Network, automation, compliance and energy for the conversion of The Failford Inn into a four‑bedroom short‑term let over a self‑contained first‑floor flat.
Everything else in this note can be revisited. These cannot, because the cost of changing them rises by an order of magnitude once the walls are closed:
Taken from the mapped building footprint at 55.5046 N, 4.4402 W: the inn and both neighbouring ranges share a long axis of 57°, so the principal elevations face SSE (147°) and NNW (327°). The SSE pitch is effectively the best roof in the building — see the energy section.
AP positions were optimised against the wall geometry taken from the drawings, using a log‑distance path‑loss model with a 5 dB penalty per wall crossing at 5 GHz. Four access points cover 100 % of the internal floor area above −67 dBm, which is the threshold worth designing to — it is where a video call stops degrading. The weakest room is Bedroom 1, right on −67.
| Ref | Room | Level | Position (m) |
|---|---|---|---|
| AP1 | Hall (south) | Ground | 6.8, 3.9 |
| AP2 | Kitchen | First | 7.3, 6.4 |
| AP3 | Lobby (east) | Ground | 9.9, 6.3 |
| AP4 | Lounge / Dining / Kitchen | Ground | 2.6, 10.9 |
Positions are metres from the south‑west datum of the drawing set. None sits in a guest bedroom — APs there get unplugged, complained about, or draped with a towel.
The utility is the right home for the cabinet: it is staff‑only, lockable, already has power and the water manifold, and sits within 30 m of every outlet. Allow a 9U wall cabinet at 1.6 m, a double socket on its own way, and ventilation — a router, an 8‑port PoE switch, an NVR and a battery will dissipate 100–150 W continuously.
Lengths are Manhattan routes through the ceiling void along the circulation spine, with a riser between floors, 10 % slack and a 3 m service loop at each end. Order 500 m of Cat6 U/UTP solid copper against the 449 m modelled.
| Ref | Serves | Room | Height (m) | Cable (m) |
|---|---|---|---|---|
| AP1 | Ceiling Wi-Fi AP (PoE) | Hall (south) | 2.45 | 15.7 |
| AP2 | Ceiling Wi-Fi AP (PoE) | Kitchen | 5.35 | 29.1 |
| AP3 | Ceiling Wi-Fi AP (PoE) | Lobby (east) | 2.45 | 21.8 |
| AP4 | Ceiling Wi-Fi AP (PoE) | Lounge / Dining / Kitchen | 2.45 | 28.4 |
| CAM1 | Camera — Front entrance | Vestibule | 2.60 | 14.0 |
| CAM2 | Camera — East gable / parking | Bedroom 1 | 2.60 | 34.7 |
| CAM3 | Camera — North / bin store | Lounge / Dining / Kitchen | 2.60 | 36.0 |
| CAM4 | Camera — Balcony head | Lounge | 2.60 | 26.9 |
| D1 | TV point | Lounge / Dining / Kitchen | 0.60 | 31.3 |
| D2 | TV point | Lounge | 0.60 | 30.2 |
| D3 | Data — bedroom | Bedroom 1 | 0.60 | 35.9 |
| D4 | Data — bedroom | Bedroom 2 | 0.60 | 23.4 |
| D5 | Data — bedroom | Bedroom 4 | 0.60 | 31.2 |
| D6 | Data — bedroom | Bedroom | 0.60 | 39.1 |
| D7 | Door controller / keypad | Lobby (east) | 1.40 | 20.7 |
| D8 | Plant — heating / meters | Store (W) | 1.20 | 30.6 |
16 outlets · 449 m routed · 80 W of PoE load. Specify a switch with at least 120 W of PoE budget so there is headroom for a fifth AP and a doorbell.
Stay on UniFi — you already run the controller and the UISP estate, so this property costs you no new operational surface. A gateway, one 8‑port PoE switch and four ceiling APs. Adopt the site into the existing controller and it inherits your alerting.
Guest — client isolation on, rate‑limited, no access to anything.
Property — locks, sensors, hubs. No inbound, no internet for anything that does not need it.
Cameras — external only, no internet route at all.
Management — switch, APs, cabinet.
Pin 2.4 GHz to channels 1, 6, 11 only, at reduced power. Do not let the controller roam it.
