The part of the quote that decides whether it works.

Line of sight is the whole game

Every accuracy number you will be shown for an indoor positioning system — including ours — is measured with a clear path between the tag and the reference point. That condition is called line of sight, and in a real building it is the exception rather than the rule. A radio pulse that reaches a receiver after reflecting off a wall has travelled further than the straight-line distance, so the calculated position is pulled away from the truth. This is not a defect to be fixed by better firmware; it is geometry.

The practical consequence is a design rule rather than a disclaimer: place reference points so that the tags you care about can usually see them, and then set your expectations to what survives when they cannot. A worn badge sits at chest height on a moving person who is often turned away, often behind a trolley, often through a door frame. Designing for that reality — and targeting certainty about which room rather than a position within it — produces a system whose record holds up. Designing to the datasheet produces one that disagrees with itself at the worst moment.

At survey we simulate this deliberately: walk the route wearing a tag the way a carer would, and note every point where something solid comes between it and a proposed mounting position. Bulkheads, lift shafts, steel fire doors, sluice rooms, plant enclosures, stacked linen trolleys and the lead-lined walls around imaging are the usual offenders.

Geometry: where the points go

For presence — knowing a tag is in a zone — one reference point is enough and geometry barely matters. For boundaries, two points along the line of travel let the sequence of detections reveal direction. It is only full area coverage where arrangement starts to decide quality, and the principles are consistent:

• Spread, do not cluster. Points arranged around a space produce stable positions; points bunched along one edge produce positions that are confident in one axis and vague in the other. A roughly square arrangement is the ideal, and long thin rectangles are where accuracy degrades first — as a working rule, keep the covered area's length within about twice its width, and the angles subtended at the middle of the space wide rather than shallow.
• Keep heights consistent. Mixing a point at 2.4 m with one at 5 m in the same group introduces a vertical ambiguity the system has to resolve from weaker information. Where a building forces it, record every height precisely.
• Mount high, within reason. Higher points see over people and equipment, which is the main enemy of line of sight. Four to five metres is a good target where a ceiling allows it; in a domestic-scale retirement cottage, above door height is usually the practical answer.
• Stand off the walls. A unit pressed flat against a wall, or tucked into a corner, gets reflections from the surface immediately behind it. Half a metre of clearance is a reasonable minimum and costs nothing at design time.
• Respect spacing. Reference points that are too far apart leave gaps; too close together and you have paid for overlap. Tens of metres between points is the normal working range indoors, set at survey against the ceiling height and what is in the way.

These are planning rules of thumb, not universal constants — the right spacing for a 2.6 m domestic ceiling and for a 6 m atrium are not the same number, which is exactly why the survey exists.

Cabled or battery: the decision that sets the price

The mistake is treating this as one decision for the whole building. Most real projects are mixed: cabled where a ceiling is open anyway, battery where the disruption of cabling costs more than the equipment. In a retirement village spread across cottages and gardens, battery points are usually what makes the project quotable at all — there is no trench to dig and no occupied unit to close.

Four South African constraints

• It has to record through an outage. Reference points, gateways, switches and the server all need backup power, and the system must keep writing locally rather than dropping the shift. This is a specification item at survey, sized against your actual outage pattern — not an accessory sold afterwards. See load shedding protection.
• It has to survive cleaning. Wearables and any equipment in a clinical area get wiped with whatever the infection prevention and control programme specifies. Sealed housings rated IP67 or better are the practical floor; anything that cannot take disinfectant will either be cleaned improperly or taken off.
• The radios have to be legal to supply. Two separate things, often confused. Equipment of this class falls within the licence-exempt schedule of the radio frequency spectrum regulations, so no per-site spectrum licence is needed to operate it. Separately, and regardless of that exemption, the Electronic Communications Act requires radio apparatus supplied and used in South Africa to hold ICASA type approval. Write it into the supply agreement rather than assuming it, whoever you buy from.
• The server is part of the building, not the cloud. A modest on-premises machine runs the positioning engine and the event database — in practice a current multi-core processor, ample memory and a disk sized to your retention policy, on the backed-up side of the comms room. Keeping processing on site is also what makes the POPIA position straightforward, as our POPIA guide explains.

What a survey must actually produce

A survey that produces only a quantity of reference points is not a survey. The output should be six things:

• The sentences. Area by area, the statements the system must be able to produce. These set the resolution tier, and the resolution tier sets the cost — see the scenario map.
• Ceiling heights per area, measured rather than assumed, because height decides coverage per point and whether a mounting method is possible at all.
• An obstruction map from walking the route with a tag — including the trolleys, screens and equipment that are not on any floor plan.
• The cable line: which areas can take a cable, which cannot, and what the disruption would cost in each.
• A coordinate reference. An origin, and every planned position recorded against it including height above floor. Commissioning depends on these numbers being right; a point installed 1.5 m from where the system believes it is will quietly bias every position near it.
• Acceptance criteria, written before installation, in the client's words: which room, which boundary, how quickly, how reliably. Prove them in the first area before the rollout continues.

How installations go wrong

• The unrecorded move. A point is relocated a metre during installation because of a beam, and nobody updates the coordinate record. Every position in that area is now biased, and the symptom — occasional wrong-room readings — looks like a software fault for months.
• The ceiling void. Mounting a unit above a suspended ceiling to hide it puts a layer of tiles and services between it and every tag. It is neat, and it is the commonest self-inflicted accuracy problem.
• Designing for an empty building. Surveys done on a quiet Sunday miss the trolleys, screens, visitors and stacked equipment that fill the space on a Tuesday. Walk it busy.
• Forgetting the outdoor boundary. In a village the real perimeter is a garden gate or a path. Indoor-rated equipment on an outdoor boundary fails in the first serious weather.
• Promising centimetres. Quoting line-of-sight accuracy as a delivery commitment sets up a dispute at handover that the physics will win. Commit to what the use case needs — room and door certainty — and specify it that way in the acceptance criteria.

See how this is put together on our nurse call and staff safety page, or the buying comparison in nurse call vs location systems.

Primary sources

• ICASA — Notice to amend Annexure B of the Radio Frequency Spectrum Regulations, Notice 1822 of 2023, Government Gazette 48643, 23 May 2023 — the licence-exempt schedule in force; type approval under section 35 of the Electronic Communications Act applies separately and regardless
• ICASA — type approval

Spacing, height and geometry figures in this article are planning rules of thumb for design discussion, not guarantees — the correct values for a specific building come from its survey.