Common temperature mapping mistakes

• Sizing the study to the loggers available instead of the space.
• Uncalibrated or out-of-date loggers, or calibration at a different range from the study.
• No unique sensor identifiers or location descriptions, so the data cannot be mapped back to the space.
• Sensors on surfaces, in airflow from a diffuser, or at ceiling height rather than product height.
• Sampling intervals too long to see door events and defrost cycles.
• Studies too short to include a weekend, a delivery day or a defrost.
• Temperature unit assumed from the values rather than confirmed.
• Timestamps in mixed time zones or with daylight-saving jumps, merged without checking.
• Deciding the hot/cold-spot criterion after seeing the data.
• A report that never says what the study could not show.

ValiTrac's pre-flight and data-quality stages exist to catch the metadata, unit and timestamp problems on this list before any statistic is computed — and to refuse to guess the unit.

The mistakes that cost the whole study

Some errors cost a sensor, some cost a conclusion, and a few cost the study. It is worth knowing which is which before the loggers go in, because the expensive ones are all avoidable in the planning hour.

Sizing the study to the loggers you own

This deserves its own heading because it is the most common of all, and the least often admitted. A plan that calls for eighteen sensors and a cupboard containing twelve produces a twelve-sensor plan and a rationale written backwards to justify it. The tell is a protocol whose sensor count happens to exactly match the equipment inventory.

The honest routes are to hire the extra loggers, or to run the study in phases with a set of common reference positions instrumented in every phase so the datasets can be stitched together. Both cost something. Both are cheaper than a study an assessor rejects.

The quiet ones, found at analysis

• Clocks not synchronised before deployment. Loggers drift, and a five-minute spread across twelve devices makes a coordinated door event look like six separate ones.
• Daylight saving falling inside the study window, producing a duplicated or missing hour that most software silently reorders.
• Loggers not equilibrated before the clock starts, so the first hour records the corridor they were carried from.
• Opening the door to check on the loggers, adding a door event to your own dataset.
• A sensor taped to a rack upright rather than in the air, reading the steel.
• Lighting, heating or plant in a different state during the study than in routine use, and not recorded either way.

Every one of these is cheaper to prevent than to explain. The protocol review is the moment: read it as though you were the assessor, and ask of each requirement how it will be evidenced.