A visitor finishes our Hexagram Diagnostic, the quiz that scores a business across six pillars, and the confirmation screen shows a score. Their emailed Gap Map shows a score too. When we put the redesigned site live at the end of September, those two numbers could disagree, because they were never calculated the same way.
We found it and fixed it the same day. The lesson is bigger than the bug.
Two scores, two formulas
The redesigned quiz page added up the raw answers in the browser and showed the sum out of 60, labelled "provisional".
The real scoring lives on the server, and has done since well before the redesign. The server averages each pillar's two answers into a score from 1 to 5, then totals the six pillars out of 30. That is the number the business owner's email uses, and the number in the copy we receive.
So the page showed arithmetic written for the page. Everything after it used the actual formula. The two were never going to land on the same number, because they were not measuring the same thing.
Nobody had to make a mistake for this to happen. The new page needed something to display the moment someone pressed submit, and a quick sum filled the gap. It looked like it worked: it was a number, it sat in the right place, and it went up when the answers went up. That is exactly why it survived the redesign.
Why "provisional" is the tell
A number labelled "provisional" is usually a number nobody has got round to replacing. The label is a quiet admission, written into the copy, that the figure on screen is not the real one.
The tempting fix was to tweak the page's sum until it matched the server's. We did not do that. Two formulas that agree today will drift the first time someone changes one of them and forgets the other.
The confirmation screen now waits for the server's response, the same one that generates the email, and shows exactly that: the total out of 30, the health status, every pillar's score and label, and the three pillars to fix first. We rescaled the example Gap Map graphic and the copy around it too, so a visitor reading the page before they start sees the same 1 to 5 scale and 30-point total they will get at the end.
There is now one function in our codebase that decides a business's score, and every surface that shows a score reads from it.
The same failure is probably in your funnel
Any business that shows a customer the same figure twice has this failure available to it. A quote on screen and a quote in the confirmation email. A price on the product page and a total at checkout. A dashboard number and the monthly report.
It rarely looks like a bug. Each number is plausible on its own. The gap only appears when someone compares them side by side, and that is the worst possible moment: the customer is deciding whether to trust what you just told them.
This is a conversion problem before it is an engineering one. A visitor who sees two different scores does not conclude that you have a rounding issue. They conclude that one of your numbers is wrong, and they cannot tell which.
The check is simple. Find every number your customer sees in more than one place, and trace each copy back to where it is calculated. If two copies are calculated separately, one of them is a placeholder wearing the other one's clothes, whether or not anyone labelled it "provisional".
This is the second time in a month that this tool has needed a second look; the tool that finds bottlenecks had one of its own covers the first. Moments like this sit under the Conversion pillar, which Charulata owns on our bench, and fixing them across a whole funnel is the work of our performance marketing and CRO service.
Matching the numbers was never the fix. Deleting the second one was.
The Hexagram Diagnostic now gives you the same score on screen and in your inbox; take it to see where your own funnel tells two stories.
