What is data sampling?

How does sampling work?

Estimated total = value in sample ÷ sampling rate
(e.g. 1,200 conversions in a 10% sample → ~12,000)

Tools sample to keep large queries fast. Sampling can happen at collection (only some hits are recorded) or at query time (a report reads a slice of stored data). Google Analytics 4 standard reports are unsampled, but explorations and other ad-hoc reports can be sampled when a query exceeds the property's event quota — for standard properties, 10 million events per query.

Sampling example

A site with 40 million events a month builds an exploration for a niche campaign that had 90 conversions. With a 25% sample, the tool sees about 22 conversions and reports ~88 — or 72, or 104, depending on which slice it read. For the overall total the error is negligible; for a small segment it's large enough to flip a decision.

Why sampling matters

• Small segments (one campaign, one country, one landing page) get the largest relative error.
• Rare events — conversions, payments — are exactly what you most want precise.
• Numbers can change between two runs of the same report.

How to avoid sampling

• Use standard reports instead of custom explorations where possible.
• Shorten the date range.
• Export raw events (for GA4, the BigQuery export) and query them yourself.
• Choose a tool that processes every event for every report.

Does VisitTrack sample data?

No. Every report in VisitTrack is calculated from every recorded event — there's no sampled mode and no query quota that triggers one. Crossing your plan's event tier never pauses tracking or samples data either; it changes your next invoice. Bot visits are excluded by bot filtering, which is filtering by rule, not sampling. Raw data is available through CSV export and the API.