# On a high-traffic site will Clarity capture every session per variant?

Asked by Priyanka Menon on 2026-09-18. Tags: clarity, sampling, analytics.

Our site does well over a million sessions a month. My worry is that if
Clarity does not record everything, my per-variant replay pool ends up patchy
and I draw the wrong conclusion analysing behaviour that was only partially
captured.

What should I expect at that volume?

## 2 answers

### Answer from Croct Bot (2026-09-20)

Your instinct about the volume is right, and the limit comes from Clarity,
not from Croct. Clarity records up to 100k sessions per project per day and
then samples beyond that. At well over a million sessions a month you will
cross that threshold on busy days, so not every session gets a replay. That
sampling is a property of Clarity itself and is unaffected by whether Croct
is connected.

What the integration changes is only the ability to filter whatever Clarity
did capture by audience, experiment, and variant. It does not raise or lower
Clarity's capture ceiling.

The important part for your decision-making: Croct's own experiment
analytics remain unsampled across 100 percent of the data, so your
quantitative results are not affected by Clarity sampling at all. The
practical pattern is to use Clarity for qualitative spot-checks per variant
and Croct for the full quantitative measure.

This is noted on the [Clarity integration page](https://docs.croct.com/immersion/integrations/analytics/clarity),
and the reasoning behind [unsampled experiment results](https://docs.croct.com/reference/analytics/experiment)
is worth a read alongside it.

### Answer from Marta Kowalczyk (2026-09-21)

Can confirm from a high-traffic store. We do a few million sessions a month
and Clarity is definitely sampling on peak days, but it never mattered for
our conclusions. We read the win or loss off Croct's numbers, which see
everything, and only dip into Clarity to watch a handful of sessions per
variant to understand the behaviour. That split has worked fine for us for
months.

If bot traffic is also on your mind at that volume, the thread on
[unsampled results and how bot filtering fits in](/answers/unsampled-results-and-bot-traffic)
is relevant.