Time
Click Count
A guest checks in, but the room remains in an energy-saving mode. Elsewhere, a motion sensor reports no activity while an occupied room’s air conditioning turns down. Both problems can arise even when the sensors work correctly: the hotel’s systems may disagree about room identity, occupancy meaning, or which controller has authority to act.
Smart sensors typically connect through a room controller or gateway that translates device messages into building automation points and application events. The building management system (BMS) uses these signals for equipment supervision and control, while the property management system (PMS) exchanges reservation, check-in, checkout, and room-status information through supported interfaces. Evaluating smart sensors for hospitality property management therefore means verifying the entire signal-to-action path—not merely checking whether a device supports an advertised protocol.
The PMS and BMS answer different questions. A PMS records the commercial and operational status of a room: whether it is assigned, checked in, checked out, or unavailable for sale. The BMS monitors and manages physical systems such as heating, cooling, ventilation, and equipment alarms. Neither system should automatically treat the other’s status as a complete description of room occupancy.
A checked-in room may be temporarily empty. Motion may come from housekeeping rather than a guest. A sleeping guest may produce little detectable movement. An open balcony door describes a physical condition, not a change in reservation status.
Integration logic should combine these inputs into explicit operating states. For example, a checked-in room with no recent presence detection might enter a conservative setback mode after a configured delay, rather than being treated as unsold. Door activity can trigger reassessment, but it does not reliably establish how many people remain inside.
Keep booking state, detected presence, and control mode as separate fields. This prevents a sensor event from inadvertently changing front-desk records and makes contradictory conditions easier to diagnose.
A typical architecture has four layers. The boundaries may be combined in one device, but each function still needs an accountable owner.

Not every sensor needs a direct PMS connection. Temperature samples usually belong in the room-control or BMS path, not in reservation records. The PMS may only need to expose room-status events to an integration service. Reducing unnecessary data exchange limits interface load and avoids giving building devices access to guest information.
Ask for an architecture diagram showing both data flows and command flows. A diagram labeled “two-way integration” is insufficient unless it identifies which system reads each value, which system can write it, and what happens when a connection fails.
Compatibility has several levels: network connectivity, message transport, data interpretation, and permitted operations. A gateway can successfully communicate with a BMS while still exposing unusable point names or omitting the status information needed for safe decisions.
| Interface | Typical role | What to verify |
|---|---|---|
| BACnet | Exchange building automation values, alarms, and commands. | Supported objects and services, units, writable properties, change-of-value behavior, and command priorities where applicable. |
| Modbus | Read or write registers on meters, controllers, and equipment. | Register addresses, data types, byte order, scaling, polling requirements, and error handling. |
| MQTT | Distribute device or gateway messages through a broker. | Topic structure, payload schema, delivery settings, retained-message policy, authentication, and authorization. |
| PMS API or event interface | Exchange room assignment and operational status. | Supported events, access permissions, rate limits, delivery guarantees, version support, and reconciliation options. |
BACnet/IP support does not automatically establish support for BACnet Secure Connect or encrypted communications. MQTT defines messaging behavior, not a universal hotel-room data model. Modbus registers require a documented interpretation before their values become meaningful engineering points.
Request the relevant protocol documentation rather than accepting a compatibility logo. For BACnet, examine the Protocol Implementation Conformance Statement and the actual point list. For a PMS interface, confirm that access is available for the deployed system version and operating environment; technical support for an API does not necessarily include permission to use every endpoint.
A practical compatibility demonstration should show a real value reaching the destination with the correct room, units, timestamp, and quality indication. Where commands are required, it should also demonstrate authorized writes and resulting equipment behavior.
Incorrect mapping can send a valid command to the wrong room. This is especially easy to overlook when the PMS uses a room number, the gateway uses a device identifier, and the BMS uses a point name derived from a floor plan.