Put Zigbee on channel 15, 20 or 25 — the gaps between those Wi‑Fi channels. This one decision prevents most "the sensors are flaky" callouts.
Do the guest SSID on both bands; some guest kit still cannot see 5 GHz.
A single letting site review mentioning bad Wi‑Fi costs more than the entire network. Put the SSID and password on the welcome card and on a QR code by the door, and never rotate it mid‑stay.
A let is not a home. The person using the building has never seen it before, will not read instructions, and cannot be asked to re‑pair anything. That inverts the usual priorities: reliability and graceful failure beat capability.
Every automation must degrade to a working manual control. If the hub is off, the internet is down, or a guest has done something inexplicable, the light switch still turns on the light. Design backwards from that and most of the other choices follow.
Cheap, battery‑efficient, mesh‑forming, and the widest device choice by a distance. Every mains device becomes a router, so a property this long self‑heals. Use it for sensors, TRVs, relays and buttons.
Better security model and it avoids the crowded 2.4 GHz band, which matters for a lock. But UK device choice is thin and prices are high. Worth it for the entrance lock if the lock you want speaks it; not worth a second network otherwise.
The right long‑term answer and the wrong one to bet a build on today. Buy a controller with a Thread border router built in so you can add Matter devices as the ecosystem matures, then keep buying Zigbee for now.
Wi‑Fi sensors flatten batteries and clutter the AP client table. Smart bulbs are worse: a guest turns the wall switch off and the bulb goes unavailable, so the automation fails and the room looks broken. Use dumb bulbs on smart relays.
Home Assistant on dedicated hardware in the comms cabinet — a Yellow, a Green, or a small mini‑PC. Wire it, do not put it on Wi‑Fi. Add a Zigbee coordinator on a USB extension lead, not directly in a port: USB 3 and cabinet metalwork both wreck 2.4 GHz reception, and this is the single most common cause of a flaky Zigbee mesh. Everything runs locally; the internet is for notifications and remote access, not for turning lights on.
Conventional circuits and conventional switches, with a Zigbee relay or dimmer module behind each switch or above each ceiling rose. The guest sees a light switch and it works. You get scenes, occupancy control and the ability to prove the lights are off. This is why the electrician needs neutral at every switch drop — it costs nothing now and rules out half the product catalogue if it is missed.
The fire alarm system is not part of the automation. Scotland requires mains‑powered, interlinked smoke alarms on every storey, a heat alarm in the kitchen and a CO alarm where there is a fuel‑burning appliance, and this is an explicit condition of the short‑term let licence. Fit a proper BS 5839‑6 system. Smart smoke sensors are supplementary monitoring so you learn about an alarm while the property is empty — they never substitute for the real one.
Short‑term letting in Scotland has been licensed since 2023 and the licence carries mandatory conditions. Several of them constrain the services design directly.
No indoor cameras anywhere, in any form, disclosed or not. The major letting platforms prohibit them outright and a listing can be removed for it. The four cameras in the schedule are all external, covering the entrance, the parking side, the bin store and the balcony stair. Disclose them in the listing, angle them off the neighbouring boundary, and do not cover the balcony itself — that is guest amenity space.
Noise monitoring is permitted and sensible, on one condition: the device must measure sound level only and never record or transmit audio. Disclose it in the listing and site it in the circulation space, not a bedroom.
A keypad lock on the entrance, and a code generated per booking that is valid from 16:00 on arrival to 11:00 on departure and dead outside that window. It goes out automatically the morning of arrival, together with the Wi‑Fi details and the parking note. Nobody meets anybody, no key safe gets photographed and shared, and an early arrival simply cannot get in — which is a feature, because the cleaner is still there.
Keep one permanent maintenance code and a separate cleaner code that only works on changeover days. When someone stops working for you, you revoke a code, not a key.
Two thresholds, not one. A sustained level above roughly 70 dB after 23:00 plays a discreet chime inside and sends you a message. A second breach within the hour sends the guest a polite templated message. Only a third gets a phone call. Most parties in let properties are not malicious, they are people who have lost track of the time, and the chime ends it. It also gives you a dated record if the neighbours ever complain.
The cleaner's code unlocking the door is the signal that the turnaround has started. That starts a checklist on their phone, turns the heating up while they work, and turns it back down when they leave. What you want from it is the evidence: door opened at 10:40, closed at 13:05, so you know the property was serviced without asking anyone.
