What Omaya works out on its own

Much of what Omaya shows is not typed in by anyone. It is a judgement Omaya makes from the data: which room a tag is in, where a person lives, whether a gateway is about to fail, whether a booked room is empty. This page lists those judgements in one place, for building administrators, privacy officers and anyone asking "how does it know?".

Each entry says what Omaya decides, from what data and with which numbers, who or what it applies to, how far you can change it, where it can be wrong, and where the full explanation is. Where an entry says Cannot be changed, the numbers are fixed for every organisation.

Times are in your organisation's time zone (System → Configuration → General → Timezone) unless an entry says otherwise. Where a judgement only counts opening hours, those are the Opening hours card on the same page, or a venue's own week.

Where people and things are (indoors)

When a tag has really moved to another room

• Decides: whether a louder gateway in another room has really taken a tag, or the change is a reflection, before the tag is moved there.
• From: what each gateway heard at every sweep, smoothed. A challenger must be louder by 8 dB to take a tag and the current room must fall 4 dB behind to lose it, two sweeps in a row, with 6 dB more for each floor of separation. A tag whose accelerometer says it has lain still for 60 s keeps its room unless a gateway is 10 dB louder.
• Applies to: every tag and sensor heard over Bluetooth, people and assets alike. The stillness hold only for tags that report movement.
• Change it: System → Configuration → Indoor Positioning → Speed: Sweeps to win in a row, Louder by, to take a tag, Louder by, to leave a room, Extra to change floor, Settling hold at rest. Stability & records → Bermuda-style zone stability replaces the fixed dB margins with a percentage difference.
• Can be wrong when: a tag sits near a wall between two gateways and the margins are low (below about 4 dB tags trade back and forth); a tag lying still in the wrong room stays there until it moves.
• More: Speed settings · The tag flicks between rooms

How long a move takes to show

• Decides: how much evidence and time a move needs before it is written to the trail, and so the shortest visit that can be recorded at all.
• From: sweeps to win multiplied by the sweep interval, plus Hold before writing. Presets: Stable 2 sweeps and 60 s hold, Balanced 2 sweeps and 30 s, Responsive 1 sweep and 15 s. With no preset chosen: 2 sweeps and 30 s.
• Applies to: everything with a trail, for the whole organisation.
• Change it: System → Configuration → Indoor Positioning → Speed: the presets and Hold before writing (0–600 s). An empty field returns to the installation default.
• Can be wrong when: a visit is shorter than the time shown at the top of the Speed tab: it is not recorded. Faster presets record more room changes on tags with weak coverage.
• More: Speed settings

Rooms that have no gateway of their own

• Decides: the room a tag is in from what every gateway on the floor heard and did not hear, so a room without a gateway can still be named.
• From: the drawn rooms cut into half-metre cells, each gateway's signal model, and the previous room (a tag tends to stay, or move to a neighbouring room). Optional: a triangulated position as extra evidence; holding the room while the tag has not moved for 5 minutes.
• Applies to: floors with rooms drawn and at least one gateway placed. Off until switched on.
• Change it: System → Configuration → Indoor Positioning → Rooms → Room estimation, with Also use triangulation as evidence, …but not beside a gateway on a room corner and Hold the room while a tag lies still.
• Can be wrong when: a floor has only three gateways, so one bad reading cannot be singled out; a gateway sits on a corner shared by several rooms; a held tag was put down in the wrong room.
• More: Rooms settings · A desk tag in the wrong room

Which floor a tag is on

• Decides: the floor before the room, so signal leaking through a slab does not move a tag upstairs.
• From: by default, weak readings from another floor are ignored: heard by one gateway only and weaker than −80 dBm. Optionally each gateway votes for its floor (full vote at −65 dBm, half at −75, ignored below −80, winning margin 0.3), or the floor is weighed over time.
• Applies to: sites with more than one floor, where each venue has a floor level set.
• Change it: System → Configuration → Indoor Positioning → Floors: Ignore faint readings from another floor (on as shipped), Pick the floor by majority vote (off), Weigh the floor over time (off) with Change floor sooner while a tag is moving.
• Can be wrong when: a venue has no floor level, so it gets no cross-floor protection; weighing over time can miss a visit of a minute or two to another floor.
• More: Floors settings · Wrong floor

Which positioning method a floor uses

• Decides: whether a floor positions by radio map, by triangulation, from fixed sensors only, or by nearest gateway.
• From: what the floor has: a radio map from a calibration walk gives fingerprint; three or more gateways with positions give triangulation; otherwise nearest gateway.
• Applies to: each floor (venue) separately.
• Change it: the Positioning method card on the floor's Calibrate gateways page: Automatic (default), Presence, Triangulation, Fingerprint.
• Can be wrong when: Fingerprint is chosen on a floor with no walk: it never produces a position.
• More: Which method a venue actually uses

Gateway positions and signal fade, from one walk

• Decides: where each gateway really is, how fast signal fades on that floor, how loud each gateway is at one metre, and the radio map.
• From: the readings collected while a tag is carried on a calibration walk. The fade is fitted between 1.5 and 4.5, and a new model is kept only if it beats the one declared by a clear margin.
• Applies to: the walked floor.
• Change it: walk again. Pressing Finish walk does the rest.
• Can be wrong when: gateways were moved after the walk; tags of another model are used without a tag power offset.
• More: What happens when you press Finish walk

