Event Property
Concept
Extra information carried on a recorded action, and the field that decides which ways you can break that action apart later.
Why it exists
Counting worked fine back when a recorded action carried nothing but a name. An event is one line saying who did what and when, and counting lines with the same name told you how many times that action happened on a given day. The wall came right after. When Mia asked which kind of sharing had gone up, there was no way to answer.
The workaround for a while was to chop the name into pieces: instrument document_shared_via_link and document_shared_via_email separately. Two branches was tolerable. Add document type, then add whether the account was free or paying, and the name count grew to twelve — and seeing total shares meant adding twelve numbers by hand. Every extra dimension doubled the number of names, and each new name had to be instrumented, documented, and remembered by everyone reading the numbers.
An event property moves those dimensions out of the name and into fields. The name stays as one document_shared, and sharing method and document type ride along as fields on the line. The total comes from one name and the breakdowns come from the fields. Adding a dimension now adds one field instead of doubling anything.
How it works
A property is a name-and-value pair attached to an event line. On a share event, a field called share_method might hold the value link, and a single line can carry several such pairs. Four things get decided when you define one.
- The value type — text, a number, true or false, or a list of values. What you'll do with the value later is what picks the type. Store document length as text and you can only break down by it; store it as a number and you can average and total it.
- How many distinct values — how many kinds of value can land in that field. Three or four, like sharing method, produces a breakdown table you can read. A field like document title, where nearly every line holds a different value, produces a breakdown that is just a long list of one-row entries and tells you nothing. The rough test is whether the split would still fit on one screen after a month of data.
- Consistent spelling — one meaning gets one spelling. If
linkcomes from the web,LINKfrom the app, andLinkfrom the extension, the screen shows three separate rows and none of them is the whole picture. A naming convention, the rule that standardizes how names are formed, applies to values and not only to field names. - When the value is fixed — a property records a fact about the moment it happened, so it never changes afterwards. A line shared in March under the free plan stays free even after that person upgrades in May. That is where properties part ways with a user property, which describes a person's current state and gets overwritten.
What to attach is decided before the feature is built. A tracking plan — the document that records in advance which actions get instrumented under which names with which properties — lists field names and the values allowed in them. Filling them in later doesn't work. Add a property in May and only the lines recorded after that carry a value; everything through April stays blank. Comparing against the previous quarter means waiting until enough time has passed with the property in place.
Where it shows up
Properties get used three ways: breaking down, filtering, and adding up.
- Breaking down — splitting one event by property value and laying the pieces side by side. Split last week's 1,000 shares by method into 800 copied links and 200 email invites, and which part grew is immediately visible. This is the first tool anyone reaches for when asking why a metric moved.
- Filtering — keeping only the lines that match a condition. Shares from paying teams only, or shares from mobile only. Different people need different scopes from the same event, so a number handed to someone else has to arrive with the filter stated alongside it.
- Adding up — totaling or averaging numeric properties. Put the amount on a payment event and the same lines give you revenue and average order value, not just payment count. Without that one field, the payment count has to be multiplied by an assumed price, which stops being true the moment discounts exist.
These fields also anchor an A/B test, an experiment that shows two versions to randomly split groups and compares the results. Record which version a person saw as a property and the two groups can be compared inside the same event. Skip it and there is no way, once the experiment ends, to work out who saw what.
How it sounds at work
"Can you attach the plan property to this event too?"
A request for the field needed to compare free and paying accounts.
"That property has too many distinct values — breaking down by it is unreadable."
Nearly every line holds a different value, so the split just produces a long list.
"The field name matches but the values are coming in two different spellings."
The same meaning is being sent two ways and has to be merged before reading.
When a number moves at Pagedeck, the first thing Mia does is break that event down by property. If shares dropped 30 percent last week, she splits by method to see whether only copied links fell, then by device to see whether only mobile fell. Narrowing it here makes the next conversation concrete. Without properties to split on, the same 30 percent produces an hour of guessing.
What people get wrong
Attaching every property you can think of feels like the safe move. You might need it later, and adding one more field to a line looks free. But an event with 40 fields is an event where nobody knows which fields are worth trusting. And once names and phone numbers start riding along out of habit, personal data ends up inside the analytics tool, and removing it later is a far bigger job than not sending it.
Treating properties as less important than names is the other common mistake. Plenty of teams settle event names together in a meeting and leave properties to whoever writes the code. Properties are the part that answers questions, though. A name alone tells you how often something happened; which circumstances make it happen more takes a property. Six months of events recorded without them leaves six months of data that can produce a total and nothing else.
Where to go next
- Event tracking
the full process of deciding what gets recorded and how, properties included
- Event
how the line these fields attach to gets made in the first place
Last updated 2026-09-16