Temperature mapping data analyser

How the calculation works

1. 01The gradient is the spread across sensors at the same instant, reported as both the worst case and the study mean. Instants where fewer than two sensors reported are excluded, because treating a lone survivor as a spread of zero would flatter a study whose other channels had failed.
2. 02Hot and cold spots are identified by each sensor's mean over the whole study, which is the figure that says where permanent monitoring sensors belong — separately from the single highest and lowest readings, which can come from a different sensor entirely.
3. 03Mean kinetic temperature follows ICH Q1A(R2) with ΔH/R = 10 000 K, reported both pooled across every reading and for the warmest channel on its own.
4. 04Per-sensor statistics reuse the same summary as the logger data check: count, minimum, maximum, mean, standard deviation, MKT and readings outside the limits you set.
5. 05Data quality is assessed before anything else: dead channels, channels that stopped part way, channels with no variation at all, duplicated columns, irregular sampling and studies too short to show a daily cycle.

Limitations

• It sets no acceptance criteria. Limits and the permitted gradient are yours to enter, and no verdict is offered until you do — the figures that decide whether a mapping study passes come from your protocol and the standard you work to.
• It reads columns, not positions. It can say which channel ran warmest but not that the warmth was at ceiling height beside the evaporator, so it needs the position table from your protocol alongside it.
• It carries no measurement uncertainty. A gradient of 3.1 against a limit of 3.0 is neither a pass nor a failure when the loggers carry ±0.3; that is a decision rule question for the guard band calculator.
• Timestamps must be year-first. A date like 01/02/2026 is January in one country and February in another, so the tool declines to guess and reports on row order instead.
• It is not a mapping report. A report also needs the protocol, the loggers' calibration status, the load condition, the rationale for each position and a conclusion a qualified person signs.

Frequently asked questions

What is a temperature gradient in a mapping study?

The gradient is the difference between the warmest and coldest sensors at the same moment — a snapshot of how far apart the room's extremes are while it is doing whatever it was doing. It is not the difference between one sensor's maximum and another's minimum recorded hours apart, which is a larger and much less meaningful number. This analyser reports the worst single instant, which is the figure a mapping report normally quotes, alongside the study average so you can see whether that worst case was typical or was one event.

How do I find the hot and cold spots from mapping data?

Rank the sensors by their mean over the whole study rather than by their peak. The peak can belong to a position that spiked once when a door opened, while a different position ran consistently warmer for the entire week — and it is the consistently warm one where a permanent monitoring sensor belongs. This tool reports both, separately, because confusing them is a common way for a study to be run properly and then used badly.

What data quality checks should I run on a mapping study?

Before any statistics, check that every channel actually recorded: that none is missing, that none stopped part way through, that none is reporting a frozen value, and that no two columns are identical, which usually means one logger was exported twice. Then check the sampling was continuous, because an excursion that fell into a gap is one nobody can report on. A study with a failed channel still produces tidy statistics from the survivors, and that tidiness is exactly what makes the problem easy to miss.

Is my mapping data uploaded anywhere?

No. The file is parsed and every figure computed in your browser, and nothing is sent, stored or logged. Mapping data describes a real facility — its layout, its weak points and when its doors open — so there is no good reason for it to leave your machine to have a mean taken. You can confirm it by watching the network tab, or by disconnecting and seeing the tool keep working.

How many sensors does a mapping study need?

That is not a question this tool answers, and it is worth being wary of any free tool that does. Expectations come from documents such as WHO TRS 961 Annex 9 and the EU GDP guidelines, which are not public domain, so a number produced here would be an invention landing in a protocol somebody has to defend. What this tool will tell you is what the study you actually ran covers, including when only two channels carry data — a difference between two points is not a map.

What format should the export be in?

A timestamp column followed by one column per sensor is what nearly every logger platform produces, and it is read directly. A long export of timestamp, sensor, value is detected automatically. Commas, semicolons and tabs all work, quoted cells are handled, and a title block or footer from the logger software is skipped and counted. Empty cells stay empty rather than being read as zero, which would otherwise drag a mean down and invent an excursion.