Create a controlled mapping record linking the property identifier, stable room identifier, displayed room number, controller identifier, and relevant BMS points. A displayed room number alone may not be unique across multiple buildings or properties. Device replacement should update the hardware association without altering the room’s identity.
Define the data contract alongside that mapping. At minimum, a sensor event should identify its source, measurement or state, unit where applicable, observation time, and validity. Distinguish an explicit “unoccupied” result from “unknown,” “sensor offline,” or “no update received.” These conditions demand different responses.
Commissioning should include a physical verification: activate a door contact or other identifiable sensor and confirm that the correct room changes in every consuming system. Repeat the test after a controller replacement or room-renumbering change rather than assuming the original mapping remains valid.
Room comfort should not depend entirely on a cloud round trip. Local controllers can maintain temperature regulation and apply approved fallback rules while supervisory systems exchange slower operational updates. The acceptable delay for a checkout event differs from the delay tolerated in an equipment-control loop.
Specify latency by function and measure it end to end. Useful measurement boundaries include physical event to gateway receipt, gateway receipt to application processing, and command issuance to equipment response. Also test event bursts at check-in or checkout periods; acceptable performance under light traffic does not establish capacity during concentrated activity.
Control authority needs equal attention. A guest thermostat, occupancy rule, central schedule, and engineering override may all request different setpoints. Document which request takes precedence, how long overrides last, and how normal operation resumes. In BACnet-based control, verify command-priority and relinquishment behavior where commandable objects are used.
Separate desired state from confirmed state. An API response acknowledging a command may mean only that it was accepted for processing. Where available, use controller feedback to establish whether the intended mode or setpoint was actually applied.
Buffered messages help preserve telemetry, but replaying old control events can create new faults. A delayed checkout event should not place a room into vacant mode after a subsequent check-in.
Use event identifiers and sequence or version information where the source provides them. Make processing idempotent so duplicate delivery does not repeat an action. Apply freshness rules to sensor events and commands, and reconcile the room’s current PMS state after reconnection rather than relying solely on queued history.
For MQTT, retained messages require particular care: a newly connected subscriber may receive an old state immediately. The consumer needs a timestamp and validity policy, not just the retained value.
Each connection introduces a different failure mode. A depleted sensor battery, a failed gateway, an unavailable PMS endpoint, and a BMS communication fault should not produce the same operating response.
Alarm routing should identify who must act. A leak alert needs a room or equipment location and an operational destination; simply placing it in a generic dashboard may leave it unattended. Routine hospitality integration should not replace required life-safety systems or their approved control arrangements.
Segment sensor and building-control networks from guest access networks, and restrict communication to necessary destinations and services. A gateway that reaches both building equipment and an external API deserves close examination because it crosses trust boundaries.
Use unique device or service identities, supported encrypted transport, and least-privilege permissions. A service consuming room-status events should not receive access to guest profiles or billing functions unless a separately justified requirement exists. Store credentials securely and define certificate renewal and credential rotation procedures.
Procurement review should cover firmware-update authentication, support periods, vulnerability handling, and audit logging. Logs need enough detail to reconstruct status changes and commands without collecting unnecessary guest information. Occupancy and door-event histories can reveal behavior, so retention and access policies should be deliberate.
A demonstration that changes one dashboard icon proves little about deployment readiness. Acceptance testing should exercise complete workflows with the intended PMS interface, gateway configuration, and BMS point mapping.
Define pass criteria before testing: permitted mode changes, required fault indications, acceptable function-specific latency, and recovery behavior. Report communication failures separately from semantic failures such as incorrect room mapping or stale state; both can interrupt service, but they require different remedies.
The procurement package should retain the interface specification, mapping records, control-authority rules, failure policies, and acceptance results. If an integration cannot expose these details—or requires undocumented assumptions to keep rooms comfortable—the unresolved issue is an engineering dependency, not a feature that can safely be deferred until rollout.
Recommended News
Join 50,000+ industry leaders who receive our proprietary market analysis and policy outlooks before they hit the public library.