▤The Test Group experiment running Open the partner account
paid
Affiliate disclosure. The partner link in the masthead and in the band beside the copy on this page is a sponsored link to a partner operator, and this site may be paid if you open an account through it, at no extra cost to you. It carries rel="sponsored noopener" and opens in a new tab. A desk about how a screen is measured and tested should not leave its own funding unsaid: one link funds the site, no operator and no product is named, rated or recommended anywhere on it, and this site runs no analytics of its own on its readers.
The Test Group / The events
One interaction, written down, and what it costs to keep

What one interaction writes down

Before a rate can be computed, something has to be written down. An event is one interaction plus the context around it: what was clicked, when, by which session, on which account, and where the reader was in the flow. On the samples one 14-minute session produces 1,240 of them, which is the point at which a description of a screen becomes a description of a person.

Desk spec
events per session
1,240
session length
14 minutes
bytes per event
240
a month of sessions
535.7 GB
the eventOne interaction, written down. A click carries a name, a time, an account, a session and what was on the screen, and it is kept whether or not the reader chose to be measured.
the funnelThe order the steps happen in. A landing visit becomes an account, an account becomes a deposit page, and 100,000 visits end in 2,074 first bets - a 2.1% path the whole loop is aimed at.
the armsTwo versions of one screen shown at the same time, and a rate for each. 4.30% against 5.16% is a 0.86-point lift, and the interval decides whether it is a result or a coincidence.
Direct answer

An interaction event is one action - a click, a scroll position, a field completed - plus the context needed to place it: a name, a timestamp, an account, a session and the state of the screen. On the samples a single 14-minute session produces 1,240 events of 240 bytes each, and a site with 1,800,000 sessions a month stores 535.7 GB of them.

The fields, and why the context is the invasive part

The event's name is the least interesting field in it. A click on a button is not sensitive; the fact that it was the second visit from one account, from a device seen on four other accounts, after eleven minutes on the deposit page, is. Context is what turns a log into a profile, and the samples record it 1,240 times in a quarter of an hour.

Sample B - the fields of one event, and what each one is for
FieldExample on the samplesWhat it is for
the namedeposit_page_viewcounting the step in the funnel
the timeto the millisecondordering events and measuring hesitation
the sessionone identifier per visitrejoining the events into one journey
the accountabsent before sign-in, present afterattaching the journey to a person
the devicetype, screen, buildsplitting a rate by the screen it happened on
the screen statewhich step, which versionattributing the event to one arm of the test
the payloadthe value typed, masked or notfinding where a form is abandoned
seven fields240 bytes in totalone action, seven answers to "who and when"
sample B - what one session costs to keep events in one 14-minute session = 1,240 bytes per event = 240 one session = 1,240 x 240 = 297,600 bytes = 290.6 KB events per minute = 1,240 / 14 = 88.6 sessions in a month = 1,800,000 a month of events = 297,600 x 1,800,000 = 535,680,000,000 bytes = 535.7 GB a year at the same rate = 535.7 x 12 = 6,428.4 GB the events are not the expensive part; the profile they assemble is.

What the mix of events looks like

Most of what is recorded is not deliberate: a scroll position and a pointer position are recorded as they happen, whether or not the reader meant anything by them. On the samples fewer than one in six events is an intentional action.

Sample B - the 1,240 events of one session, by kind
Kind of eventCountShareDeliberate?
scroll position64051.6%no, recorded passively
pointer movement38831.3%no, recorded passively
click18615.0%yes
page view221.8%yes
form field completed40.3%yes, and the most sensitive
one session1,240100%212 of them, or 17.1%, were deliberate
sample B - how much of the log is a reader's own action deliberate events = click 186 + page view 22 + field 4 = 212 share of the session = 212 / 1,240 = 17.1% the other 1,028 events, or 82.9%, were recorded without the reader doing anything they would describe as an action of the four completed fields, three are masked in the replay and the fourth, an e-mail address, is not so the smallest group of events carries the most about the person.
The counts here are invented. What a particular site records, how long it keeps it and whether it masks a field are decisions of that site, and a reader can usually see part of the answer in its privacy notice and in the consent choices it offers. A page like this cannot tell a reader what a named operator does; it can only show what the machinery consists of.
What a reader can actually find out about their own events
  • Read the privacy notice for the list of event kinds, not the summary sentence about "usage data".
  • Check whether a session recording is named, because a replay is a different thing from a count.
  • Look for the retention period per category, since logs and recordings rarely share one.
  • Use the data-access route to request a copy, which turns a description into a list.
  • Turn off non-essential analytics where the choice exists, and remember it applies going forward only.

Read next

The funnelHow events become steps and ratesThe replayWhen the events become a pictureThe consent lineWhat may be recorded at allThe vendorsWho else receives the events