Production monitoring accuracy is a real specification, and it is usually looser than owners assume. The number in your app comes from sensors inside the inverter or from current transformers clamped onto conductors, interpreted by software, and reported at a point in the system that is not the same point your utility meter reads. Disagreement between the two is normal. Large or growing disagreement is a diagnostic signal worth chasing.
Two instruments doing different jobs
A revenue meter is a legally regulated instrument, tested to a defined accuracy class, sealed, and periodically verified because money depends on it. A monitoring system is a diagnostic and informational tool. Its sensors are chosen for cost and adequacy, not for billing-grade precision, and the manufacturer publishes a tolerance that is typically wider than a revenue meter’s. Neither is wrong; they are built to different purposes. When you compare an app figure to a bill, you are comparing a diagnostic estimate against a regulated measurement, and expecting them to match to the kilowatt-hour misunderstands both. Accuracy class also describes behavior across a range, not a single figure. An instrument that is accurate near full scale can be considerably less accurate at a small fraction of it, which is exactly where a solar system operates at dawn, at dusk and under heavy cloud. A monitoring channel sized for a large array will therefore report the low-light hours less precisely than the midday ones, and those hours accumulate across a year.
The measurement points are not the same
An inverter reports what it produced at its own output terminals. Between there and the utility meter sit conductors with resistance, sometimes a subpanel, and the entire consumption of the house. If your meter records net flow, it shows production minus consumption, which is a completely different quantity from gross production. Even with a dedicated production meter the reading is taken after the conductor run and after any parasitic loads on that circuit. Monitoring that uses clamp-on current transformers introduces another gap, since the accuracy of such a sensor depends on correct placement, correct orientation, the conductor being centered, and the sensor being matched to the expected current range. A transformer installed backward will faithfully report the wrong sign.
What else distorts the numbers
Sampling and rounding contribute. Systems that log at intervals and integrate between samples will miss short transients, and the error compounds differently in changeable weather than in steady sun. Communication gaps create holes that some platforms fill by interpolation and others simply omit, which changes a monthly total depending on which convention the software uses. Firmware updates occasionally change calibration or reporting conventions, creating a step in the data that looks like a hardware problem. Time zone and daylight-saving handling shifts energy between days. None of these are failures; they are why a monitoring total and a meter total drift apart in ways that are not a linear offset.
Using the discrepancy as a diagnostic
The useful practice is to establish the normal gap early and then watch it. If your app consistently reads a few percent above the production meter, that is your baseline, and it is uninteresting. If the gap suddenly widens, something changed: a sensor moved, a string dropped out of monitoring but kept producing, a module-level device stopped reporting, or the meter was replaced. A gap that grows slowly over months can indicate a current transformer drifting or a connection degrading. This is why writing down the relationship at commissioning is worth doing. Without a baseline, any later comparison is just two unrelated numbers.
Which number to use for which purpose
For anything financial, including incentive claims, net metering reconciliation and warranty disputes, use the regulated meter. For diagnostics, per-string comparison and spotting a failed module, use the monitoring system, because it has granularity the meter never will. For assessing degradation over years, use whichever source you have consistently, corrected for weather, and do not switch sources midway. And if an incentive program requires a specific class of production meter, make sure the one installed satisfies it, because discovering years later that the wrong instrument was used is an expensive way to learn the difference. It is also worth knowing how long your platform retains data and in what resolution. Many services keep high-resolution intervals for a limited window and then compress to daily or monthly totals, which erases exactly the detail a later investigation needs. Exporting the raw data periodically, or choosing a platform that allows a full export, preserves the record independently of the vendor’s retention policy.
What to check before you act
- Record the normal gap between app and meter during the first months.
- Verify current transformers are oriented and placed as the manual specifies.
- Note firmware updates, since they can create steps in the data.
- Use the regulated meter for anything financial, monitoring for diagnostics.
Common questions
My app shows more production than my bill credits. Am I being shortchanged?
Usually not. If your meter records net flow, it shows production minus household consumption, which is always lower than gross production. Check which quantity each instrument reports.
Can monitoring accuracy be improved?
Sometimes, by relocating or replacing current sensors or adding a dedicated production meter. Ask whether the platform supports calibration against a reference before spending on hardware.
The practical takeaway
Expect the app and the meter to differ, learn your normal gap early, and treat a change in that gap as the signal. Use the regulated meter for money and the monitoring platform for finding faults, and leave any sensor work to a qualified installer.
References and further reading
- NREL PVWatts: modeled production to compare against measured output
- NREL: photovoltaics research including field performance measurement
- NREL: a technical report from the laboratory’s photovoltaics publication library
Numerical scenarios are illustrative unless identified otherwise. Follow the exact product instructions; component ratings and local installation requirements can differ.
