Radon Detector Accuracy: What the Numbers Actually Mean

As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.

Why a radon detector cannot be accurate in the first ten minutes

Radon arrives in a room one atom at a time. The detector counts decays inside its own chamber, and the count rate is proportional to concentration. Counting statistics then govern the uncertainty: if you count N events, the one-sigma uncertainty is the square root of N. To get a 10 percent uncertainty you need about 100 counts; to get 1 percent you need about 10,000. A domestic detector with a small chamber simply cannot accumulate that many counts in ten minutes at typical indoor concentrations. Any instrument that displays a stable-looking figure after ten minutes is either displaying a rolling estimate with wide uncertainty or smoothing the display so it does not flicker.

This is why the published response times differ so much. The X-Sense XR0B-SR listing states results in 24 hours and accuracy reaching 95 percent after seven days or more. The GZAIR detector states a first reading in 10 minutes and reliable data after 1 hour. The AKSTEST RM-60 states a first reading after 12 hours and updates every 60 minutes. These are not contradictory claims about the same quantity; they are different statements about how long the instrument needs before its internal average stabilises. The shorter the claim, the more it is a statement about display behaviour rather than about counting statistics.

For a homeowner, the practical consequence is simple. A reading taken over an hour tells you whether radon is present and roughly at what order of magnitude. It does not tell you your annual average. Radon varies with weather, season, ventilation and pressure, often by a factor of two or more across a year. The long-term average is the number that maps onto risk guidance, and no consumer detector can shortcut the time that average requires.

Sensor type sets the accuracy floor before any percentage is quoted

Consumer radon detectors use one of two sensing approaches, and the listing usually tells you which. Ionisation chambers, as named in the GZAIR listing, collect charged particles produced by alpha decays and measure the resulting current or pulse rate. Semiconductor detectors, as named in the AKSTEST listing, use a photodiode or similar junction to count alpha particles directly. Both are legitimate. Both have a floor on what they can resolve, set by chamber volume, shielding from external radon progeny, and the electronics’ noise floor.

The GZAIR listing publishes a sensitivity figure: 30 counts per hour per pCi/L, described as 15 times higher than the minimum professional standard. That is a useful number because it lets you do the arithmetic yourself. At 4 pCi/L, a common action level, 30 cph/pCi/L gives 120 counts per hour, or about 2 counts per minute. Over one hour the one-sigma uncertainty is the square root of 120, about 11 counts, which is roughly 9 percent of the signal. Over a day it falls to about 2 percent. That is the mechanism behind the claim that longer averaging gives better accuracy, and it is why the same instrument can quote a short response time and a much better long-term figure without contradiction.

Where a listing quotes an accuracy percentage without saying what it is relative to, treat it as a claim about the instrument’s agreement with a reference under unspecified conditions. The X-Sense figure of 95 percent after seven days is the clearest example: it is a statement about a seven-day average, not about a single reading. A seven-day average is still short compared with the seasonal variation described above, so 95 percent agreement with a reference over that window does not mean 95 percent agreement with your annual average. No listing here publishes an annual-average accuracy figure, and you should not infer one.

What the published figures let you compare, and what they do not

The table below collects only what the listings state. Where a listing is silent, the cell says so. This is deliberately unflattering: it shows how little of the information a buyer might want is actually published.

Two things stand out. First, only one listing publishes a sensitivity figure, and only one publishes a detection range. Without those, you cannot compute the counting uncertainty yourself, which is the only accuracy statement that is not a marketing claim. Second, the accuracy claims are not commensurable. One is tied to seven days, one to one hour, and two are absent. A comparison table of accuracy percentages across these four products would be misleading, which is why this one does not contain such a column.

The Safety Siren Pro4 is a useful counterexample. Its listing states continuous monitoring, short-term and long-term averages, and audible and visual alarms, but publishes no accuracy figure, no sensor type and no response time. That is not a criticism of the instrument; it is a statement that the listing gives you nothing to compare numerically. If a manufacturer does not publish a figure, the honest reading is that you do not know it, not that it is average.

