Why Timestamps and Source Images Matter

Why this lesson matters

Two fields carry disproportionate weight in a monitoring record. One makes the reading calculable. The other makes it defensible. Most sites have neither, and both are essentially free to add.

Part one: the timestamp

Capture time is not entry date

The date written on a sheet is normally the day of transcription. The capture time is the moment the technician stood in front of the instrument. These can differ by days, and when the value is cumulative that difference propagates directly into every rate you calculate.

Why elapsed time, not nominal period

A "weekly" round is rarely seven days. Monday 6 am to Friday 4 pm is 4.4 days. The next reading, Friday 4 pm to Monday 7 am, is 2.6 days. If you divide each period's consumption by seven you will report one period as far too low and the next as far too high, and you will then spend time investigating a variation that exists only in your arithmetic.

The correct form is always:

rate = consumption / actual elapsed time between captures

which requires the timestamp to the minute on both readings.

Time of day changes the meaning of an instantaneous value

A boiler pressure at 5 am on a cold start and at 2 pm at steady load are different measurements of a healthy plant. Compare them without timestamps and you will see a trend that does not exist. Any instrument whose value depends on load, occupancy, ambient temperature or plant sequencing is uninterpretable without the time.

Practical timestamp rules

• System generated, never typed. Any human entered time will eventually be rounded to the hour, copied from the row above, or guessed.
• Store with a timezone or in UTC. Australian sites crossing daylight saving boundaries will otherwise produce a duplicated hour once a year and a missing hour once a year, and the affected deltas will be wrong.
• Capture at the instrument. If readings are queued offline, the stored timestamp must be the observation time, not the time the queue synchronised. This distinction matters enormously for basements, pits and plant rooms with no signal.
• Preserve the distinction between captured at and recorded at. Both are useful; conflating them is not.

Timestamps make the schedule enforceable

Once every reading has a real capture time, the system can answer questions a sheet never could:

• Is this asset overdue against its schedule, and by how long?
• Was this round completed within its grace period?
• What is reading coverage for the period, with an explicit denominator?
• Are readings clustering at the end of the month, which suggests retrospective completion rather than observation?

Part two: the source image

What the image actually does

A photograph of the register or dial, bound to the reading record, converts an assertion into a verifiable observation. Three specific functions:

Verification. Anyone can confirm the recorded value against the instrument face without a site visit. Transposition errors, which are otherwise undetectable, become trivially findable.

Preservation of detail beyond the recorded value. Trailing digits, needle position within the graduation, unit and scale markings, tag plate, and the physical condition of the instrument and its fittings.

Traceability. The evidence chain runs instrument, image, value, person, time. That chain is what survives scrutiny in an audit, a dispute over a utility invoice, or a failure investigation.

Image capture standards worth enforcing

• Frame the register or dial face square on to avoid parallax on a needle.
• Include the asset tag in the frame where physically possible. Where it is not, rely on the asset key in the record and photograph the tag separately once.
• Capture the full index, including any trailing or red digits, not just the digits being recorded.
• Avoid flash reflection off glass; move the angle rather than relying on the flash.
• If the instrument cannot be read, photograph it anyway and flag the reading as unreadable with the reason. An image of a fogged glass is far more useful than a blank cell or, worse, a carried forward value.

Storage and retention

• Store images in private storage, bound to the reading, not in a shared gallery or a mail thread. An orphan image file is not evidence.
• Keep the original capture, uncompressed enough that the index remains legible when zoomed.
• Retain for at least the period your reporting obligations require, and remember that energy and emissions reporting typically expects records to be kept for several years.
• Treat images as immutable. If a reading is corrected, the original image stays attached to the original value and the adjustment carries its own record.

How the two fields work together

Question	Needs timestamp	Needs image
What was the consumption rate last week?	Yes	No
Is this asset overdue for a reading?	Yes	No
Did the technician actually read this instrument?	Partly	Yes
Is 12,437 a transposition of 12,347?	No	Yes
Was this reading taken during production or overnight?	Yes	No
Can this figure be defended in an audit?	Yes	Yes
Was the glass fogged when this was read?	No	Yes

Neither field is sufficient alone. Together they take a manual reading to a standard that is, for most monitoring and reporting purposes, equivalent to an automated point.

Common mistakes at this stage

• Recording the sync time as the capture time on offline captures. This silently breaks every rate calculation on poor signal sites.
• Storing images without binding them to the reading record. A folder of photographs named by date is not traceability.
• Compressing images below register legibility to save storage. The whole value is in the digits.
• Ignoring timezone and daylight saving. Two broken periods a year, every year, quietly.
• Allowing a carried forward value when the instrument is unreadable. Flag it. A visible gap is honest data; a fabricated value is permanent contamination.
• Deleting the original image after a correction, which removes the evidence for the very reading most likely to be questioned.