Tags and gateways that are louder than others

• Decides: how much louder or quieter each tag model transmits than the tag the walk was done with, and how much louder each gateway hears than the others, and removes the difference before comparing.
• From: the Calibration Bench and the walk. A tag offset above 25 dB, or a gateway offset above 20 dB, is treated as implausible.
• Applies to: tag offsets to every tag of that model; gateway offsets only while the switch below is on.
• Change it: Device → Tag Power; System → Configuration → Indoor Positioning → Rooms → Allow for gateways that hear louder than others (off as shipped) and each gateway's offset on the floor plan.
• Can be wrong when: no offsets are set: on a typical floor a 6 dB difference makes a distance look 1.7 times longer.
• More: The Tag Power page · One model is always wrong · One gateway hears too loudly

Repeated gateway messages

• Decides: that a message a gateway sends again, unchanged, is not fresh evidence of where the tag is.
• From: the same gateway, tag, content and signal within 5 seconds. Some gateways repeat the last message for about two minutes after losing a tag.
• Applies to: positioning only. The message is still stored and the tag still counts as heard.
• Change it: Cannot be changed.
• Can be wrong when: a tag lingers in the room it left for up to two minutes, because the repeats still say it was heard.
• More: Dusun gateways: good to know

When a stay starts and ends

• Decides: when a stay in a room starts and ends, and that a quick bounce out and back is the same stay.
• From: a stay starts at arrival and ends when the tag has not been heard for 5 minutes. Returning to one of the last few rooms within 5 minutes reopens the earlier stay instead of starting a new one. Long stays are split at midnight.
• Applies to: every device with a trail; entity Timelines follow the devices bound to the entity.
• Change it: System → Configuration → Indoor Positioning → Stability & records → Split long stays at midnight (on as shipped). The 5-minute rules cannot be changed.
• Can be wrong when: a tag goes unheard for more than 5 minutes inside a room: the Timeline shows two stays with a gap.
• More: Read the Timeline · There is a gap in the Timeline

Online, stale or offline: people and assets

• Decides: whether an entity is currently heard.
• From: the time since any of its tags was last heard. On Monitor → Entity Tracker: online up to 2 minutes, stale up to 15, offline after that or with no tag bound; someone outside on GPS stays online for 15 minutes between fixes. The view opened from the Entity page in the menu uses 5 minutes for online and 15 for stale, so it can say online while the Entity Tracker already says stale.
• Applies to: every entity with a tag bound.
• Change it: Cannot be changed.
• Can be wrong when: a tag reports less often than every two minutes, for example a tag in power-saving mode: it shows stale on the Tracker while present.
• More: Online, stale, offline

What the positioning advisor recommends

• Decides: for each positioning switch, whether this site should have it on or off, and warns when a switch is on without what it needs.
• From: your floors and the last 24 hours: floor vote when more than 5 % of decisions are contested between floors; room estimation when drawn rooms have no gateway; Bermuda when devices change floor 4 or more times a day; warnings for fingerprinting without a walk, floors without a floor level, or fewer than three placed gateways.
• Applies to: the whole organisation.
• Change it: nothing to set. Apply and Try on/off on the last day act on its advice.
• Can be wrong when: no tags were seen in the last 24 hours: it has nothing to judge.
• More: Overview settings

Whether a floor's tags are steady

• Decides: on a floor plan's Health tab, which rooms can be named, how often positions rest on only one or two gateways, and how often tags flap.
• From: a flap is leaving a room for under 2 minutes and coming straight back; a longer absence is a round trip.
• Applies to: each floor.
• Change it: Cannot be changed.
• Can be wrong when: a genuine quick errand into the next room counts as a flap.
• More: The floor's Health tab

Outdoors and GPS

Where a person's home is

• Decides: where a person spends the night, so that place can be hidden.
• From: the person's GPS fixes from 00:00 to 05:00 over the last 30 days, grouped within 200 m. The place seen on the most nights is home if it was seen on 3 or more nights.
• Applies to: people only. Assets and vehicles never have a home.
• Change it: System → Configuration → Outdoor & GPS → Privacy → Homes on maps and replays: Hidden, "At home" (default) or Shown. The 30 days, the hours, the 200 m and the 3 nights cannot be changed.
• Can be wrong when: someone works nights at the same place every night (that place is found as home); two homes within 200 m merge; fewer than 3 nights of fixes means no home is found and it is drawn like any other stop.
• More: How Omaya knows someone is at home

Likely homes among frequent stops

• Decides: which unnamed places trackers keep stopping at, and which of those is probably someone's home.
• From: stops outside every geofence over the last 30 days, grouped by place. A place is listed after 3 stays; it is a likely home when at least half its stays run through 03:00.
• Applies to: every GPS tracker, people, assets and vehicles together.
• Change it: System → Configuration → Outdoor & GPS → Privacy → Likely homes in frequent stops: Hidden (default) or Shown, flagged.
• Can be wrong when: a depot open all night, or a night-shift site, is flagged as a home.
• More: Outdoor settings

Trips, stops and geofence visits

• Decides: turns GPS fixes into rows on the trail: a visit inside a geofence, a trip while moving, a stop once still.
• From: a stop is 5 minutes within 50 m; a new row starts after 30 minutes with no fix if the tracker moved meanwhile; a fix rated worse than 150 m does not move the trail.
• Applies to: GPS trackers and tags heard by Omaya Scout phones.
• Change it: System → Configuration → Outdoor & GPS: Stop after, Stop radius, New trip after silence, Ignore fixes worse than.
• Can be wrong when: the stop radius is set below the 5–20 m that GPS wanders on its own, which splits one stop into several.
• More: Outdoors: visits, trips and stops

