Analytics

Session Sampling Explained — What Sampled Analytics Hides From You

Many session replay and analytics tools record only a fraction of your visits. Learn what sampling is, when it is harmless, when it hides the exact sessions you need, and how to check whether your tool samples.

Purushottam Kumar Suman
Purushottam Kumar SumanSeptember 13, 20269 min read
Founder & CEO, DeepSync
Chart showing a sampled subset of website sessions

Short answer: sampling means your tool records or analyses only a portion of visits and treats it as representative. That is fine for broad trends on high-traffic pages. It fails when you need a specific session, such as the one where a customer's payment broke.

Table of Contents

  1. What Sampling Is
  2. Why Tools Sample
  3. When Sampling Is Harmless
  4. When Sampling Hurts
  5. How to Tell Whether Your Tool Samples
  6. How DeepSync Handles It
  7. Key Takeaways

What Sampling Is

There are two kinds, and they are easy to confuse.

TypeWhat happensWhere you see it
Capture samplingOnly some visits are recorded at allSession replay tools with a recording rate
Query samplingAll data is stored, but reports are calculated from a subsetAnalytics tools on large date ranges

Capture sampling is permanent. A session that was not recorded cannot be recovered later. Query sampling is an estimate you can sometimes avoid by narrowing the date range.

Why Tools Sample

Recording a session costs storage and processing. A tool with a fixed monthly quota has two ways to spread it across a busy month: record a percentage of every day's traffic, or record everything until the quota runs out. Vendors often choose the first because the dashboard never looks empty.

Neither choice is dishonest. But they produce very different data, and many teams do not know which one they have.

When Sampling Is Harmless

  • Heatmaps on high-traffic pages. A click map built from 20,000 visits looks the same as one built from 100,000.
  • Broad trend lines. Whether mobile traffic is growing does not need every session.
  • Qualitative browsing. If you only want a general feel for how people use a page, any twenty sessions will do.

When Sampling Hurts

Support and debugging

A customer writes in saying checkout failed. If you record one visit in five, there is an 80% chance that session does not exist. This is the most common way teams discover their tool samples.

Rare events

Problems that affect 1% of visitors are the ones nobody notices and the ones that quietly cost money. At a 10% sample rate you would see a tenth of them, which often is not enough to recognise a pattern.

Low-traffic pages

A pricing page with 300 visits a month sampled at 20% leaves 60 sessions. Split those by device and you are making decisions on a handful of recordings.

Funnels

If each step is sampled independently, conversion between steps becomes an estimate of an estimate. The steps with the fewest visitors, which are usually the ones you care about, have the largest error.

Segments

Sampling error grows as you filter. "Mobile Safari users from paid search who reached the cart" may be a few dozen real sessions. Sampled, it may be none.

How to Tell Whether Your Tool Samples

  1. Read the plan page for words such as "up to", "daily limit", or "recording rate".
  2. Compare session counts with your analytics tool for one page on one day. A consistent ratio such as one in four suggests a fixed rate.
  3. Search for a session you know happened. Visit your own site, note the time, and look for it.
  4. Look for an indicator on reports saying the result is based on a percentage of data.
  5. Ask support directly: "Is every session recorded until my quota is reached, or is a percentage recorded?"

How DeepSync Handles It

Every DeepSync plan records 100% of sessions up to the plan limit. Nothing is sampled. What changes between plans is volume, retention, and which analysis features are unlocked (plans overview).

The honest trade-off is what happens at the limit: recording pauses for the rest of the billing month and resumes at the start of the next one, or immediately if you upgrade. You are never charged an overage. So a busy month can end with a gap at the end rather than thin coverage throughout (session limits).

Protect your quota before you upgrade

Internal and staging traffic often use a surprising share of a session allowance. Excluding your own team's visits frequently buys back more headroom than a tier change. See how to exclude internal traffic.

Key Takeaways

  • Capture sampling is permanent; query sampling is an estimate.
  • Sampling is fine for trends and bad for specific sessions, rare events, and small segments.
  • Check your plan wording and test with a session you know happened.
  • DeepSync records everything up to the limit and pauses rather than sampling.

Conclusion

Sampling is a reasonable engineering trade-off. It becomes a problem when you do not know it is happening and go looking for a session that was never recorded.

Start free and record every session up to your limit.

Frequently Asked Questions

Was this article helpful?

Ready to understand
users like never before?

Join thousands of teams who use DeepSync to uncover insights,improve experiences, and build better products—faster.

Quick & easy onboarding
See results in real time
Enterprise-grade security