Vayyar Care

A radar point cloud is a set of reflections in space. Turning that into a fall, a bathroom visit or a period at rest is entirely inference, and there is no version of this product that works without it.

The company is explicit that this is where it competes. Its own positioning against other manufacturers of comparable radar sensors rests on the software and algorithms rather than the hardware, which is an unusual admission from a semiconductor business and an accurate one. Several suppliers can produce a millimetre wave point cloud; classifying a minor fall from it, or distinguishing a person at rest from a person who has collapsed, is the part that is hard.

The classification set is broader than a trip alarm. Falls both major and minor, room presence, mobility, time at rest and bathroom frequency are separate inferences from the same signal, and the last two support trend analysis rather than alerting, which requires the models to be stable enough to compare across days.

Graded A on the same basis as the other radar record in this lane. There is no nurse call hardware, access control or location beacon business sitting alongside to carry value independently, so the models are not merely central but constitutive, and every question about this product is a question about them.

Alerting into the system staff already watch, with formal clinical risk process behind it and no performance calibration published.

The integration decision is the oversight decision here and it is the right one. Detections surface inside existing nurse call platforms rather than in a separate vendor application, which means an alert arrives where care staff are already trained to respond and where escalation, coverage and audit already exist. A monitoring product that creates a second console for an understaffed night shift is a product that gets ignored, and this design avoids that failure mode by construction.

The clinical safety standard achieved in January 2026 carries real oversight weight and is unique in this lane. That standard requires a manufacturer to appoint a clinical safety officer, maintain a hazard log and produce a clinical risk management file, so hazards of this system have been formally identified and mitigated by a named accountable person under an externally assessed regime. That is process accountability of a kind no competitor here demonstrates.

What remains unpublished is the calibration. No sensitivity, no false alarm rate, no alert volume per resident, no time to notification and no stated behaviour on an unacknowledged alert. The concept of fall verification appears in partner material, implying staff confirm detections, and nothing describes what proportion require it.

The shared occupancy limitation noted in third party review is an oversight matter too, since a system that struggles with two people in a room may misattribute an event.

Ask for the false alarm rate at a named installation, alert volume per resident, and whether the hazard log can be shared.

Hardware specified to a level almost no vendor in this index matches, and no performance figure attached to any of it.

The silicon is public. The radar on chip platform is documented as covering imaging and radar bands from 3 to 81 gigahertz with up to 72 transceivers and an integrated digital signal processor, which are the specifications that physically bound what resolution and range are achievable. A reader with radar knowledge can evaluate whether the claims are plausible from first principles, which is a different and stronger kind of transparency than an accuracy assertion.

The coverage disclosure is the part that costs something. Range is published as 13 feet to the front and 6 feet 6 inches either side, together with an explicit statement that large rooms may require more than one device. Publishing the boundary of your own sensor, and volunteering that a single unit will sometimes be insufficient, is disclosure against commercial interest, and it stands in direct contrast to unquantified multi room coverage claims elsewhere in this lane.

The method is described concretely too: point cloud imaging from 4D radar, wall or ceiling mounting, penetration of obstructions including shower curtains and steam, and an explicit statement that no optical data is collected.

What is absent is the model layer. No detection accuracy, no false alarm rate, no operating point, no model card and no architecture description. The classification set is enumerated without any measure of how well each classification performs.

Graded B because the physical and functional disclosure is unusually complete and the performance disclosure is absent. Ask for detection performance per classification at a stated operating point.

The most consequential component in the chain is designed in house and specified publicly, which answers the question this axis usually has to leave open for hardware vendors.

The silicon is the company's own. Vayyar designs the radar on chip platform this product runs on and publishes its specification, covering 3 to 81 gigahertz with up to 72 transceivers and an integrated digital signal processor. For every other sensing vendor in this lane, the radar front end is almost certainly a third party component from one of a small number of semiconductor suppliers, undisclosed and representing a concentrated dependency. Here the dependency does not exist, because the sensing company is the semiconductor company, and it says so with numbers.

Inference likewise appears to run on the vendor's own stack, described as a system on chip with a proprietary software stack and machine learning algorithms, which places no third party model provider in the classification path.

The distribution and integration layer is named openly: nurse call manufacturers, care platform partners and regional distributors are all identified, and each is independently verifiable.

What remains undisclosed is the backend and the data. No cloud or hosting provider is named, no sub processor register was located, and no position is published on training data provenance or on whether data from deployed sensors contributes to model development. That last question carries weight given the deployment base includes long term care residents with dementia who cannot meaningfully consent.