Indoors or outdoors

• Decides: whether the room or GPS owns a tracker's trail at a given moment.
• From: for 120 seconds after a gateway last heard the tracker, indoors wins: GPS fixes go on the map but not the trail. After that GPS takes over and the indoor stay is ended at the last time a gateway heard it.
• Applies to: trackers with GPS that are also heard by gateways.
• Change it: System → Configuration → Outdoor & GPS → Indoor wins for (30–900 s).
• Can be wrong when: expect a gap of one to two minutes between the last room and the first outdoor row.
• More: The handover between indoors and outdoors

Outside right now

• Decides: for each tracker outside, whether it is on a trip, stopped, in a geofence, at home, quiet, or back inside.
• From: quiet is nothing heard for 15 minutes while outside; back is heard by a gateway since, within the hour.
• Applies to: trackers with a GPS fix in the last 30 days.
• Change it: Cannot be changed, apart from New trip after silence, which sets how far back it looks.
• Can be wrong when: a tracker reports less often than every 15 minutes and so reads quiet between fixes.
• More: Outside

Where a tag heard by a phone is

• Decides: places a tag at the position of the Omaya Scout phone that heard it best, with an accuracy figure.
• From: the phone's position and accuracy (25 m when it gives none) plus a distance estimated from the signal, capped at 60 m. Per minute, the most accurate fix wins.
• Applies to: tags heard by phones running Omaya Scout.
• Change it: Cannot be changed.
• Can be wrong when: the distance from signal is a generic estimate, so the tag can be up to the stated accuracy away from the phone.
• More: Phones as trackers: Omaya Scout

Working outside in Attendance

• Decides: counts a field-staff person's GPS time outside, within working hours, as working outside, except time stopped at home.
• From: the site's opening hours, else 08:00–18:00 Monday to Friday; home as described above.
• Applies to: people carrying the label chosen as field staff.
• Change it: on Reports → Attendance, the field-staff label (Which label marks field staff?).
• Can be wrong when: home is not known yet (fewer than 3 nights), so time at home counts as working outside.
• More: Attendance in the report glossary

Counting people and space

Staff, visitors and unassigned tags

• Decides: which tagged people count as staff and which as visitors.
• From: a person carrying the chosen staff label is staff; any other person is a visitor; a tag bound to nobody is unassigned. Who held a tag on a past day comes from the binding history; the label is read as it is today.
• Applies to: Footfall, Monitor → Outside and AI Reporting.
• Change it: on Reports → Footfall, Choose the label that marks staff. With none chosen there is no split.
• Can be wrong when: a person's label changed: rebuilt past days use today's label.
• More: Staff, visitors and unassigned tags

The most inside at once, from door counters

• Decides: how many people were inside at the peak, from entries and exits.
• From: entries minus exits, hour by hour, reset at local midnight and never below zero. A counter that counted only one direction in the period is left out and named.
• Applies to: places with door counters.
• Change it: Cannot be changed.
• Can be wrong when: a door is not counted: the figure is wrong in one direction and stays wrong.
• More: Door counts, and the most inside at once

Days with missing or partial data

• Decides: that a day with no data is missing, not empty, and that a day with data for fewer than half its expected hours is partial.
• From: the gateways' hourly reports, including fixed sensors, so an empty building still counts as reporting.
• Applies to: Footfall and the reports that share it.
• Change it: the Opening hours card changes which days and hours are expected.
• Can be wrong when: gateways reported but the tags or counters behind them did not: the day looks complete.
• More: Missing days and older days

How many people are in a zone right now

• Decides: one "people right now" figure per zone, from whichever sensor actually counts people there, and whether the zone is empty, normal, busy or over.
• From: in order: the room's calibrated radar; counting sensors that reported in the last 10 minutes; the number of tags. Door counters are not a source for a zone. Busy is 50 % of capacity or more; over is capacity or more.
• Applies to: every zone. Without a capacity only empty or normal is possible.
• Change it: the zone's capacity (Space → Zone). The order, the 10 minutes and the 50 % cannot be changed.
• Can be wrong when: the source is tags: the figure counts only people carrying one.
• More: The four sources · Capacity and what busy means

A venue's occupancy from its doors

• Decides: whether a venue's entries minus exits may be shown as its occupancy.
• From: only when you have said every way in and out is counted, exits do not run ahead of entries by more than 10 % plus 5, and during opening hours the counters reported in the last 15 minutes. The day starts at opening time, or midnight without hours.
• Applies to: venues with door counters.
• Change it: Space → Venue → Every way in and out of this venue is counted.
• Can be wrong when: one door is missed: the figure drifts one way all day.
• More: When the doors are all counted

A venue's capacity when none is set

• Decides: how many people a venue is meant to hold.
• From: the venue's own capacity; if empty, the sum of its active zones' capacities; if neither is set, unknown (never 0).
• Applies to: every venue.
• Change it: the capacity on the venue form, or on its zones.
• Can be wrong when: some zones have no capacity: the sum undercounts.
• More: Venues

Occupancy forecast for the next 24 hours

• Decides: the expected people count per zone, hour by hour, for the next day.
• From: 30 days of hourly counts from the trail, plus holidays. Refreshed hourly.
• Applies to: zones with at least a day of history.
• Change it: Cannot be changed.
• Can be wrong when: many people carry no tag: the forecast follows tagged people only.
• More: Right Now

