This week I found out that the human genetics team changed the names of 27 genes because Microsoft Excel kept turning them into dates.
I am not kidding.
A gene called SEPT1 gets typed into a spreadsheet, and Excel decides it means the first of September.
MARCH1 becomes 1-Mar.
Nobody typed a date. Just dumb machine trying to be helpful.
In 2016 three researchers went through 35,175 spreadsheets from 18 journals.
One in five papers with a gene list in spreadsheets had corrupted data.
Peer reviewed and published.
And they could not switch the autocorrect off, because Redmond geniuses hadn't thought about shipping that feature back in the day.
Solution for this mess was brilliant.
By 2020 the committee that names human genes gave up and renamed the genes.
SEPT1 became SEPTIN1, MARCH1 became MARCHF1.
Twenty seven of them.
Excel: 1 — Humans: 0
Scientists changed naming convention to accommodate a dumbass spreadsheet.
Why is this even possible?
Because Excel was built to save people time.
Selling time means trade-offs. And the trade-off here is ambiguity.
When someone types in SEPT1, it does not stop to ask what kind of data this column holds.
It assumes, based on its own settings, that you meant September 1st.
Most of the time this guess is useful but there are edge cases that corrupt peer-reviewed science.
In advertising & analytics we have our own version of this trade-off.
UTMs.
UTMs are an easy way to track where traffic comes from and credit the channels that drive your business.
Same as Excel. Easy to set up, saves your time but also comes with a huge trade-off.
The trade-off shows up in your data warehouse when you try to stitch everything together.
If your UTM source is typed as facebook, Facebook, fb or Meta, your systems can record four different sources.
They all mean one thing.
Excel turns one label into the wrong meaning. UTM turns one meaning into a dozen labels — better say humans do.
When I started building Bratrax I thought the solution was quite simple. Enforce tracking template.
If our clients were more careful with their UTMs, the mess would disappear.
So we created tracking templates and naming rules.
Pretty much everything the industry has tried for the last thirty years.
It didn't take a lot of time to realize how silly this was.
Someone somewhere will break it. I guarantee it will happen to you too.
Your VA or your media buyer will change UTM_source from Facebook to Facehook.
Then you stitch everything together and for a second you're excited about sales coming from a new channel called Facehook.
A moment later, you realize the whole reporting setup is hopeless.
My mistake when building these templates was not assuming that everyone will either ignore them or use them wrongly.
Teams change, agents change, platforms change, naming conventions drift. Everything that can change will.
So same as the scientists I gave up.
With one difference.
They gave up on the software and changed the names.
I gave up on the names and changed the software.
We stopped asking whether our UTM was correct and started asking what the person behind it meant.
facebook, Facebook, fb and Meta are different labels for one concept.
So we built a semantic layer inside Bratrax that pulls the meaning out of the label.
When the system sees these four, it does not blindly report four sources.
It reads through surrounding signals, works out they point to one underlying source, and stitches them together.
You see one source. Meta.
On most accounts it now resolves more than 90% of ambiguous UTM tags.
Not all of them. Humans are too creative for a complete solution and I will not pretend otherwise.
Attribution should not break because someone capitalized a UTM wrong.
The downstream system should recalibrate itself.
The scientists renamed 27 genes to accommodate the software.
We changed the software to accommodate the humans.
— Brat
P.S. You can take my word for any of this, or you can check it against your own numbers. Connect your store and your ad accounts to Bratrax and you're looking at your own dashboards, with every sale traced back to what actually drove it. Run it beside whatever you use today, and keep whichever one you trust.
P.P.S. Not ready to connect anything yet? Poke around our demo — synthetic data, but modeled on how real stores actually behave. A minute to get in, then click around like you own the place.
