Platform
Storing the number is the easy part. Storing what it means is the rest.
Field measurements are retained at full resolution and queryable across arbitrary ranges — and every stored value keeps the four facts that let it be defended later.
- JR14.TK03.LEVEL14.62ftgood
- JR14.MTR.GASRATE318.4MCF/dstale 41m
- JR14.TK01.LEVELno datano comms
Full-resolution retention
Measurements are stored at the resolution they arrived at, and remain queryable across arbitrary ranges rather than only recent windows.
- History is not thinned into averages that destroy the event you later need to find.
- Queries span arbitrary ranges rather than fixed reporting periods.
Four facts on every value
A stored measurement is never a bare number and a timestamp.
- The source device that produced it.
- Its units, retained and never silently converted.
- Its acquisition time, distinct from the time it was stored.
- Its quality flag.
Late, interpolated and failed stay distinguishable
A value that arrived late, arrived interpolated, or failed a quality check is marked as such for as long as it exists.
- Backfilled data is identifiable as backfilled.
- An interpolated value never passes as a measured one.
- A value that failed a check is retained and flagged rather than dropped.
Feeds everything downstream
Trends, alarm evaluation, reconciliation and exports all read the same stored history, so the number on a chart and the number in a report have one origin.
- One source of history rather than a separate store per surface.
- Exports carry the same provenance the interface shows.
Why quality flags have to survive storage
Quality information tends to get lost at write time. A reading is validated on the way in, the validation result informs a decision about whether to store it, and then only the number is kept. Six months later the number is indistinguishable from any other.
That is the failure this design avoids. The quality state is stored with the value, permanently, because the question it answers — should I trust this specific reading — gets asked long after the moment it arrived.
It also means a bad reading is worth keeping. A sensor that drifts produces a recognisable pattern in flagged values, and that pattern is only visible if the flagged values were retained rather than discarded.
Common questions
Does gauge capture go into the historian too?
Manual gauges and derived volumes are stored with the same provenance discipline and are available to the same downstream surfaces. They carry different provenance from a telemetry reading, and the platform records which is which.
Is history ever rewritten?
A correction never overwrites the original. It is stored as a new record referencing the one it amends, so history remains reconstructible as of any point in time, and figures that changed as a result are flagged as restated.
See it against your own wells.
A demo runs against synthetic data first, then against a read-only connection to your system of record if you want to see real numbers.