Whether a room is underused, used or busy

• Decides: a verdict per zone on Space Utilisation.
• From: an hour is occupied when its peak is one person or more, counted over opening hours that have passed. Under 25 % of hours occupied is underused, 70 % or more is busy. Counter coverage below 80 % is flagged as the sensor being down, not the room being underused.
• Applies to: every zone with readings.
• Change it: opening hours and zone capacity. The percentages cannot be changed.
• Can be wrong when: a counting sensor was off for part of the period and its coverage stayed above 80 %.
• More: Space Utilisation

Buildings scorecard

• Decides: a class per building (Opportunity, Alert, Resilient or not scored) and which floors could be handed back.
• From: floors with the smallest typical need are fitted into the spare seats of the others, with a 10 % margin. A floor is scored after 20 measured opening hours. Below 25 % use is an alert; 25 % or more seats spare is an opportunity.
• Applies to: buildings with floor areas set.
• Change it: floor area and rent on the venue, Area Unit under General, and opening hours. The percentages cannot be changed.
• Can be wrong when: occupancy on a floor comes from tags only, so its need is underestimated.
• More: Space Utilisation

How much of the data arrived

• Decides: grades the data behind Space Utilisation and Attendance as good, fair or poor, and names the worst devices.
• From: each fixed sensor is expected once per its own silent threshold (at least an hour) during opening hours on open days; gateways by their availability. Good is 95 % or more, fair 80 % or more. Tags and event-driven devices such as door counters are not penalised for quiet periods.
• Applies to: the period and places on screen.
• Change it: each device's Silent threshold (minutes) and the opening hours.
• Can be wrong when: a sensor's silent threshold is much longer than its real reporting interval, so gaps are not seen.
• More: Data health on Space Utilisation and Attendance

Attendance: visits, here now, late and left early

• Decides: when each person was on site, how visits split, and who arrived late or left early.
• From: gaps up to 30 minutes are bridged into one visit (a GPS stretch always splits it); last seen under 5 minutes ago is here now; late is first in more than 15 minutes after opening, left early is last out more than 15 minutes before closing. A stay of zero seconds is dropped.
• Applies to: people with a tag bound.
• Change it: on Reports → Attendance, Gap that ends a visit and Late / early after (or Don't mark); opening hours.
• Can be wrong when: a person forgets their tag: they are absent, not late.
• More: Who was in, and for how long

Office days by team

• Decides: average office days a week per team, with the anchor day and the quiet day.
• From: a team is an entity label (a person with two labels counts in both). Only people with a tag bound in the period are measured; others are not measurable, never absent. Only open days count.
• Applies to: people.
• Change it: the policy picked on the page, and the operating week.
• Can be wrong when: a person did not carry their tag on a day in the office.
• More: How many office days each team really comes in

Where in a room people spend time

• Decides: the hotspots in a room and how long it was occupied, on the Room Usage Map.
• From: calibrated position radars only, in half-metre cells; a spot needs 30 readings. A radar awake for under 80 % of the hours judged gives a lower bound, not a verdict.
• Applies to: rooms with a calibrated position radar.
• Change it: Cannot be changed.
• Can be wrong when: the radar sees through glass or reflects off a screen.
• More: Room Usage Map

Devices and gateways health

Gateway online, stale or offline, and the down alert

• Decides: the status of each gateway, and tells admins once when one goes down and once when it comes back.
• From: the time since its last packet, checked every minute. Bluetooth: stale after 5 minutes, offline after 15. LoRaWAN: 30 and 60 minutes. Stale counts as up, so nobody is told. Inside a gateway's quiet hours nothing is announced; if it is still down when they end, the alert goes then.
• Applies to: every active gateway.
• Change it: System → Configuration → Gateway timing: Stale after and Offline after (Bluetooth only); per gateway, Expected silence (quiet hours).
• Can be wrong when: a gateway's connection drops for a few minutes at a time: raise Offline after so short drops do not page anyone.
• More: Being told when a gateway drops

When a gateway outage started and ended

• Decides: the real start and end of each outage on Gateway Availability.
• From: the start is the gateway's last packet, not the moment it was noticed (about 15 minutes later); the end is the first packet after.
• Applies to: every gateway, from 6 October 2026; earlier time is filled from hourly records.
• Change it: Cannot be changed. Quiet hours are left out of availability.
• Can be wrong when: an outage began and ended while Omaya's own processing was stopped: it is not seen. A recovery read late can overstate an outage, never understate it.
• More: Gateway Availability

Too few gateways for positioning in a venue

• Decides: that a venue no longer has enough gateways online to triangulate, so its devices fall back to nearest gateway.
• From: fewer than 3 active gateways online in a venue that has at least 3. It must last 3 minutes before it is announced, and 5 minutes back to normal before that is announced.
• Applies to: venues with 3 or more gateways.
• Change it: Cannot be changed.
• Can be wrong when: a gateway is stale (quiet for 5 to 15 minutes): it still counts as online here.
• More: The venue events

