Choosing sampling interval and study duration

Two parameters decide what a mapping study can prove: how often each sensor records, and how long the study runs. Both are usually set by habit. Both are testable, and the interval in particular has a threshold beyond which the study stops being able to see the thing it exists to find.

What a slower interval actually costs

The table below is one 24-hour cold-room trace, generated at 1-minute resolution with compressor cycling, two defrost cycles and a goods-in door opening, then resampled at each interval exactly as a logger set to that interval would record it. Limits 2 to 8 °C. The true event is 17 minutes out of limits, peaking at 9.64 °C.

Read the 30-minute row carefully. The logger did not record a smaller excursion, or a shorter one. It recorded a clean day. The maximum in the dataset is 7.22 °C, comfortably inside limits, and there is nothing in the record to investigate, report or learn from. The event happened; the evidence does not exist.

The 15-minute row is the subtler failure. The excursion is flagged, so the study looks like it worked, but the recorded peak is 8.56 °C against a true 9.64 °C. Any risk assessment based on that peak is assessing an event that was more than a degree hotter than the number in front of it.

A slow logger does not average or smooth the peak it missed. A value between two readings is simply absent, so the recorded maximum understates the real worst case, and by an amount nothing in the dataset reveals.

Choosing the interval

• One to five minutes for any storage-area mapping study. This is the range WHO Technical Supplement 8 and the EU GDP guidance are written around, and the table above is why.
• One minute where the space has fast transients: a walk-in freezer with frequent door use, a small refrigerator, anything where a defrost cycle is short and sharp.
• Match the monitoring system's interval if the mapping is being used to justify monitoring sensor placement. Mapping at 1 minute and monitoring at 30 gives you two incomparable datasets.
• Never choose the interval to keep the file small. At one minute a 72-hour study on 12 sensors is about 52,000 readings, which is a trivial spreadsheet and a few hundred kilobytes.

Choosing the duration

Duration has to cover at least one complete operating cycle of the space, and for a refrigerated area that means at least one full defrost. A study that stops before a defrost has not tested the condition most likely to produce an excursion.

• 72 hours minimum for a storage area is the common expectation, and it is what an assessor will look for.
• Long enough to include a defrost cycle, with the defrost timing confirmed from the plant rather than assumed.
• Long enough to include normal operational use: deliveries, picking, door openings. An empty weekend is not representative of a working warehouse.
• Seasonal repetition for spaces whose performance depends on ambient conditions. Summer and winter studies answer different questions, and a single study in October answers neither.

Recording the choice

Both parameters belong in the protocol before the study starts, with the reasoning: the interval chosen against the transients expected, and the duration chosen against the operating cycle. Deciding afterwards, or leaving them at whatever the loggers were last set to, is the difference between a study and a data collection.