Understanding an application's experience | Docs

The Experience indicator qualifies how an application feels to use — Good, Average or Poor — based on the signals reported by Kabeen agents. Do not confuse it with the hardware health of a workstation, which aggregates CPU, memory, storage and battery: experience is about the application, health is about the machine.

The Experience indicator

The overall experience level (Good, Average, Poor) is a synthesis of two technical indicators measured over a rolling one-month window:

performance (loading latency), measured by the LCP;
reliability (errors / incidents), measured by the availability rate.

The level kept is the more severe of the two: if performance is Good but reliability is Poor, the overall experience is Poor (it is neither an average nor a weighting). When only one of the two indicators has data, that one determines the level; if neither has data, the experience is Undetermined (shown as "-" in grey).

Level	Colour
Good	Green
Average	Orange
Poor	Red
Undetermined	Grey (insufficient data)

Performance (latency)

Performance is based on the LCP (Largest Contentful Paint, the loading time of the page's main element). The score and value shown correspond to the 75th percentile (P75) of the LCP over the period — in other words, 75 % of loads are faster than this value. It is neither an average nor the maximum value.

Performance	Threshold (LCP P75)
Good	< 4 s
Average	from 4 s to 16 s
Poor	≥ 16 s

Reliability (errors / incidents)

Reliability is based on the availability rate: the proportion of successful requests over the period. A request is counted as an error / incident when it returns an HTTP code ≥ 500 (server error); client-side errors (4xx) are not counted.

Reliability	Availability rate
Good	≥ 99 %
Average	from 95 % to 99 %
Poor	< 95 %

Satisfaction

User satisfaction is presented as a separate indicator in the Experience tab. Unlike performance and reliability, it does not enter into the calculation of the overall Good / Average / Poor level: it complements the reading without determining it. See Application user satisfaction.

An application's experience (how it feels to use) is distinct from a workstation's health (its hardware state). For the latter, see Managing workstations.

Time spent

Time spent is the application's foreground time during the session, timestamped. Detection depends on the application type:

Web application: detected via the browser extension;
native application: detected via the process name.

An application that runs only in the background cannot be tracked, since it is never displayed in the foreground.

Average usage

Average usage is the number of distinct users who used the application over the period, shown in the chart on the Usage tab. A given user is counted only once, regardless of how many times they open it (it is neither a number of sessions nor a usage time). It can be filtered along three axes:

by team;
by period (from daily to yearly);
by application type (Web or Native).

Anonymization

The usage charts show the number of distinct users. Whether identities (user and workstation names) are displayed depends on the workspace's identity tracking mode (nominative, anonymous or forced anonymous) and on the permission to read users. See Data privacy.

Do not confuse

Machine readings (CPU / RAM), transmitted as snapshots about every 10 minutes, measure a workstation's load. They measure neither time per application nor application experience. For a workstation's load and health, see Managing workstations.