One missing identifier hid an entire purchase process at Szallas.hu.
You cannot decide about what you do not see in the report.
| Finding an invisible purchase process | Before | After |
|---|---|---|
| Purchases started on the special offers site, in Google Analytics | none were visible | 100% visible |
Nobody notices the missing number
Sooner or later somebody notices a wrong number. A missing number they do not, because there is nothing to notice. The report shows no error, it just stays silent about something.
This number happened to be easy to notice, because the entire revenue was missing.
How big the system is
Szallas.hu is an accommodation booking system, roughly to Central Europe what Booking is to the world. The public figures of Szallas Group, the company behind it:
- It works in 5 countries: Hungary, the Czech Republic, Poland, Romania and Croatia.
- It runs 16 products, among them szallas.hu, hotel.cz, noclegi.pl and travelminit.ro.
- In 2022, 87,000 accommodations offered their rooms on it, and more than one million bookings arrived on its sites in a year.
- In 2024, Hungarian guests booked accommodation in Hungary worth HUF 41.1 billion on Szallas.hu.
- In 2022, the Polish company Wirtualna Polska bought it for EUR 82 million.
In a system this size many websites work together, and the measurement has to be right on every one of them.
The purchase happened, the report did not show it
We put several smaller things right in the measurement of Szallas.hu. The most serious case belonged to the site where they publish their special offers.
Here the purchase started on one website and finished on another. There is nothing special in that, several websites work this way at other companies too.
What made it special was that the purchase was not reported to Google Analytics by the visitor's browser, but by the company's server. (Google Analytics is the program that shows how many people come to the site and what they do there. When the server reports to it, that is called server-side tracking.) For some reason these purchases were not visible at all.
So there was a process nobody knew the performance of.
Feeling our way in the dark
I could only see the Google Tag Manager container. (This is the box that holds a website's tracking codes, and it runs in the visitor's browser.) I could not see into the server side, and that was exactly where the measurement started.
It was like having to find the fault in a car without being allowed to open the bonnet.
What remained was elimination. I went through the possible causes one by one and crossed out what it could not be. One was left at the end: with 99 percent probability, the user identifier was missing from the events.
An event is a signal that something happened, a purchase for example. The identifier tells Google Analytics which visitor the signal belongs to. Without an identifier, it rejected the events.
The developers confirmed it later: this was the fault.
From complete invisibility to 100 percent visibility
This measurement did not concern the main site, only the special offers site. Still, a whole process came to light: what was not visible at all before is now 100 percent visible.
The difference is that from now on, decisions about this process can be made from numbers, not from a feeling.