On departure the property resets itself — everything off, heating to 12 °C frost protection, lock re‑secured, and a flag raised if a window was left open.
Modelled hourly for the year: solar geometry for the site latitude, an isotropic sky model, and monthly clearness and diffuse fractions for west‑central Scotland, calibrated so a south‑facing array at 35° returns the 870 kWh/kWp that this part of the country actually delivers. Demand is built from a letting pattern — heavy hot water and laundry on changeover days, near‑base load when the property is empty — not a domestic profile.
The lounge roof is the balcony, so it is out. That leaves the pitched roof over the two‑storey block (11.4 × 6.5 m on plan, so roughly 45 m² per pitch) and two single‑storey roofs of about 38 and 25 m². One good pitch alone takes around 7.5 kWp, so roof area is not the constraint. Demand is.
| Configuration | Gen | Import | Export | Gas | Run cost | Saving | Capex | Payback |
|---|---|---|---|---|---|---|---|---|
| A · Do nothing (gas boiler) | — | 2756 | — | 14987 | £2,004 | — | — | — |
| B · 4 kWp PV | 3480 | 1446 | 2170 | 14987 | £1,337 | £667 | £5,400 | 8.1 y |
| C · 6 kWp PV | 5220 | 1331 | 3795 | 14987 | £1,063 | £941 | £6,900 | 7.3 y |
| D · 6 kWp PV + 10 kWh battery | 5220 | 242 | 2508 | 14987 | £972 | £1032 | £11,400 | 11.0 y |
| E · 8 kWp PV + 15 kWh battery | 6960 | 119 | 4113 | 14987 | £699 | £1305 | £15,950 | 12.2 y |
| F · 8 kWp + 15 kWh + HW diverter | 6960 | 119 | 2262 | 13161 | £860 | £1144 | £16,850 | 14.7 y |
| G · ASHP + 10 kWp + 20 kWh | 8700 | 2429 | 3554 | — | £310 | £1694 | £35,200 | 20.8 y |
| H · 4 kWp + HW diverter (no batt) | 3480 | 1446 | 1193 | 14014 | £1,421 | £583 | £6,300 | 10.8 y |
| I · ASHP + 8 kWp + 10 kWh | 6960 | 2917 | 2367 | — | £615 | £1389 | £29,700 | 21.4 y |
| J · 6 kWp + 10 kWh on time-of-use | 5220 | 1378 | 3651 | 14987 | £834 | £1170 | £11,400 | 9.7 y |
| K · 0 kWp + 10 kWh on time-of-use | — | 3153 | — | 14987 | £1,556 | £448 | £4,500 | 10.0 y |
| L · ASHP + 8 kWp + 15 kWh on ToU | 6960 | 4585 | 3818 | — | £91 | £1913 | £45,450 | 23.8 y |
kWh/yr. Modelled at 26.1 p import, 15 p export, 6.4 p gas, with standing charges. Capex at £1,020–1,350/kWp for PV and £450/kWh installed for storage.
The table above is the property without the tub, and it is worth keeping because it shows why the answer changes. On that demand, 6 kWp with no battery paid back in 7.3 years and nothing beat it; a 10 kWh battery cost £4,500 to save £91 a year, and a hot water diverter actually lost money because it displaced 6.4 p gas to give up 15 p of export. The property simply did not use enough electricity for anything clever to be worth doing.
The hot tub changes all of that, and it is dealt with next.
Superseded by the hot tub — see work package T. On the building alone the answer was 6–8 kWp and no battery. With the tub added, both a larger array and storage pay their way.
The heat pump case is a long payback on energy alone (21–24 years) — take that decision on EPC banding, future gas prices and the property's carbon story instead, and if you might take it later, run the duct now.
Tilt is forgiving here; aspect is not. The same 6 kWp returns 5,220 kWh facing south at 35° and 3,579 kWh facing north — a 31 % swing that moves payback by three years. East or west lose about 12 %, which is quite survivable. Confirm the orientation before anyone quotes.