A gateway likely to fail soon

• Decides: nightly, whether each gateway is healthy, weak or fading, or likely to fail within 30 days.
• From: a straight line through 30 days of its daily average signal. A 30-day average below −75 dBm, or a fall of 0.1 dBm a day or more, is a warning. Likely to fail needs the line to reach −85 dBm within 30 days, a good fit, at least 30 days since installation and 7 days of data. Under 3 days of data: unknown.
• Applies to: active gateways.
• Change it: not by your organisation; your Omaya provider can adjust the lines for it.
• Can be wrong when: a gateway was recently installed or moved and is still settling.
• More: Live Monitor

A device that has gone silent

• Decides: that a device has stopped reporting.
• From: nothing heard for the device's silent threshold: its own, or 3 days for tags and asset tags, or 2 hours for other devices. Critical at three times that. Heard again, it closes itself as recovered. Checked hourly.
• Applies to: devices with Anomaly detection on.
• Change it: Device → Sensor, edit the device: Anomaly detection and Silent threshold (minutes). Leave it blank for the defaults above.
• Can be wrong when: a device is meant to report rarely, such as an on-demand button: switch Anomaly detection off for it.
• More: Anomalies · What counts as down

A device heard for fewer hours than usual

• Decides: that a fixed device is reporting far less than it normally does.
• From: hours heard per day over the last 30 days (at least 3 days needed). It is flagged when the last day is 2 standard deviations below normal and at least 4 hours below the average; a warning when half the average or less. Days run in UTC.
• Applies to: fixed devices with Anomaly detection on; never tags.
• Change it: Anomaly detection per device. The numbers cannot be changed.
• Can be wrong when: a site was closed for a holiday: the quiet day is flagged.
• More: Anomalies

A battery draining faster than it should

• Decides: that a device's battery is falling faster than a healthy cell.
• From: the 14-day trend behind the Battery report: 5 % a day or more is flagged, 15 % a day or more is a warning. Needs a battery reading under 3 days old.
• Applies to: devices with Anomaly detection on that report a battery level.
• Change it: Cannot be changed, apart from Anomaly detection.
• Can be wrong when: a tag's level jumps around because of cold or heavy use.
• More: Anomalies

Battery level from a voltage

• Decides: a percentage for tags that report only their battery voltage.
• From: a straight line from empty to full per model: MOKO M2 and most MOKO BeaconX tags 2.0 V to 3.0 V, MOKO M3 Mini 3.0 V to 3.6 V. A model with no known battery type shows the voltage and no percentage. Other models report their own percentage.
• Applies to: those models.
• Change it: Cannot be changed.
• Can be wrong when: a cell holds its voltage until nearly empty, so the percentage falls quickly at the end.
• More: Battery

Low battery by voltage

• Decides: that a tag's battery is low from its voltage, before the percentage says so.
• From: checked nightly against the lowest voltage the tag reported in the last day. Below the line set for its model is a low alert that names the voltage; a work order raised from it quotes the voltage too ("2,540 mV, below 2,600 mV"). The alert closes itself once the tag reads 100 mV above the line.
• Applies to: tags that report their voltage, of a model with a line set.
• Change it: the line per model, under System → Configuration → IoT → When a tag's battery is low, by voltage. A model left empty is judged by percentage only.
• Can be wrong when: a cold room or a burst of radio work dips a cell for a moment: the lowest reading of the day counts.
• More: Battery

Low battery, and days left to replace

• Decides: when to raise a low or critical battery alert, and roughly how many days a battery has left.
• From: checked nightly. Below 20 % is low, below 10 % critical; the alert closes itself at 25 %. The days left come from the daily lowest level over 14 days (at least 5 days), projected to 20 %; admins are told when it is 14 days or fewer. Days run in UTC.
• Applies to: every device that reports a battery level.
• Change it: Cannot be changed.
• Can be wrong when: a device's last battery reading is days older than its last packet.
• More: Battery

A device whose data is being dropped

• Decides: that a device is sending readings that cannot be kept because it has no venue or zone, and its gateway has none either.
• From: the readings as they arrive. One open warning per device; it clears when the device is placed.
• Applies to: fixed devices, for example those bridged from Home Assistant.
• Change it: place the device in a venue or zone.
• Can be wrong when: the device's gateway has a venue set: its readings are kept under that venue, which may not be where the device is, and no warning is raised.
• More: Anomalies

Door counters that do not add up

• Decides: daily, whether a door-counted room is counted correctly.
• From: a counter with no inward side set; two counters on the same pair of rooms; a busy day with far more people in than out; door counts and radar disagreeing.
• Applies to: rooms and venues counted at their doors.
• Change it: fix the counters; the checks cannot be changed.
• Can be wrong when: a day with an event that ends after midnight looks unbalanced.
• More: What the audit checks

Radar reflections

• Decides: that a spot a radar keeps reporting is glass, a screen or a cabinet, not a person, and that one radar frame heard by several gateways is one frame.
• From: over 24 hours, a half-metre cell hit in 3 or more different hours, 30 or more times, without moving. A new reflection inside the counting area is announced once. Nothing is removed automatically.
• Applies to: position radars.
• Change it: Cannot be changed. Adjust the radar's counting area if the reflection is inside it.
• Can be wrong when: furniture is moved: the old spot stays listed until it has not been hit for 14 days.
• More: Central Alerts

