The problem
You make decisions on the numbers I send you. Whether to keep an ad running, who to call back, how much money has come in. If one of those numbers is wrong, you make the call on a wrong number, and you may never find out. Reporting software has a bad habit. It can say "done" or "sent" when nothing useful happened. So I check my own numbers against the place they came from, and when I find a mistake, I fix whatever made it. Here are three of my own mistakes from this summer, what I found, and what I changed.
My own website's lead tracking
Between June 9 and August 2, my website's tracking reported eight leads as delivered to Google Analytics. Analytics had recorded none of them. One setting was written in lowercase where Google expects capitals. Google threw the whole record away and still answered that it had arrived. We sent the same record to Google's testing address, and it named the bad value. We changed it, watched a test lead land in Analytics, and added an automatic check that fails if the old mistake comes back. Then we checked my clients' setups directly. Neither of the two was affected.
A storm-restoration contractor's morning report
For more than three weeks after it started, the 8 a.m. report understated the money the contractor had collected. My sync wasn't refreshing the financial data on a schedule, and the gap came to $338,531. This one reached the client. The low figure sat on the report every morning until I fixed it. We started a daily refresh of the financial data on July 26. Overnight, the collected figure moved from $60,962.55 to $399,493.81, matching what was already in the contractor's CRM.
A weekly call report for a client
On a Monday morning in September, I asked to read the weekly report before it sent itself at 8 a.m. We held it about an hour before the send and checked it line by line against the phone system's records. The numbers matched. Four statements didn't. One "longest wait" counted the conversation after a transfer as waiting, so a call answered in 8 seconds showed a wait of nearly ten minutes. Another line promised a section the report didn't have. We held it, fixed the four statements in the program that writes the report instead of just that week's copy, and sent the corrected report the same day. One fix didn't fully take. A footer line about how caller numbers are stored was still wrong in that corrected report, and the next Monday's read caught it. From that week on, the report got held every Monday until somebody read it against the call records.
The numbers
In each case, once the fix was in, the number you'd act on matched its source. See how the measurement engine works.
What I can't show yet
Checks catch the kinds of mistakes somebody already thought to look for, so I can't prove nothing else is wrong. The weekly reviews that followed kept finding things to fix, and those fixes went into the program itself.
Whoever keeps your numbers, it's fair to ask where a decision from six months ago is written down and what stops an untested change from reaching your reports. A good answer includes the last time that stop caught something. If you want a second look at a report you rely on, I'll read it against its source and tell you what doesn't match. [Talk to me at No Brainer Media](https://nobrainersmma.com/contact/).