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
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.
What it costs
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.
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.
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.
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
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.
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.
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.
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
Quality is not a single number. Each dimension fails in its own way, and each one is worth understanding on its own terms.