The whole picture, not the part that happened to arrive
Data quality
Completeness is knowing that what you are looking at is everything there is. When a sensor drops out or a site stops reporting, the numbers still add up — they are just quietly wrong. A missing measurement should show up as a missing measurement, not as a slightly lower average.
What it means
Completeness is coverage, not volume
A record is complete when every measurement that should be there is there, and when the ones that are not can be named. Size is not the test — a very large set of readings can still be missing the one hour that mattered.
Most operations never find out. The report renders, the average looks reasonable, and nobody asks which assets stopped reporting last Thursday.
- Every asset that should be reporting is reporting, and you can see which ones are not
- Holes in a history stay visible as holes rather than being smoothed into an average
- A late reading fills its own place in the record instead of landing at the end
- What is missing is described as clearly as what is present
What it costs
A partial picture, priced at full confidence
Incomplete data rarely announces itself. It surfaces later, in a decision that was reasonable given what was on the screen and wrong given what was not.
Decisions made on whatever arrived
A planner comparing two sites does not know that one of them stopped sending readings halfway through the period. The comparison is made in good faith, and the site with less data looks calmer than it actually was.
The gap nobody found until the audit
Missing records tend to surface when someone outside the business asks for a full year and the year has holes in it. By then the equipment that would have explained the gap has been serviced, replaced or scrapped.
Trust that erodes quietly
Once an engineer finds one absence the system never flagged, they start checking everything by hand. The platform keeps running, but the team goes back to spreadsheets for the numbers that actually matter.
Our approach
Absence should be as visible as presence
Nothing can make a sensor report while it is offline. What a platform can do is make sure the silence is recorded, attributed to a source, and impossible to mistake for a low reading.
Every measurement stays tied to its source
A reading carries the asset it came from, so the full set of things that should be reporting is known rather than only the things that did. A feed that goes quiet stands out against that expectation instead of disappearing from the picture.
Nothing is quietly averaged away
Where a value is absent it stays absent, rather than being filled with a nearby number to make a chart look tidy. A chart with an honest hole in it starts a conversation; a smooth line ends one.
History stays open to data that comes late
Field data arrives out of order — a link comes back up, a logger is collected, a correction is issued weeks later. Records land in the period they belong to, so a history that was incomplete on Monday is genuinely more complete on Friday rather than merely longer.
The other dimensions
Six dimensions, one standard
Quality is not a single number. Each dimension fails in its own way, and each one is worth understanding on its own terms.