A hotel conversion where the corridor design was the problem

A Victorian building converted to a boutique hotel, with a wireless network installed during the conversion on a corridor-mounted design: 22 access points, roughly four or five per floor, in the corridors.

On a survey walked along the corridors, the network looked excellent. In the rooms, where guests are, it did not.

Why corridor designs fail in old buildings

A corridor is a waveguide. Signal propagates along it efficiently and through the walls either side poorly. In a modern building with plasterboard partitions the penalty is a few decibels and a corridor design can work.

This building has solid masonry internal walls, some of them 400 mm thick, plus cast iron soil pipes, lath and plaster, and — in fourteen rooms — original metal-backed mirrors that turned out to matter.

Measured attenuation through a representative internal wall was 18 to 24 dB at 5 GHz. A corridor access point delivering -50 dBm in the corridor was delivering -70 to -75 dBm two metres into the room, and worse in the bathroom beyond.

In-room coverage at -67 dBm or better was 41 percent. Guests were reviewing the hotel accordingly.

The commercial context

Twenty-three negative online reviews per month mentioned the Wi-Fi. For a boutique hotel competing on experience, that was the entire business case — it did not require a technical justification.

The design decision

Two options were costed:

More corridor access points. Cheaper to install, using existing containment. Modelling showed it would improve the worst rooms and still leave roughly a quarter of rooms below target, because the problem is the wall, not the distance.

In-room wall plates. An access point in each room or shared between two, wall-mounted at desk height, with the room's data outlet. Far more cable, far more disruption, and a genuine solution.

The hotel chose in-room. In a building of this construction it was the only option that would actually work, and the review scores made the economics straightforward.

Implementation

• 61 wall-plate access points, one per room in 28 rooms with difficult geometry and one shared between two rooms elsewhere.
• Phased by floor, one floor at a time, with the floor taken out of service for six working days. This was the constraint that set the nine-week timeline.
• Cable routed in existing risers where possible; two new risers were formed with conservation approval, which added three weeks to the programme and had to be applied for before the project started.
• Per-room VLAN isolation so guests cannot see each other's devices, with the access point's built-in client isolation as a second layer.
• 2.4 GHz retained at low power in roughly a third of rooms for guest IoT — chiefly personal devices guests bring with them.
• Channel plan at 20 MHz throughout. With 61 access points in a building of this size, narrow channels were not a compromise but a requirement.

What the masonry gave back

The thick walls that caused the problem also solved the channel-planning problem. 18 to 24 dB of isolation between rooms means co-channel reuse at very short distances is viable — a channel can be reused two rooms away rather than two floors away. A building that is hostile to coverage is generous with isolation.

In-room coverage at -67 dBm or better went from 41 percent to 99 percent. Mean in-room throughput went from 12 Mbps to 145 Mbps.

Negative reviews mentioning Wi-Fi fell from 23 per month to 1. The hotel's own guest satisfaction score for Wi-Fi moved from 2.9 to 4.6 out of 5.

Cost, honestly

The in-room design cost approximately 2.4 times the corridor-densification option. It was still the right decision, because the cheaper option would not have solved the problem and the hotel would have paid for both.

The conservation approvals for two new risers were the single largest programme risk and had to be started before anything else. In a listed or historically significant building, that approval is the critical path — not the network.