Independent living sensor review, 1 August to 6 September 2026: what the data can and cannot show

The movement and presence sensors recorded nothing at all from 24 to 29 August 2026 (only two readings, neither of them a detection, on 24 August), and 23 August produced about half a usual day's activity events (212, against roughly 400 to 650 on other days). Over the same six days the temperature and light sensors in four to five areas continued to report several hundred readings a day, so the property's telemetry link was working, and a low-battery indication was recorded from the movable hub device on 28 August, inside the gap. Once these days are set aside and complete periods are compared like for like, detected activity averaged about 98 episodes per day from 1 to 22 August and about 92 per day from 30 August to 6 September, a difference of about -6%, with movement between monitored areas showing a similarly small difference. The supplied trend figure of about -28% compares the first third of activity days with a final third that contains the interrupted days, and the multi-scale view of the same series does not agree on a direction: the most recent windows are higher than the preceding period while the intermediate windows are lower.

The six-day absence of activity data is the most important thing in this report, for two reasons. First, it must not be read as six days without movement: no detection was recorded because the activity sensors produced no readings at all, and the available evidence cannot determine whether that was an equipment or communication problem or a period when the home was unoccupied. There is no door sensor that could help distinguish the two. The hub low-battery indication dated inside the gap is a concrete lead worth checking, though its timing alone does not establish the cause. Second, the gap distorts the headline picture: on a like-for-like comparison of complete periods before and after it, no clear sustained change in the overall level of detected activity was identified, so the reported decrease should not be relayed as a fall in the individual's activity.

The bathroom device reported around 400 to 600 readings a day in the first ten days of August, falling to under 100 a day from about 11 August, with several days from then on producing no bathroom readings whatsoever; over that same stretch of becoming intermittent it also produced device-status readings on 4, 15, 18 and 19 August. All four bathroom streams (movement, presence, light and temperature) last recorded anything on 2 September 2026, while the other three monitored areas continued to 6 September. Across the same period bathroom-area episodes fell from about 7 a day in the first week to none in the final week, and overnight bathroom-area episodes averaged 0.8 per night across the 29 analysed nights, with none on the later nights.

The fall in bathroom-area detections runs in step with the bathroom device's own loss of reporting, so it cannot be used as evidence about how the individual is using the bathroom, by day or overnight. Bathroom-area and overnight bathroom figures for the later part of this period describe what the equipment could detect, not what the person did. Because bathroom activity is often a signal assessors want to see, restoring or replacing this sensor is likely to be necessary before any future report can address it.

First detected morning activity moved later at every comparison scale: about 07:25 across the first half of the period against about 08:33 across the second half, and about 09:10 across the final week. However, checking the underlying events shows that every first morning detection earlier than about 07:30 came from the bathroom sensor, with one exception on 17 August in the hallway. Once bathroom reporting ceased, the first detection of the day is always the hallway, between about 09:00 and 09:21. Across the 30 days with a value, first detected morning activity ranged from 04:08 to 09:45. Over the same period the longest nightly stretch without any detected activity in the 22:00 to 06:00 window also lengthened at every scale, from about 305 minutes in the first fortnight to about 362 minutes later and about 384 minutes in early September.

On the face of it these two patterns look like a consistent shift towards later mornings and longer undisturbed nights, and both are arithmetically consistent across short, intermediate and broad windows. The event-level evidence shows a simpler explanation that cannot be ruled out: the bathroom sensor supplied nearly all early-morning and overnight detections, and when it stopped reporting the earliest thing the remaining sensors could see was hallway movement at around 09:00. The later timing therefore reflects the loss of that sensor as much as any change in routine. These measures cannot establish when the individual woke, got up or moved about overnight, and a period without detected activity is not evidence about sleep or rest. The wide day-to-day range, from 04:08 to 09:45, is also a reason not to read any single day's timing as a settled routine. If morning routine matters to this assessment, it would need to be revisited once bathroom sensing is restored, alongside what the individual and anyone supporting them describe.

Comparing the complete period before the data gap (1 to 22 August) with the complete period after it (30 August to 6 September), kitchen-area detections fell from about 38 to about 30 episodes a day (-22%) and kitchen-area detected time fell from about 155 to about 115 minutes a day (-26%). Kitchen sensors reported on every day of both periods, so this is not explained by a loss of sensing. It is also the only continuously monitored area with a difference of this size: overall detected activity was about 6% lower and hallway episodes about 5% lower between the same two periods, while lounge-area episodes were slightly higher and lounge-area detected time slightly lower, so no single-direction change in lounge activity was identified.

This is the one area-level difference in the period that is not accounted for by missing data, and both the number of separate detections and the total detected time point the same way. It is worth checking against other information, because kitchen-area activity is often used as an indirect signal about preparing food and drink. The data cannot establish what caused it or whether anything about eating and drinking has changed: a room episode is detected movement in the kitchen area, not a meal prepared. The comparison also rests on eight days after the gap, and no matching rise appears elsewhere, so this is better treated as a question to explore than as an established change. Asking the individual, or anyone who supports them with shopping and meals, is likely to be more informative than the sensor data alone.

Three safety-classified alarms were recorded from a worn device: two coded as fall-detected, on 8 August at 12:41 and 21 August at 09:04, and one coded as a manual SOS on 20 August at 15:49. These are the only safety-classified events in the period. The supplied quality warnings state that the meaning of this telemetry has not been confirmed and that it must not be treated as ordinary activity.

These records show that an alarm signal was generated by a worn device at those times. What actually happened on those occasions is not established by the data: they do not show that a fall occurred, that the individual pressed the button deliberately, or that help was needed or provided, and device codes of this kind can also be raised by knocks, drops or testing. Because they are the only events of their type, and two of them fall on consecutive days on 20 and 21 August, they are worth checking against the alarm service's own call records and any care notes for those dates rather than being counted as incidents on the strength of the telemetry alone.

Activity sensing exists only for four areas: bathroom, hallway, kitchen and lounge. There is no door or exit sensor, so the property-absence evidence is unavailable and the routine check on exits per day was skipped. One appliance-classified stream is present but holds only three readings and no usage sessions were derived from it. Continuous sensor availability was not confirmed for any of the 38 overnight windows that the assessment intersects.

Three practical questions cannot be answered from this data. Without a door or exit sensor, the data cannot describe leaving or returning to the home, and hallway movement must not be used as a substitute. Appliance use cannot be assessed, and the very small number of appliance readings must not be read as no appliance use. Nothing can be said about areas with no sensor, such as a bedroom. If any of these matters to this assessment, it would need either different equipment or direct enquiry.