Here is a thing that still annoys me after twenty years. A business will argue for an hour about whether the campaign worked, and nobody in the room will ask whether the numbers they are arguing over are right.
They usually aren’t. Not wildly, and not on purpose. But when I open an analytics account for the first time, I expect to find at least one fault that changes a number someone reports upward. I am rarely disappointed.
Analytics doesn’t break. It drifts.
A broken setup would be easier. Someone would notice a flat line and fix it. What happens instead is slower. The site gets a new booking widget and the tag never follows it. A developer tidies up a template and a tag starts firing twice. A thank-you page gets a new address and the conversion quietly stops counting, or starts counting on the wrong page.
Every one of those leaves the reports looking normal. There is still a line, and it still goes up and down. It is just no longer measuring what you think it is.
The usual suspects
These are the faults I find most often:
- Conversions that aren’t conversions. A page view or a button click is being counted as a lead, so the lead number is two or three times what the sales team actually received.
- Double counting. The same purchase or enquiry recorded twice, once by the website and once by a plugin somebody installed years ago.
- Your own people in the data. Staff, the agency and the web developer all sitting in the traffic numbers, sometimes as your most engaged visitors.
- Consent gaps nobody has sized. A banner went in, some tracking stopped for people who decline, and no one has worked out how much of the picture is now missing.
- Payment and booking tools taking the credit. The sale is attributed to the payment page it passed through, so the channel that earned it gets nothing.
None of these are exotic. All of them change decisions.
Why it stays wrong
Because nobody owns it. The agency assumes the web developer set it up properly. The developer assumes the agency checks it. The marketing manager inherited it and has a campaign to get out. Each of them is being reasonable, and the result is a setup that hasn’t been looked at properly since the day it was installed.
The other reason is less comfortable. Inflated numbers are pleasant to report. A fix that halves the lead count overnight is a hard conversation, even when the real number was always the lower one. So the incentive to look is weak.
What I’d do
- Compare analytics with something real. Take last month’s leads or orders from the CRM, the inbox or the till, and put them beside what analytics says. If they are far apart, you have your answer about whether to trust it.
- Test the conversions by hand. Submit the form. Make the booking. Watch what gets recorded, and how many times.
- Write down what each conversion means. One sentence each. If nobody can say what “generate_lead” counts, it shouldn’t be in a report.
- Decide who owns it. One named person who checks it after every site change.
- Then fix it once, properly, and keep a short record of what was changed and why.
This is a day or two of work for most businesses, and it is the cheapest marketing improvement there is. Every decision after it gets better, because it is finally based on something true.