Ask for the hosting provider and sub processor register, and whether deployed sensor data trains models.

The strongest field deployment evidence in this lane and no measured performance behind it.

What exists is multinational, multi year and named. A county council in the United Kingdom reported eliminating hospital admissions and long lies following one of the largest single installations of this technology, published in January 2026. A care home published a resident safety case study in April 2026. An Ontario long term care provider piloted 122 units across a facility housing more than 250 residents. A distribution partnership dating to 2018 is described as having detected hundreds of real falls across assisted living residences and private homes. Deployment across four countries through nurse call partners is verifiable through those partners rather than resting on vendor assertion alone.

That is materially more than the other radar record here, which offers channel announcements and a laboratory demonstration.

What is missing is the denominator. Hundreds of falls detected says nothing about falls missed, and no sensitivity, specificity or false alarm rate was located for any deployment. Eliminating long lies is a meaningful operational outcome and it is reported by the deploying authority in a case study rather than measured against a control period or a comparator technology, so the counterfactual is unknown.

No peer reviewed publication was located, and for a technology deployed at this scale over this many years the absence is notable rather than expected.

Ask for detection sensitivity and false alarm rate from a named installation, and for the before and after measurement behind the hospital admission claim.

A modality choice that removes the worst risk outright, with no operational handling detail behind it.

The company states plainly that its sensors collect no optical data. That is a stronger guarantee than a policy commitment not to view images, because the technology cannot produce one. The consequence is most valuable exactly where it is hardest to achieve: radio waves pass through shower curtains and steam, so the bathroom, which is where falls concentrate and where a camera is least acceptable to any resident, is covered by a sensor that cannot see. Solving the highest risk room with the lowest exposure modality is a genuinely good design outcome and it is the central reason this product exists.

Coverage limits being published rather than obscured is a second point in its favour on this axis, because a buyer knows which parts of a dwelling are and are not observed.

Everything operational is missing. No encryption statement, no retention schedule, no access control model, no deletion process and no position on whether deployment data contributes to model development.

The channel adds a specific ambiguity. Detections flow into a partner's nurse call or care platform, so the record of when a resident used the bathroom and how long they rested is held in a system Vayyar does not operate. Nothing describes what Vayyar retains versus what the partner retains, and a buyer asking where their residents' data lives will get an answer from neither party's public material.

Ask what leaves the sensor and to whom, what Vayyar retains centrally, and how retention is divided with platform partners.

A dedicated pass located no United States health privacy position: no compliance statement, no business associate agreement template, no execution requirement and no description of which entity contracts with whom.

The channel makes the contracting question genuinely unclear rather than merely undisclosed. This product reaches facilities embedded in nurse call platforms sold by other manufacturers, so the operator's counterparty is usually TekTone, Austco, K4Connect or a regional distributor. Whether Vayyar signs anything directly with a covered entity, sits as a subcontractor to the platform vendor, or never enters the health privacy chain at all is unstated, and those are three different answers to a question an operator's compliance officer has to resolve before purchase.

Evidence of governance elsewhere does not close this gap and should not be read as closing it. A clinical safety standard achieved under United Kingdom health technology rules concerns clinical risk management rather than data protection, and a county council deployment implies European data protection compliance without publishing any of it. Neither speaks to United States health privacy obligations, which is what this axis measures.

The design does reduce what is at stake, since no optical data exists to be disclosed, but reduced stakes are not a stated posture.

Graded D on the same standard applied to the other sensing vendors in this lane, all of which likewise published nothing here. Ask whether Vayyar signs agreements directly with operators or only with platform partners, and what the partner agreements require downstream.

Two dedicated passes located no security page, no information security attestation, no trust center, no penetration testing statement and no vulnerability disclosure policy.

The clinical safety standard achieved in January 2026 does not fill this gap and should not be credited here. It governs clinical risk management, requiring hazard analysis and a named safety officer, and it says nothing about encryption, access control, secure development or breach response. Two different regimes, and only one is evidenced.

The device surface is substantial. These are connected sensors installed permanently in bedrooms and bathrooms across care facilities in several countries, integrated into nurse call systems that are themselves life safety infrastructure. Device authentication, firmware signing and update practice, and how a sensor is provisioned and decommissioned are all unaddressed, and a compromised sensor in a nurse call chain is a compromise of an emergency system rather than of a reporting tool.

The partner channel adds an access question. Installation and configuration are performed by nurse call manufacturers and their distributors, so installer credentials and platform access exist between the vendor and the resident, and nothing describes how that access is scoped or revoked.