Safety: falls, buttons and roll call

A fall, or a person who has stopped moving

• Decides: that the wearer of a tag has fallen (a sharp change in which way the tag faces, then stillness) or has stopped moving, with no calibration per person.
• From: the tag's accelerometer. At Normal sensitivity: a swing of 45° or more, then 60 s still. Stopped moving: active, then still for 15 minutes. A position the person rests in at least 3 times is learned and does not alert. In a rest place it can wait (10 minutes as shipped) before asking, or only record it.
• Applies to: people wearing a supported tag, covered by a fall rule or with Watch every supported tag on.
• Change it: System → Configuration → Fall detection: Fall detection, Sensitivity, Alert people, Time to answer, Rest places. A toilet can never be a rest place.
• Can be wrong when: the tag is worn on the wrist (about 1 in 6 alerts false); someone lies down on purpose.
• More: How it decides · How accurate is it

Is this a real button press

• Decides: whether a message from a button is a new press, and whether single, double or long.
• From: Minew tags: only the first press message after 90 seconds of silence counts; a tag sending its press message for 2 minutes without stopping is treated as idle and never as a press; the same press is not counted twice within 6 minutes. MOKO buttons: each press carries its own count. Damaged messages are refused.
• Applies to: tags with a button.
• Change it: Devices → Panic Buttons, and the press slot under device configuration.
• Can be wrong when: a Minew tag is left on its factory settings, so its idle message looks like a press.
• More: What happens after a press

Panic button or just a button

• Decides: whether a press raises a critical panic incident or a warning.
• From: only a button registered under Devices → Panic Buttons raises Panic button (critical); any other button in a watched zone raises Button pressed. A movement alarm is never a press; an inactivity alarm counts only on a registered panic button. Presses older than 60 minutes are ignored.
• Applies to: Security Watch and Central Alerts.
• Change it: Devices → Panic Buttons; Panic button pressed under Security Watch Detectors.
• Can be wrong when: a desk button is registered as a panic button and times out on inactivity.
• More: Panic Buttons

Safe, late or missing in a roll call

• Decides: each person's status during a roll call.
• From: their live zone. For the first 60 seconds an absent person is late, then missing; someone not updated for 5 minutes goes back to late. A badge already in the gathering zone at the start must be heard again 2 minutes later. A GPS fix newer than any gateway sighting counts as outdoors, accounted for. A warden's mark always stands.
• Applies to: badges bound to a person, or labelled staff, employee, human or visitor.
• Change it: gathering points and wardens in the app; the timings cannot be changed.
• Can be wrong when: a badge is not heard at the gathering point, for example in a car park without a gateway.
• More: How to read the board · Badge reliability

Alerts and the AI assistant

What still needs a look in Central Alerts

• Decides: which alerts are still open, and which repeat.
• From: an alert is done when acknowledged, when it is information only, or, for sources nothing closes, after 24 hours. Three or more alerts of the same kind from the same device or rule fold into one row.
• Applies to: Monitor → Central Alerts.
• Change it: Cannot be changed.
• Can be wrong when: an alert from a source nothing closes is still relevant after 24 hours: it drops out of Needs a look but stays in the list.
• More: Central Alerts

OAIA: gateways offline, alone or together

• Decides: which gateways are offline, and when several in one venue dropped together (a power or network event, not a faulty gateway).
• From: offline gateways, critical after 24 hours; together means 2 or more down within 15 minutes, or half the venue's gateways.
• Applies to: organisations with OAIA on.
• Change it: System → Configuration → AI → Heartbeat every (2 to 24 hours) sets how often it looks.
• Can be wrong when: two gateways fail on their own within 15 minutes.
• More: The heartbeat

A gateway that keeps dropping

• Decides: that a gateway drops out more often than a healthy one does, even if it is online now.
• From: its outages in the last 7 days. OAIA raises a warning at 3 or more of its own; an outage that began within 15 minutes of another gateway's, or that took every gateway down, is site-wide and does not count. The TV Dashboard's Gateways widget names a gateway from 2 drops in 7 days.
• Applies to: every active gateway; the finding needs OAIA on.
• Change it: Cannot be changed.
• Can be wrong when: two gateways each drop on their own within 15 minutes of each other: both are treated as site-wide.
• More: The heartbeat · When a gateway outage started and ended

OAIA: data no longer reaching positioning

• Decides: that a venue's positioning has gone quiet even though its gateways may look online.
• From: the room decisions in the last heartbeat window (at least 2 hours) compared with the same window a week earlier: 10 % or less is flagged. Needs 50 decisions in the earlier window.
• Applies to: organisations with OAIA on; not new venues.
• Change it: Heartbeat every.
• Can be wrong when: the venue was closed this week but open the same day last week.
• More: The heartbeat

OAIA: low battery

• Decides: which devices have a low battery.
• From: a reading from the last 7 days at 20 % or less (warning) or 10 % or less (critical).
• Applies to: organisations with OAIA on.
• Change it: Cannot be changed.
• Can be wrong when: a device stopped reporting more than 7 days ago: it is a silence, not a battery finding.
• More: The heartbeat

OAIA: temperature off its normal

• Decides: that a zone's latest temperature is far from its own normal for that hour of day.
• From: 14 days of readings for the same hour (at least 7): 3 standard deviations and at least 2 °C away; critical at 5 °C. Hours are UTC hours, compared like for like.
• Applies to: zones with temperature readings, with OAIA on.
• Change it: Cannot be changed.
• Can be wrong when: the air conditioning schedule changed this week.
• More: The heartbeat

