"Where did the extra power go?" is the usual way this problem arrives, and the first useful move is to slow down. A surprising number — "the battery dropped 20% overnight," "the van uses more than it should" — is not yet a fault. It is a discrepancy, and a discrepancy can come from a real new load or from a measurement boundary, a configuration setting, a timing mismatch, or an estimate that drifted. This guide is an evidence-led workflow for telling those apart — using only normal controls and closed-equipment observation, and knowing exactly where an owner's investigation has to stop.

Safety screen — read before you start. This is owner-level troubleshooting: displays, records, and normal user controls only. It never involves opening equipment, moving conductors, probing a shunt or fuse, or resetting protection repeatedly. If any of the stop conditions in the last section are present, this workflow ends and the system stays out of service until someone qualified checks it.

Define the discrepancy before explaining it

Two numbers can only be compared when they describe the same thing. Before reasoning about causes, pin down five properties of each value you are comparing:

  • Boundary — where is it measured? Battery outflow, an appliance's output, and a utility meter are three different places, separated by conversion losses and system scope.
  • Direction — is discharge positive or negative on this device?
  • Unit — energy (Wh) or charge (Ah)? They are not interchangeable.
  • Interval — the same clock window, start to end?
  • Operating context — same loads, mode, and conditions?

A battery-side watt-hour total cannot be compared directly with an appliance's output energy or a utility bill without accounting for what sits between them. Get the boundary and interval aligned first; most "mysteries" resolve here, before any hardware is suspected. If the units themselves are the confusion, what a battery monitor measures sorts measured from derived from estimated.

Rebuild what should have happened

Now build the expected side: what energy your known loads should account for over one defined interval, at the same boundary. This is the daily energy load profile method applied to a single day:

E_accounted = Σ(Pᵢ × tᵢ) + measured cycling-load energy + standby/idle + conversion

  Illustrative reconstruction at the battery-outflow boundary:
  E_accounted ≈ 1.00 kWh over 24 h   (≈ 41.7 W average)

Two traps live here. Cycling loads (a fridge, a pump) must be entered as measured or time-averaged energy, never a nameplate wattage times 24 hours. And standby draws are easy to omit entirely — the U.S. Department of Energy's standby-power guidance exists precisely because idle devices consume real energy that no one notices. Include inverter idle and control/monitoring draw at the chosen boundary. The average-power back-check (41.7 W here) is a quick sanity test: if it looks implausibly low, something is missing from the inventory.

Quantify the unexplained remainder

Subtract expected from measured, at the same boundary and interval:

Remainder   D = E_measured − E_accounted

  E_measured  = 1.30 kWh   (monitor-reported battery outflow, same 24 h)
  E_accounted = 1.00 kWh   (from above)

  D = 1.30 − 1.00 = 0.30 kWh   →  300 Wh  →  ≈ 12.5 W average over 24 h

The remainder defines the question; it does not name the cause. And it is only as sharp as its inputs: if the combined measurement-and-accounting uncertainty is about ±0.08 kWh, the real unexplained range is roughly 0.22–0.38 kWh, not a crisp 0.30. Quantify that band from your actual devices and estimates rather than treating the displayed total as exact. A 12.5 W average is small enough that a single forgotten standby device could explain it — which is why the next steps check the boundary and the estimate before hunting for a phantom load.

Check whether the data can see the whole path

A remainder can appear because the monitor cannot see part of the system. On the inspected shunt monitor, any load or charger wired on the battery side of the shunt is excluded from the current total and the SOC — and the readings still look plausible. So before concluding "something new is drawing power," ask whether the measurement boundary is even complete. A balancing worksheet is evidence, not proof: if both your monitor and your load inventory share the same blind spot, they can agree with each other and still both miss a circuit.

Confirming or correcting a sensor boundary physically — inspecting the shunt, tracing wiring — is qualified work, not part of this flow. What you can do as an owner is reason about which loads and charge sources are known to pass through the monitor, and flag the boundary as "unverified" when you cannot tell.

Separate an SOC symptom from an energy discrepancy

"Battery dropped 20% overnight" is an SOC symptom, and SOC is an estimate, so it can move without any change in real consumption. The manufacturer's troubleshooting list is instructive: an incorrect configured capacity, wrong Peukert or charge-efficiency settings, an incomplete current path, a lost synchronization, or plain accumulated drift can all shift the percentage on their own. Before you accept an SOC change as proof of energy use, look for direct evidence over the same interval — accumulated amp-hours or watt-hours, or measured current — and check when the monitor last synchronized to a full charge. A stale, unsynchronized SOC is a weak witness.

Preserve evidence before changing the system

The instinct to start flipping settings is the enemy of a diagnosis. Changing configuration can erase the very record that would tell you what changed first. Before you touch anything, capture the history the system already holds. Platform event logs — the VRM Portal manual is one current example — record which device raised which alarm, when it began and cleared, and any firmware, control, or setting change, down to the minute. That timeline often answers "what changed first?" by itself: a load that appeared, a setting someone edited, a firmware update, a communications gap. Screenshot or export it before you begin testing.

Test safe hypotheses one at a time

Now, and only now, test — reversibly, through normal controls, one variable at a time:

  1. Hold everything else constant — same boundary, same interval length, same conditions.
  2. Change exactly one reversible, user-level thing — switch off one appliance at its normal switch, disable one schedule, unplug one device.
  3. Re-measure over a comparable interval and see whether the remainder follows.
  4. Restore it, and move to the next hypothesis.

Changing one variable at a time is what makes the inference reproducible; changing three at once tells you nothing. None of this requires opening equipment, moving conductors, defeating a protective device, or repeated resets — if a hypothesis can only be tested that way, it is past the owner boundary.

Do not switch off a critical load to run a test. Medical equipment, refrigeration for food or medicine, heating or pumps preventing freeze or flood, communications, and security are not fair game for a comparison test unless their own plan permits it and an alternative exists. Classify these before you start, and leave them running.

Know when normal troubleshooting ends

Some conditions outrank every number on the screen. Stop the workflow, keep the system out of service, and get qualified help — following the applicable stop-use or emergency response — on any of:

  • burn marks, staining, exposed or damaged conductors, damaged insulation or enclosure, or arcing;
  • abnormal heat, odor, swelling, venting, hissing, leaking, or smoke from a battery or any component;
  • liquid ingress, or physical/impact damage;
  • repeated operation of a fuse, breaker, or protective device, or any fault you cannot explain.

The UK Health and Safety Executive's guidance on electrically powered equipment takes exactly this line for visible damage and overheating: take it out of use and have it checked by a competent person. A physical abnormality is not a data point to weigh against a normal display — it wins outright. For the battery-specific signals and what they mean, see when to stop using a battery system and the safety fundamentals. If you reach the end of the safe checks without a confirmed, benign cause, that is itself a result: preserve the record and route it to qualified support rather than pushing past the boundary.

Key takeaways

  • Treat surprising consumption as a discrepancy, not yet a fault — a boundary, setting, timing, or estimate can cause it.
  • Align boundary, direction, unit, interval, and context before comparing any two numbers.
  • Rebuild expected energy from measured cycling loads plus standby, then subtract: D = E_measured − E_accounted, and carry its uncertainty band.
  • A bypassed measurement path and a drifting or unsynchronized SOC both mimic new consumption; check them before hunting a phantom load.
  • Preserve event history before changing settings; test one reversible variable at a time.
  • Physical damage, heat, odor, swelling, arcing, or repeated protection trips end the flow — stop, stay out of service, get qualified help. Never interrupt a life-safety load to test.

Where to go next