Sentys — Export spec (one page)

Everything below comes straight out of your SRO/SCADA or historian. Expect it to take about twenty minutes, most of it waiting for the export to finish.

The export — five lines

One machine you care about — a pump, a heat exchanger, a motor. The one that surprises you.

All its logged tags: temperatures, pressures, flow, current, vibration, speed, valve/setpoints. Don't filter — more tags is better.

1-minute averages are plenty; hourly still works. Whatever your system exports by default.

6–12 months of history — ideally spanning a breakdown or repair you remember. So the report can check whether it had warning.

CSV or JSON, any delimiter. Semicolons, decimal commas, Danish headers — all fine.

Format details (for whoever runs the export)

Aspect	Accepted
CSV	Any delimiter (comma/semicolon/tab), decimal comma or point, one header row. One row per timestamp is ideal; one file per machine.
JSON	Array of records, JSON Lines/NDJSON, one wrapped array (`{"readings": [...]}`), or column-oriented lists. Nested objects are fine.
Timestamps	Any consistent format (ISO 8601 preferred). Local time is fine — say which.
Units	Whatever you log. A tag list with units/descriptions is helpful if it's one click, not required.
Gaps & noise	Expected. Don't clean anything — missing values and sensor glitches are information.
Size	Anything up to a few GB is fine. If the file is too big for email, we'll agree a transfer that suits your IT.

Aspect

Accepted

CSV

Any delimiter (comma/semicolon/tab), decimal comma or point, one header row. One row per timestamp is ideal; one file per machine.

JSON

Array of records, JSON Lines/NDJSON, one wrapped array ({"readings": [...]}), or column-oriented lists. Nested objects are fine.

Timestamps

Any consistent format (ISO 8601 preferred). Local time is fine — say which.

Units

Whatever you log. A tag list with units/descriptions is helpful if it's one click, not required.

Gaps & noise

Expected. Don't clean anything — missing values and sensor glitches are information.

Size

Anything up to a few GB is fine. If the file is too big for email, we'll agree a transfer that suits your IT.

What you get back — within a week

A written findings report (one self-contained file, forwardable).
What drifted and when — health-over-time per machine, warning points marked.
Which sensor saw it first, and how far from healthy it moved.
The breakdown check: whether your last failure had warning in the data.
Every caveat stated — including "we found nothing", if that's the truth.

How your data is handled

Processed on a single machine in Denmark; raw data never uploaded to any cloud service.
Stated up front: schema inference sends column names, statistics, and a few sample values to an AI API — never the bulk data. Object, and it's done by hand.
Used for your analysis only; never used to train models for anyone else.
Deleted when you say so — confirmed in writing. A mutual NDA is signed before any data moves.

What to export — and what you get back

The export — five lines

Format details (for whoever runs the export)

What you get back — within a week

How your data is handled