OAIA: something left on after hours

• Decides: that a meter is still drawing as if the venue were open.
• From: at least an hour after closing, a meter drawing 70 % or more of its own open-hours average and at least 200 W. Skipped if the room shows people in the last hour.
• Applies to: organisations with Energy and OAIA.
• Change it: opening hours.
• Can be wrong when: the room has no occupancy sensing, so a late worker cannot be seen (the finding says so).
• More: The heartbeat

OAIA: outdoor findings

• Decides: a long stop away from your sites compared with the person's own stops, a tag outside that has gone quiet for longer than usual, a low battery on a tag used outside, an asset moving outside site hours.
• From: 30 days of each tracker's own history; a stop counts as long from 45 minutes, quiet from 60 minutes. A stop at home is never a finding, and no finding carries coordinates.
• Applies to: organisations with GPS trackers and OAIA on.
• Change it: Cannot be changed.
• Can be wrong when: a person's routine changed this month.
• More: The heartbeat

OAIA: app findings

• Decides: for each app you have, the waiting items nobody is dealing with.
• From: Facility: an urgent work order overdue by an hour (critical at 4), unassigned for 2 hours. Meeting Room: calendar sync stopped for a day, bookings open 2 hours past their end. Security Watch: unattended 30 minutes, open 4 hours. Toilet Hygiene: dirty score 60 or more with no job. Visitor: 30 minutes past the visit, 15 minutes waiting for approval. Roll Call: drill overdue (every 90 days). Data Centre and Warehouse: critical alerts unacknowledged for 15 minutes, any for 2 hours.
• Applies to: enabled apps only.
• Change it: Cannot be changed.
• Can be wrong when: work is done outside Omaya and not closed in it.
• More: The heartbeat

Security Watch: activity when a zone is armed

• Decides: that someone is in a watched zone while it is armed, or, in a zone with no armed hours, at an hour that is normally empty.
• From: presence in the last 6 minutes, checked every 5 minutes; for a zone with no armed hours, the zone's own history for that weekday and hour. Counting sensors count as presence. One incident per armed night.
• Applies to: watched zones.
• Change it: Apps → Security Watch → Settings: Armed hours, each zone's Armed, the whitelist, and the detector Unexpected or after-hours activity.
• Can be wrong when: a cleaner or guard is not on the whitelist.
• More: What it watches for

Security Watch: out of habit

• Decides: with the AI, whether a person somewhere new to them, or a room in use at an hour it never is, is worth an incident.
• From: a zone the person has not been in for 30 days, or a zone in use at an hour that has been unused for weeks; needs at least 5 days of the person's history. Judged once per person, zone and day.
• Applies to: watched zones.
• Change it: the detector Out of habit — judged by AI.
• Can be wrong when: someone has just changed role or desk.
• More: Out of habit

Security Watch: coverage lost, heat, unknown devices

• Decides: that every gateway covering a watched zone is offline; that a temperature means possible fire; that an unregistered device is lingering.
• From: coverage: all gateways offline (stale counts as up, quiet hours left out). Heat: 45 °C or a rise of 5 °C within an hour. Unknown device: heard for 5 minutes or more, and in the last 10.
• Applies to: watched zones.
• Change it: Apps → Security Watch → Settings → Detectors: Heat alert at, Or rising faster than, Lingering for at least; Follow-up: Alert again if a critical incident is not acknowledged after (10 minutes) and Close as unattended when not detected for (12 hours).
• Can be wrong when: visitors' phones are heard as unknown devices.
• More: Settings

The apps

Toilet Hygiene: the dirty score

• Decides: how dirty a toilet probably is right now, from 0 to 100. Recomputed every minute.
• From: time since cleaned (25 points), visits in the last hour (20), people inside now (15), ammonia or a leak (10), humidity change (10), a cleaner stuck inside (10), bad feedback in 24 hours (5), a missed scheduled clean (5). A sensor that is missing reduces the points available instead of counting as clean.
• Applies to: every toilet.
• Change it: the toilet's Normal time between cleans and Dirty score limit.
• Can be wrong when: the toilet has no door counter or air sensor: the score rests mostly on time since cleaned.
• More: The dirty score

Toilet Hygiene: extra cleans

• Decides: when to send a cleaner outside the schedule.
• From: busy: visits since the last clean reach the toilet's level, worked out from its stalls, urinals and area (at least 15); waits 15 minutes before the next. Dirty: the score reaches the limit, the toilet is open, no job is open and no cleaner is inside.
• Applies to: toilets with the switches on (off as shipped).
• Change it: the toilet's Advanced settings: Send a cleaner when traffic is high, Busy when visits reach (blank is automatic), Wait before next extra clean, Send a cleaner when the dirty score passes its limit.
• Can be wrong when: the door counter counts passers-by.
• More: Busy-hour auto clean · Extra cleans

Toilet Hygiene: a job closed by the cleaner's badge

• Decides: that a clean was done, and whether the cleaner arrived on time.
• From: the badge staying inside for the toilet's clean time closes the job; a shorter stay reopens it. Arrival is the first badge visit of at least 2 minutes; shorter is a badge passing.
• Applies to: cleaners with a badge.
• Change it: the toilet's One clean takes about and Job is late after.
• Can be wrong when: the badge is not heard inside, so a real clean reopens.
• More: Closing a job · Service level

