Time between events

Time between events is the average elapsed time that separates two defined events in a sequence, measured across many entities such as users, orders, or tickets. You pick a starting event and a paired ending event, calculate the gap for each entity, then average those gaps into a single duration. A subscription product might track the time between account creation and first invite sent. A support desk might track the time between a ticket being opened and the first agent reply.

The metric matters because most journeys are sequences of steps, and the speed of movement between steps is often a better signal than the steps themselves. Two cohorts can have identical conversion rates while one moves twice as fast. Faster movement usually means lower friction, stronger intent, and earlier value, all of which tend to improve downstream retention rate.

Time between events is a building block rather than a single fixed metric. Time to first payment, time to first response, and gap between repeat purchases are all specific instances of it. Whenever a question takes the form of how long does it take to get from one action to the next, time between events is the underlying measure.

Time Between Events = Average (Event B Timestamp - Event A Timestamp)

For each entity that completed both events, subtract the timestamp of the earlier event from the timestamp of the later event, then average the results. If 500 users signed up and later made a first purchase, and the combined gap across them is 6,000 hours, the average time between events is 12 hours.

The definition of the two events does most of the work, so define them precisely. Decide what counts as the start and end, decide what happens when an entity never completes the second event, and decide the unit of measurement, whether seconds, hours, or days. Excluding entities that have not yet completed the second event prevents you from understating the interval, but it can hide a growing tail of people who stall.

1. 1

   Starting event

   The earlier action that opens the interval, recorded with a reliable timestamp. Choose an event every entity reaches, such as account creation or order placement, so the population is well defined.

2. 2

   Ending event

   The later action that closes the interval. It must be unambiguously linked to the same entity as the starting event, usually by a shared user or order identifier.

3. 3

   Matching rule

   The logic that pairs each starting event with the correct ending event. For repeatable events, decide whether you measure the gap to the next occurrence or to the first occurrence after the start.

4. 4

   Censoring rule

   How you treat entities that completed the start but not the end within the window. Either exclude them or cap their interval at the window length, and state which, because it materially changes the average.

A metric tree turns time between events from a single number into a map of what controls it. The headline interval breaks down into the durations of the individual stages between the two endpoints, and each stage has its own causes and its own owner. Decomposing the gap this way shows you exactly which stage is slow rather than just telling you the overall journey is slow.

There is no universal benchmark for time between events because the right interval depends entirely on which two events you chose. A useful number is the one measured against your own history and your own segments. That said, broad ranges exist for the common journeys this metric is applied to, and they give a rough sense of what good looks like.

Read the trend before the absolute value. A median interval that is steadily falling means the journey is getting smoother, even if the number is still higher than a published benchmark. A rising interval at stable volume is the warning sign, because it means friction is creeping in somewhere between the two events. Always pair the average with the share of entities that complete the second event at all, since a fast interval among a shrinking population is not the win it appears to be.

Shortening the interval means removing delay from the slowest stage, not from the journey in general. Find the stage that contributes the most elapsed time, fix the specific cause there, then re-measure to confirm the overall gap actually moved.

1. 1

   Reporting the mean on a skewed distribution

   Time-based metrics almost always have a long tail. A few very slow entities pull the mean far above the typical experience. Use the median, or report both, so the headline reflects what a normal entity actually sees.

2. 2

   Silently dropping incomplete journeys

   Excluding entities that never reach the second event makes the interval look faster while hiding a growing group of people who stalled. Track completion rate alongside the interval so the two cannot diverge unnoticed.

3. 3

   Mismatched event pairing

   For repeatable events, measuring the gap to the wrong occurrence produces a meaningless number. Define whether you mean the next occurrence or the first occurrence after the start, and apply it consistently.

4. 4

   Ignoring segment mix shifts

   A rising interval can be caused entirely by a change in who is in the population, not by any stage getting slower. Always check whether the segment mix moved before concluding the process degraded.