A smart plug that looks right on a kettle can read way low on your gaming PC, and that gap is not just a rounding quirk. For steady, simple loads the energy count usually lines up, but for things that turn on and off fast or use motors and chargers the plug is guessing more than measuring.
The real difference is what happens inside: a heater’s math is simple, but a PC or fan forces the plug to estimate power behavior and smooth the numbers over time before showing them. true-power meter accuracy research shows even a lab-grade meter slips from about 1% to over 6% on reactive loads, and a smart plug accuracy study found one good brand stayed within 1.7% while four others were 54-100% off on variable loads. Run a kettle, then a small motor, through your plug and a cheap reference meter to see quickly whether yours is usable for tracking costs.
Why your plug can show 0.3 kWh while the wall meter shows 1.4 kWh
Budget plugs use a low-cost metering chip that samples voltage and current and then estimates power factor rather than doing a full four-quadrant true-power measurement. Power factor is just how much current is in step with voltage. On a pure resistive load like a kettle, current stays in phase, so estimated power matches what a lab meter would calculate.
On a switching power supply or small motor, current shifts out of phase and brings harmonics and fast spikes. That breaks the simple estimate. The chip still multiplies voltage and current, but the timing is wrong, so wattage reads low or jumpy. Two other behaviors hide the problem.
First, many apps do not show live watts. In an Tapo average power reporting thread, TP-Link support told users the displayed current power is actually an hourly average, not real-time. Second, firmware adds reporting delay. The peer-reviewed test of five brands documented delays of 3 to 6 seconds normally and up to 20 seconds on some models. So peaks get flattened and the graph never shows what you saw on the reference meter.
On AVForums, one owner logged a gaming PC for 8 hours and saw the P110 report 0.3 kWh while a second plug showed 0.427 kWh early, then 1 kWh versus 1.494 kWh by evening. As the poster put it, “half a kwh difference seems substantial”. The fix others used was to stop trusting the variable gaming load first and test with a stable kettle plus maths.
Diagram showing voltage and current sensing path in smart plug, estimation block for power factor, and comparison to lab-grade true-power calculation
At the plug, look for: does live watts update within 3 to 5 seconds when the kettle clicks on, or does it freeze then jump minutes later, which signals averaging and delay rather than a broken sensor.
How to test your smart plug against a reference meter without lab gear
Accuracy only means something when both meters see the same load at the same time for the same full cycle. That is the method used by the Home Assistant community in its reference meter comparison method, where a calibrated Kopp reference ran in series with smart meters and percent deviation was calculated from the complete kWh dataset.
You’ll need a reference like a Kill A Watt accuracy P3 P4400, which lists about ±0.2% voltage and ±1 to 2% power on steady loads, plus one resistive load around 1000W and one reactive or variable load like a small desk fan or an older PC supply. Do not exceed your plug’s rating — many are 10A continuous, 15A / 1800W peak per listings noted in a continuous load rating roundup.
Wire it wall → reference meter → smart plug → load, all on one outlet strip, no extension coils. Note start kWh on both, run a full cycle, note end kWh, time, and ambient. Use cumulative kWh, not instant watts, because watts bounce. Remember resolution matters: some wall meters only show 0.01 kWh while Home Assistant can log 0.001 kWh, so calculate deviation from the full dataset, not a single screen snapshot.
Flowchart showing reference meter test setup — wall to reference to smart plug to load, logging steps, and two load-type branches
Try this before you buy: plug a kettle through both meters, run one timed boil, and compare kWh. If the delta is more than about 3% on that pure resistive load, that plug is a poor tracker and you will see bigger drift on motors and PCs.
What the side-by-side numbers actually mean by load type
On a resistive heater, a decent budget plug should stay within roughly 3% of a reference when properly calibrated, because power factor is near 1. That is where reactive load error increase research still shows only about 1.13% error even for a true-power research meter.
On reactive or variable loads, expect more spread. The same PowerBlade paper shows error rising to about 6.5% across power factors, and the variable load error study notes highly variable 0.01 to 2 Hz loads caused substantial errors for all five brands tested. In that study, Belkin stayed ≤1.7% on energy, but four other commercial brands hit 54 to 100% error on variable loads. That explains why a fridge compressor cycling, a washer motor, or a gaming PC looks so much worse than a heater.
For bill tracking, that distinction matters. If you mostly track space heaters, kettles, and dehumidifiers, a few percent is fine. If you try to bill-split a gaming rig or an AV receiver, you may be 10% or more off, even before app averaging adds its own layer.
Table comparing plug vs reference meter readings across resistive and reactive loads with percent deviation and notes on usability for bill tracking
Comparison table — plug reading vs reference meter by load type
The table below isn’t a specific product test — it’s a worked illustrative example showing the deviation pattern this article’s mechanism explanation predicts: readings track closely on a steady resistive load, drift further on a load with a moving power factor, and drift the most on a load whose power factor swings quickly. Use it to understand the shape of the problem, then run the two-meter comparison above on your own plug and appliance to get real numbers for your setup.
Illustrative deviation pattern between smart plug and reference meter across load types — run your own comparison for exact numbers
This rubric is a practical evaluation tool created for this guide based on load-type dependence, power-factor estimation error, and reporting averaging described above, not a published industry standard. Use it as a quick check before trusting monthly totals.