Toilet Hygiene: a stall occupied too long

• Decides: whether each listed stall is occupied, and when one has been occupied long enough to send someone.
• From: the stall's linked sensor: a presence sensor's occupied reading, or a door contact, where closed counts as occupied. A sensor silent for 30 minutes is No reading. Checked every 5 minutes; occupied past the stall's limit raises one high-priority inspection job, never a second while it is open.
• Applies to: stalls listed on a toilet's Stalls card with a sensor.
• Change it: each stall's Limit (min), 5 to 240 (20 as shipped).
• Can be wrong when: a stall door normally stands shut while free: a door contact reads it as occupied.
• More: A stall occupied too long

Toilet Hygiene: who gets the job

• Decides: which cleaner is sent.
• From: points for being in the toilet's zone, the same venue, seen there in the last hour, assigned to the zone, and being its primary or backup cleaner; absent cleaners are skipped.
• Applies to: every job raised.
• Change it: cleaner assignments and absences in the app.
• Can be wrong when: a cleaner's badge is left in a locker in the venue.
• More: Moving a job to someone else

Meeting Room: check-in, release and no-show

• Decides: whether anyone is in a booked room, and from that checks the booking in, releases it, marks a no-show or ends it.
• From: in order: radar, a presence sensor, a people counter in the last 10 minutes, tags in the room in the last 5 minutes. Empty needs a sensor saying so, or a room tags normally visit. No evidence means no release. Released 10 minutes after the start; ended after 5 minutes empty.
• Applies to: bookable rooms.
• Change it: Meeting Room settings: Ghost meeting auto-release (Enable ghost meeting detection, Trigger after (minutes)) and Auto-release on no-show.
• Can be wrong when: the room relies on tags and attendees do not carry one.
• More: How the room checks you in and out

Meeting Room: a people counter that stopped

• Decides: that a room's people counter needs attention.
• From: nothing heard for more than 4 hours opens one Facility work order.
• Applies to: people counters in meeting rooms, with Facility.
• Change it: Cannot be changed.
• Can be wrong when: the counter is switched off on purpose.
• More: How the room checks you in and out

Warehouse: misplaced stock and trucks at docks

• Decides: opens and clears warehouse alerts by itself.
• From: misplaced: a tagged unit in another zone than its bin for 15 minutes. A truck at a dock for 90 minutes, critical at twice that; a dock is occupied from its sensor (distance under 2.5 m for a distance sensor). A condition clears after 5 minutes back in range. Sensors that stopped within 15 minutes of each other are grouped as one cause.
• Applies to: warehouses with tagged stock and dock sensors.
• Change it: Warehouse Setup: Truck at a dock too long after (min), Tagged stock in the wrong place after (min), the dock threshold per area.
• Can be wrong when: a dock has no sensor: there is no guess.
• More: How to handle alerts

Data Centre: power and what keeps happening

• Decides: whether power comes from mains, battery or generator; that UPS alerts at the same moment are one mains dip; which alerts repeat and how often.
• From: the UPS flags; on-battery alerts within 2 minutes of each other are one dip; a pattern needs 3 repeats within 20 minutes of the same time.
• Applies to: data halls with UPS and environment sensors.
• Change it: the UPS side of each asset in Site limits.
• Can be wrong when: two separate dips less than 2 minutes apart are read as one.
• More: What keeps happening · How to read the power page

Energy: kWh when a meter resets or sticks

• Decides: how much energy a meter used.
• From: a rise in its counter is use; a drop is a counter reset and adds nothing; a zero is a placeholder. When the counter did not rise but power readings show use, kWh is estimated from power and marked estimated.
• Applies to: every meter.
• Change it: the meter's parent and generates settings on the meters page.
• Can be wrong when: use across a reset is not counted.
• More: Energy: questions and problems

Energy per person and kWh while empty

• Decides: net kWh per person, and how much was used while nobody was there.
• From: people are counting-sensor person-hours, else footfall visitor-days; tag counts are never used. Use is split into open hours empty, open hours with people, after hours, and no count.
• Applies to: venues with meters.
• Change it: opening hours and where each meter is placed.
• Can be wrong when: the venue has no counting sensors: the split cannot be made and is shown as no count.
• More: Energy per person and kWh while empty

Environment: comfort bands and fixed or tag readings

• Decides: labels each room-hour as over-cooled, comfortable, warm or hot, and quiet, moderate or loud; and whether a figure comes from a fixed sensor or a tag.
• From: below 23 °C over-cooled, up to 26 °C comfortable, above 28 °C hot. Noise: below 45 dBA quiet, up to 55 moderate, up to 70 loud. Fixed sensors are used first; tags only when there is none, never averaged together. A tag's reading belongs to the room it is placed in.
• Applies to: Reports → Environment.
• Change it: Cannot be changed. The temperature unit is under General.
• Can be wrong when: a tag measures its own case and reads warmer than the room.
• More: Environment

Environment: the weather outside

• Decides: the outdoor temperature and humidity for each location, hour by hour.
• From: a public weather model at the location's coordinates (about a 10 km grid), not a thermometer on the roof. Older hours are filled from the archive, about 5 days behind.
• Applies to: locations with coordinates.
• Change it: the location's coordinates.
• Can be wrong when: the building is in a valley or by the sea, so its own weather differs from the grid.
• More: Environment

Your Apps home: what needs you now

• Decides: the order of waiting items on the apps home, and which toilet is likely to need someone next.
• From: critical items first, then the oldest wait. With nothing open, the toilet busiest since its last clean is named.
• Applies to: the apps you have.
• Change it: Cannot be changed.
• Can be wrong when: a toilet's door counter is off, so another one is named.
• More: Toilet Hygiene: the dirty score