The data generated is exactly what would be useful to someone targeting a vulnerable adult, since it indicates when a person is alone, asleep or in the bathroom.

One pre emptive note: further clinical or regulatory credentials cannot move this grade. Only an information security attestation, or security documentation available under agreement, will.

Ask for any external security assessment, the firmware update and authentication design, and how partner installer access is controlled.

Two named regulatory compliances and a product scope deliberately kept clear of the device boundary, which together make this the strongest regulatory position in this lane.

The communications authorisation is stated. The device is described as fully approved by the United States communications regulator, which is the spectrum and emissions authorisation any radio frequency emitting device operating continuously in an occupied dwelling requires. That regime applies to every radar vendor in this category and this is the only one to state its position on it.

The clinical safety standard is the more substantive credential. Achieving DCB0129 in January 2026 places the manufacturer under the United Kingdom clinical risk management standard for health technology, which requires a clinical safety officer, a hazard log and a clinical risk management file assessed against a published standard. It is not a product approval and should not be read as one, but it is externally assessed evidence that clinical hazards were formally analysed, which is more than an unaudited safety claim.

The scope discipline matters as much as the credentials. Vayyar Care claims falls, presence, mobility, rest and bathroom frequency, and makes no vital signs claim at all. Contactless measurement of heart rate or respiration is where a comparable product in this lane approaches the device line without stating a determination; declining to make that claim keeps this one on the safety and alerting side where the wellness path is defensible.

No United States device determination is published and none is obviously required on current claims.

Ask for the written device determination for the United States market, and for the clinical risk management file.

A dedicated pass located no fairness testing, no subgroup performance, no calibration data, no validation methodology and no external algorithmic audit.

The bias mechanism in radar is physical and applies here exactly as it does to the other radar record in this lane. What the sensor receives is energy reflected from a body, and both the strength and the shape of that return vary with body size, body composition and posture. A wheelchair is a large metal reflector that dominates a scene. Walking frames, hoists, adjustable beds and oxygen equipment introduce static and moving metal into precisely the rooms where this product is installed. Bedding and clothing attenuate differently.

The deployed population is defined by those characteristics. Long term care and dementia settings are where mobility aids concentrate, and the residents most likely to fall are the residents least likely to move like a typical training subject. A model separating a fall from a controlled descent by reading body position through the movement is directly exposed to gait variation from stroke, Parkinsonian movement and general frailty.

One performance boundary is publicly known and unquantified. Third party review describes limitations with multiple residents in a room, which is a fairness relevant constraint for shared long term care rooms and for couples aging together, and no figure describes how much performance degrades.

The clinical safety standard achieved would ordinarily require hazards of this kind to be identified in a hazard log, and none of that analysis is public.

Ask for detection performance among residents using wheelchairs and walking frames, in shared rooms, and whether mobility aids were represented in development.

Nothing allocates responsibility, for a product whose primary function is detecting an emergency.

A dedicated pass located no service level agreement, no detection warranty, no uptime commitment, no indemnity and no remediation position for either the facility or the consumer offering.

The channel makes allocation genuinely difficult rather than merely undisclosed, and more so here than for a directly sold product. A care operator contracts with a nurse call manufacturer, which embeds a sensor built by Vayyar, running models trained by Vayyar, alerting through the manufacturer's platform, monitored by the operator's own staff. When a fall is not detected, or an alert does not arrive, responsibility could rest at any point in that chain, and no published material from the vendor describes how it is divided or what it warrants to its own partners.

The absent number compounds it. Without a published detection rate there is no basis for characterising residual risk, so an operator replacing pull cords or pendants with ambient sensing cannot establish whether coverage improved or degraded, and cannot price the change.

One mitigating fact belongs on the record. The clinical safety standard achieved requires a named clinical safety officer and a maintained hazard log, which is real process accountability and more than any competitor here demonstrates. It is not contractual recourse, and it does not tell a customer what happens when the system misses.

One pre emptive note: further integrations, deployments or safety certifications cannot move this grade. Only a contractual commitment, or published detection and alert delivery performance, will.

Ask what Vayyar warrants to its platform partners, and what those partners pass through to operators.

Multiple named production integrations with the systems care staff actually use, which is the strongest interoperability position in this lane.

The named list is specific and verifiable from the partners' own material rather than only from the vendor: TekTone Tek-CARE and Austco Tacera nurse call platforms, K4Connect K4Community, TLS Evolve and SenSights.AI. The company states interoperability with any nurse call system and care platform, and unusually the claim is supported by several independent implementations rather than resting on assertion.

