone. The trouble starts when we ask a tool to answer a question it was never built to answer.
The dashboard can tell you what it saw.
The business has to decide what happened.
Source note: Anonymized from a December 11, 2023 client call. The transcript records three platform-reported book purchases against four business orders, 130 platform clicks against 62 funnel page views, and a discussion of reconciling platform spend with contacts and sales. No client identity or private destination data is used.
I turn this kind of reading into a working document for one business at a time: who your buyers imitate, what they actually want, and the words they use when nobody is selling to them. The first bounded read is free. Mimetic Intelligence, built by hand with source-linked evidence and stated confidence.