Validation, estimation and editing — and a record of which was applied.
Communications drop, a clock drifts, a CT is disturbed during maintenance. The readings still arrive and they still look like numbers. MOL applies a rule-based VEE process to every interval, across every utility, and keeps the outcome attached to the value — so what reaches billing carries its own audit trail rather than arriving as a clean number of unknown origin.
The expensive reading is the plausible one.
A gap gets noticed and fixed. The costly case is the interval that is present, within range, and wrong — a CT reconnected reversed after maintenance, a clock ninety seconds out, a channel mapped to the wrong quantity.
Nothing downstream will flag it, because nothing downstream has anything to compare it against. It gets billed and reported at face value until something forces a reconciliation, by which point the correction spans several cycles.
Every reading carries its own history
MOL applies a rule-based validation, estimation and editing process to all raw data, regardless of which utility it measures. Each value ends up in one of four states, and it keeps that state for as long as it exists.
Raw
As it came off the meter, before any check has been applied. Never billed from.
Valid
Passed every required check — or failed one, was investigated, and was confirmed. This is what billing runs on.
Verified
Failed at least one check, but was examined and found to represent actual usage. A genuine spike is still a genuine spike.
Estimated
Could not be recovered, so it was calculated by rule — and it carries that label permanently.
The distinction that matters in a dispute is the last one. An estimated reading is not a problem. An estimated reading that nobody labelled is the problem.
Checked three times, at three different moments
Some things can only be caught at the meter, some only once a month has closed. So the checks are split by when they can usefully run.
01
As the data is read
- Meter clock against system time, before anything is stored
- Communications integrity and expected record count
- Interval size, meter constants and channel-to-quantity mapping
02
Any time before the cycle closes
- Gaps — is this interval missing, or is it a genuine zero?
- Spikes and step changes against the meter's own history
- Reactive energy plausibility against active energy
- Reverse energy where no generation exists
03
At the end of the billing period
- Usage against the same month last year, or last month if there is no year
- Register readings reconciled against the sum of the intervals
- Estimated intervals re-checked for reasonableness
A failed check raises an exception, not an estimate.
Auto-estimating every failure produces a complete dataset and a clean month-end, at the cost of an accumulating layer of derived values that cannot afterwards be separated from measured ones.
MOL escalates instead. A failure is inspected against the meter's recent history, against the same period last year, and where relevant against a known change on site. Data that proves genuine is verified and retained at full value. Only what cannot be recovered is estimated, and the label persists.
The distinction survives into billing, reporting and any subsequent audit, which is what makes an estimated interval a disclosed assumption rather than a hidden one.
Want to see this on your own data?
We will run it against a month of your readings and show you what comes out.
Talk to us