Choosing nurse call as the integration target is the correct judgement for this setting and deserves saying. In senior living the nurse call platform, not the medical record, is where alerts are received, escalated and audited, and where staff spend the shift. Integrating there means detections enter an existing accountable workflow rather than a parallel one. One partner alone carries the integration across four countries, which indicates the interface is productised rather than bespoke per site.

What is missing is the clinical record. No electronic health record or care management system is named, no interface standard such as HL7 or FHIR is described, and nothing states whether a detected fall is written to a resident record as a discrete, trendable entry.

That gap has a consequence worth naming. Mobility, rest and bathroom frequency are longitudinal signals whose value is in trend, and a trend that lives only in an alerting platform does not reach care planning or incident review.

Ask which care management systems are integrated, through what standard, and whether detections write to the resident record.

The physical topology is clear and the backend behind it is not.

What can be established is more than usual because of how the product is delivered. Sensors are mounted on premise in the room they observe, and detections feed into a nurse call or care platform that is itself typically installed on site. That architecture keeps a substantial part of the data path inside the building rather than routing every event through a vendor cloud, which is a materially different residency picture from a platform product.

Everything about the vendor side is absent. No cloud or hosting provider is named, no region is stated, no residency commitment is made, no tenancy model is described and no retention position is given for whatever Vayyar itself receives.

The international footprint makes the omission consequential rather than academic. Deployments span the United Kingdom, Canada, Australia and Japan alongside the United States, and a United Kingdom public sector installation in particular operates under data protection rules with explicit location and transfer requirements. Nothing published addresses how those are satisfied, or whether processing differs by market.

The partner model introduces a further ambiguity that cuts both ways. Data landing in a partner platform may never reach Vayyar at all, which would be a strong residency answer, or may be mirrored for model improvement, which would be a weak one. Nothing indicates which, and the difference is the whole question.

Ask what data reaches Vayyar infrastructure and where it sits, how European deployments are handled, and whether processing is regionalised.

A consumer price in the public record, an enterprise price nowhere, and one disclosure that partly compensates.

The consumer figures come from third party review rather than a vendor rate card: roughly 240 dollars per unit through a retail channel plus a monthly subscription in the region of 20 dollars. Usable, and not published by the vendor.

What does compensate is the coverage specification. Because the sensor's range is published as 13 feet to the front and 6 feet 6 inches either side, with an explicit statement that large rooms may need more than one device, a buyer can compute how many units a floor plan requires before speaking to anyone. That is the calculation the other radar vendor in this lane makes impossible by claiming multi room coverage without a specification. A published limit converts an unknowable cost into an arithmetic problem, and this index should say so.

Enterprise pricing is absent and structurally so. The product reaches facilities through nurse call manufacturers and care platform partners, so the price an operator pays is set by TekTone, Austco, K4Connect or a regional distributor rather than by Vayyar. Nothing states whether the sensor is licensed, sold outright, or bundled into a nurse call refresh, and nothing indicates recurring cost at the facility tier.

No return proxy is published. Fall reduction, avoided admission cost and staff time released are all argued and none is quantified with a method.

Ask for the facility pricing basis, whether the sensor is sold or licensed, the recurring cost per unit, and what the partner channel adds.

The most international footprint in this lane, evidenced by named deployments rather than asserted.

Settings span independent living, assisted living, long term care, dementia and memory care, private homes, and hospital and remote patient monitoring through nurse call partners. Public sector social care is included through a county council deployment, which is a procurement context almost nothing else in this lane has entered and one that imposes its own scrutiny.

Geography is the differentiator. Named channel and deployment evidence covers the United States, the United Kingdom, Canada, Australia, Japan and Israel, with one nurse call partner alone carrying the product across four of those markets. Most vendors in this lane are single market, and international deployment implies the product has satisfied several regulatory and procurement regimes rather than one.

Clinical scope is deliberately narrow and should be read that way. Falls, presence, mobility, time at rest and bathroom frequency is the whole set. There is no activity of daily living inference, no condition prediction and no vital signs, so a buyer comparing this against the wearable or multimodal records in this lane is comparing different products rather than better and worse ones.

One known boundary appears in third party review rather than vendor material: performance in rooms with multiple occupants is described as limited, which matters for shared rooms in long term care and for couples aging at home.

Ask how the system performs in shared occupancy, the device count for a typical long term care room, and which markets carry local regulatory approval.