2026-08-07

Low traffic sites: why GA4 hides your best data

Your landing page got 380 visits last month.

One of them became a deal worth more than your annual marketing budget.

So you open the report to find out where that person came from. You get a row marked (not set), a table with fewer entries than you expected and a message about thresholds.

The tool you installed to answer exactly this question has decided not to tell you.

I want to be fair to Google here, because this is not a bug and it is not incompetence. It is a deliberate trade, made for a different kind of website than yours. Once you see which trade it is, the fix becomes obvious.

What GA4 actually does with small numbers

Two mechanisms get blamed for this and only one of them is really your problem.

Sampling is what people usually mention first. It is real but it works the other way round. Sampling kicks in when there is too much data, not too little so a site with a few hundred visits will rarely meet it.

Thresholding is the one that affects you. When a report could allow an individual to be identified, GA4 withholds the row. The likelihood of that goes up when demographic or signals data is involved and when the underlying counts are small.

Small numbers have a second enemy. On a quiet site a few hundred machine visits can double your monthly total.

The result is a report that looks broken but is behaving as designed. Rows vanish. Totals stop adding up. A dimension you care about shows (not set) where you expected a source.

On a site with 50,000 monthly visits this almost never bites. On a site with 380, it can hide the single most valuable fact of your quarter.

Why this trade is wrong for your site

A privacy threshold protects an individual from being singled out in a report. That is a legitimate goal and on a consumer site with large volumes it costs you almost nothing.

Now change the economics.

"The purpose of business is to create and keep a customer."

Peter Drucker, The Practice of Management, 1954

Drucker was making a point about what a business is for. It also describes what your measurement is for.

When a single customer is worth five figures, you are not doing statistics. You are trying to understand a handful of specific, high value events. Averages and percentages are almost meaningless at that scale and the individual path is the whole story.

Your tool applies a privacy control by suppressing exactly that path. You get the protection designed for a mass audience and you pay for it with the only data you needed.

The other half of the problem

Thresholding is not the only thing removing rows from your report.

Consent declines. On a B2B audience this hurts more than average, because the people you sell to are more likely to refuse.

Ad blockers and managed devices. Corporate networks filter tracker domains. Your ideal buyer sitting in an enterprise office may be invisible before their browser is even involved.

Small absolute numbers. When your month is 380 visits, losing 25% of them is not a rounding difference. It is the difference between seeing a pattern and seeing noise.

Stack the three and a micro site can end up measuring less than half of what happened, with no indication that anything is missing.

Still with me? Good, because the practical part is short and it is not what most analytics advice tells you.

What to measure on a micro site

Forget rates. Rates need volume to mean anything and you do not have volume. Measure counts and paths instead.

• Every individual source. Not the top five. All of them. On 380 visits the long tail is short enough to read and the interesting referral is usually in it.
• The full path, not the landing page. Which pages a visit touched and in what order. With few visitors you can actually read these and they tell you what the person was evaluating.
• Raw counts on the actions that matter. Form submissions, downloads, calls. Count them, do not rate them.
• The gap against your own records. Compare measured enquiries with the enquiries you actually received. On a small site you can reconcile this exactly, which is a luxury a large site never has.

That last one is the most valuable habit on this list. If your CRM shows six enquiries and your analytics shows four, you have learned something about your measurement before you learn anything about your marketing.

The trap of averages on small numbers

One warning, because low traffic sites make this mistake constantly.

Twelve conversions this month against nine last month is not a 33% improvement. It is three events and three events can easily be chance.

Do not redesign a page on a difference of that size. Wait for a pattern that repeats or judge the change on the quality of the enquiries instead of the count.

Small numbers give you rich detail about individual journeys and almost no statistical power. Use them for the first thing and be sceptical about the second.

Where StatJolt fits in

StatJolt does not apply reporting thresholds, because it never collects the personal data that thresholds exist to protect.

If five people visited, you see those five visits with their sources and their paths. Nothing is withheld to prevent identification, because there is nothing identifiable in the data to begin with.

The user journey view is built for exactly this reading. On a micro site you can follow what an individual visit did without profiling anyone.

There is also no consent banner so the B2B visitors who would have declined stay in your numbers. On 380 visits a month, that group is not a rounding error.

If you only read one box

GA4 withholds report rows when counts are small, which is a sensible privacy trade on a mass consumer site and the wrong one when a single lead is worth thousands. On a micro site, measure individual sources, full paths and raw counts and reconcile them against your own records rather than chasing rates. Three extra conversions is not a trend so use small numbers for detail and stay sceptical about the percentages.

Back