Does energy monitoring actually help you track real electricity usage
It helps for spotting waste, less for acting like a utility submeter. If your goal is to find standby draw, compare two appliances, or confirm the dryer actually finished, a plug is great even with a few percent error.
The cited accuracy threshold for bill management roundup makes the line explicit: around 3% error makes data genuinely useful for bill management, while 15 to 20% off is more decorative, and it gives the example of a TV plus cable plus soundbar sitting at 47W continuous standby. At that rate, about 11 kWh a month, roughly a dollar or two depending on tariff, adds up without you noticing.
Where it falls short is monthly cost projection on variable electronics. Tapo’s hourly average display and 3 to 6 second delay means you may never see the 400W spike that actually drives cost. For a heater that runs steady, estimated monthly cost is typically within a few percent. For a gaming PC that bounces between 80W and 350W, the monthly estimate can be off by tens of percent because the averaging hides the high end.
Why wattage reading wrong usually isn’t a broken plug
If you see 29W when you expect 39W minimum on an AV receiver, that is usually not a broken sensor. In a inaccurate peak reporting discussion, a user noted a P110 double pack far from accurate, gaming 8 hours with peak not close to actual draw when game loaded, and 30W idle receiver recorded as 29W peak despite eco off and volume at 60.
Three things stack up. The app shows an averaged power, not instantaneous. Graph resolution rounds energy to 0.1 kWh, so small differences disappear. And Home Assistant users discovered that using a Riemann sum helper to integrate watts into kWh added error versus using the plug’s native energy entity. In one comparison results test, an Innr SP240 plus Riemann helper showed +12.6% deviation on a combi oven cycle, while Shelly’s native entity stayed at +0.02%.
Firmware can also fix billing. TP-Link noted a release that fixed Bill of this Month inaccurate, which matters if your totals drift month to month without hardware changes.
The mistake that makes every plug look inaccurate
Testing first with your highly variable gaming PC or AV receiver sounds reasonable because that is your real use case, but it fails because variable load triggers the worst-case error path documented in the research. Variable loads in the 0.01 to 2 Hz range cause substantial measurement errors for all units tested, even decent ones.
Start with resistive, then move to reactive. If kettle matches within a few percent, your plug and integration are fine and the larger error on PC is load-type dependent, not a hardware fault. If kettle is already 10% off, no software tweak will save the PC test.
How to get closer to the real kWh without buying lab gear
You can tighten results without lab gear. Use the plug’s native energy entity where available instead of a Riemann helper, especially for Shelly and Innr where native stayed near +0.02% in community tests versus several percent for the helper.
Update firmware and reset stats. One Tapo release note in community posts cited firmware 1.1.6 fixing Bill of this Month inaccurate — if your monthly total looks off, update, reset energy stats, then rerun a 24-hour kettle plus fridge test.
For bill tracking, log kWh over 24 hours, not instant watts, and avoid daisy-chaining plugs, which adds voltage drop and extra standby. Set a low-power threshold alert for laundry finished detection rather than trusting watts at any single second. If you have solar, a plug’s averaging is often too slow for import versus export; many Home Assistant users switched to a Zigbee bidirectional clamp meter for faster refresh, keeping plugs for spot appliance checks.
Before committing, after a firmware update reset energy stats and rerun the 24-hour kettle plus fridge test to see if month total aligns closer to the reference.
The last check
Resistive loads can be within a few percent on a decent plug, but reactive and variable loads can drift substantially because budget plugs estimate power factor and average readings over time. The single most useful step is running the same load at the same time through both your plug and a reference meter and comparing cumulative kWh for a kettle and then for a small motor or PC.
If you skip separating by load type, you’ll think every device is equally off and misjudge which ones actually drive your bill and which automation actually saves energy.
Frequently Asked Questions
Is my smart plug’s kWh reading accurate enough to estimate my electricity bill?
It depends on load type. Resistive heaters within about 3% are typically usable, while reactive variable loads can show 10 to 100% error per peer studies. Use cumulative kWh over full cycles, not instant watts, and expect accuracy by load type to shift from about 1.13% to 6.5% even on lab meters.
Why does my Tapo P110 show lower power than my dumb plug meter?
TP-Link support notes the app’s current power can be an hourly average explained rather than live, plus reporting delays of 3 to 6 seconds up to 20 seconds hide peaks. Check a 2 to 3 second updating view or native entity and compare kWh over a full boil instead of watts.
How accurate is a smart plug vs a Kill-A-Watt or calibrated reference?
A reference like Kill A Watt is about ±0.2% voltage and ±1 to 2% power on steady loads, while budget plugs estimate power factor. Calculate percent deviation from the full kWh dataset due to 0.01 vs 0.001 kWh resolution, and note native Shelly stayed +0.02% while Innr plus Riemann helper hit +12.6% in community tests per comparison results.
Should I replace my smart plugs with a clamp meter if I have solar?
If you need bidirectional import versus export tracking and sub-5 second refresh, a plug’s averaging is usually insufficient. Many Home Assistant users switched to a Zigbee clamp meter for solar, keeping plugs for spot appliance tracking where cumulative kWh is enough.