The accuracy question buyers actually have is about the alarm threshold

Most people asking about radon detector accuracy are really asking a different question: will this thing tell me if my home is dangerous? That is a question about the alarm threshold and its relationship to the guidance level, not about the detector’s agreement with a laboratory reference. The action level used in guidance is 4 pCi/L as a long-term average. A detector that alarms at 4 pCi/L on a single short-term reading will alarm frequently in ordinary homes during weather that raises radon, and will stay silent in homes whose annual average is above the level but whose current reading is below it.

The listings here describe adjustable or selectable alarm behaviour without publishing the default threshold. The X-Sense detector states sound-and-light warnings when radon exceeds the threshold. The Safety Siren Pro4 states three selectable audible alarm modes plus a visual alert. The AKSTEST RM-60 states that the alarm threshold is adjustable. None of the four listings publishes the factory default threshold or the hysteresis around it. That is a real gap for a buyer, because an alarm that triggers on a one-hour spike is not the same product as an alarm that triggers on a seven-day average, even if both are described as radon alarms.

The practical advice is to use the alarm as a prompt to look at the long-term average, not as a verdict. If the alarm sounds, read the long-term figure the instrument displays. If the long-term figure is below the guidance level and the alarm is a short-term spike, ventilate and carry on. If the long-term figure is at or above it, the next step is a confirmatory measurement, not another consumer detector.

Calibration, maintenance and the drift nobody mentions

An accuracy figure is only meaningful if the instrument is calibrated and stays calibrated. The GZAIR listing states that all sensors have been calibrated and describes the calibration as lab free, which in context means the user does not send the unit away for calibration. That is convenient, but it is not the same as a traceable calibration against a reference standard at a stated interval. None of the four listings publishes a calibration interval, a reference standard, or a drift specification. Not stated in the listing bullets, across all four.

Radon detectors also have a maintenance item that listings rarely mention: the chamber must stay free of dust and the intake must stay unobstructed. A detector placed on a shelf above a radiator, or in a room where candles are burned, will collect particles that affect the count. The vented housing named in the AKSTEST listing is there for a reason, and it needs to be kept clear. Battery state matters too: the X-Sense listing claims a 2-year battery life and low-power notifications, and the AKSTEST is rechargeable. A detector that has been sitting unpowered for a season has no useful history, whatever its accuracy specification says.

The honest position is that consumer radon detectors are not calibrated instruments in the metrological sense. They are indicators with a stated or implied accuracy that is good enough to tell you whether to take the next step. If you need a number that will stand up in a property transaction or a mitigation decision, the next step is a passive test kit analysed by a laboratory, not a more expensive consumer detector.

Where detector accuracy stops being the right question

There are three situations where the accuracy of a consumer detector is beside the point. The first is a real-estate transaction or a regulatory requirement: what is needed is a measurement made to a defined protocol with a defined device, and a consumer detector’s percentage claim does not satisfy that. The second is post-mitigation verification: after a mitigation system is installed, you need to confirm that the annual average has fallen below the guidance level, and that requires a measurement over a season, not a spot reading from a detector whose accuracy is quoted over seven days. The third is a home with a known elevated level: at that point the question is not which detector is more accurate but which contractor to call.

This is the same boundary that applies across environmental monitoring. A handheld instrument tells you whether to look closer; it does not replace the reference method. If you are choosing between detectors, the useful comparison is sensor type, published sensitivity, averaging period and whether the listing publishes a threshold, not the headline accuracy percentage. If you are choosing between a detector and a test kit, the detector wins on convenience and the test kit wins on defensibility. For a first look at a home, a detector is the right instrument. For a decision, it is the start of a process, not the end of one.

The short version

Radon detector accuracy figures are not comparable across listings because they are tied to different averaging periods and different definitions. Read the averaging period first, then the sensor type and any published sensitivity, then the alarm threshold. If a listing publishes none of those, you do not know the accuracy, and you should not assume it. And if you need a number that will be acted on formally, no consumer detector is the right instrument.

Frequently Asked Questions