Three things change the numbers above and all of them need the accountant, not the engineer: a self‑catering property is likely on a commercial electricity contract with no price cap, 20 % VAT rather than 5 % and possibly the Climate Change Levy, which makes avoided import worth more and improves PV; capital allowances on the installation improve the effective payback materially; and the domestic grant schemes (Home Energy Scotland, Boiler Upgrade) are aimed at owner‑occupied homes, so assume a let does not qualify until someone confirms otherwise.
A tub in the rear garden is the single biggest load on the site, and it changes the energy answer more than the building fabric does. Modelled physically rather than from a rule of thumb — standing loss through the cover against Ayrshire monthly air temperature, plus the cover‑off loss and pump energy of each bathing session:
| Component | kWh/yr | Note |
|---|---|---|
| Standing loss through the cover | 2,289 | 7.3 kWh/day in January, 5.0 in July |
| Bathing sessions | 1,497 | cover off, jets running, ~1.4 sessions per occupied day |
| Total | 3,786 | about £989 a year at 26.1 p |
For scale: the entire rest of the property uses 2,756 kWh of electricity a year. The tub more than doubles it, to 6,542 kWh.
Standing loss is the dominant term, and it is entirely a function of cover quality. A waterlogged or ill‑fitting cover roughly doubles it — call it £600 a year thrown away. Buy the best rigid insulated cover available, budget to replace it every three to four years, and meter the tub circuit so a failing cover shows up as a rising baseline instead of an unexplained bill.
Everything that made storage and diversion look pointless was a lack of load. The tub supplies that load, and it is a thermally buffered one — it can be heated whenever energy is cheapest and coast for hours, which is exactly what a solar array wants alongside it.
| Configuration | Gen | Import | Export | Self-use | Run cost | vs unmanaged | Capex | Payback |
|---|---|---|---|---|---|---|---|---|
| Tub, no PV, unmanaged | — | 6,542 | — | — | £2,993 | — | — | — |
| Tub + 6 kWp, unmanaged | 5,153 | 4,244 | 2,854 | 45% | £1,965 | £1,028 | £6,900 | 6.7 y |
| Tub + 8 kWp, unmanaged | 6,870 | 4,123 | 4,450 | 35% | £1,694 | £1,299 | £9,200 | 7.1 y |
| Tub + 8 kWp, solar-led heating | 6,870 | 2,569 | 2,904 | 58% | £1,520 | £1,473 | £9,200 | 6.2 y |
| Tub + 8 kWp + 10 kWh, solar-led | 6,870 | 1,567 | 1,720 | 75% | £1,436 | £1,557 | £13,700 | 8.8 y |
| Tub + 8 kWp + 10 kWh, solar + time-of-use | 6,870 | 4,144 | 4,243 | 38% | £959 | £2,034 | £13,700 | 6.7 y |
| Tub + 10 kWp + 15 kWh, solar + ToU | 8,588 | 4,037 | 5,851 | 32% | £692 | £2,301 | £16,950 | 7.4 y |
Same 8 kWp array, same tub: heating it on solar rather than on a dumb thermostat saves £174 a year for no capital at all. That is the best return on the whole project, and it is a software setting.
The battery that was a fifty‑year proposition without the tub pays back in 6.7 years with it, on a time‑of‑use tariff — before counting anything for outage resilience.
The cheapest configuration has the worst self‑consumption, 38 %, because the battery fills from the 7 p night rate and the array exports at 15 p. Optimise the bill, not the dashboard metric.
Reheating from cold takes 12–24 hours and costs more than holding temperature. Lock the setpoint out of guest reach and manage it centrally — a tub that is cold on arrival is a refund.
A tub provided to paying guests falls under HSE guidance HSG282, and the regime is far more onerous than domestic owners expect: a written control scheme, daily recorded checks of chlorine or bromine, pH and temperature, a filter cleaning regime, and scheduled drain‑downs. It must be named in the property's legionella risk assessment and disclosed to the insurer.
The saving grace is that you are already building the system that makes this cheap to comply with — the cleaner's changeover checklist becomes the daily log, with temperature read automatically and the chemical readings typed in once. Keep the records; they are the defence if anyone ever gets ill.
Failford is a rural hamlet on a river gorge, on overhead distribution, with trees. Outages here will be longer and more frequent than the national average, and they will cluster in exactly the winter storms when the property most needs to stay warm. That changes the battery question completely — and it changes the recommendation.
An outage costs a let more than it costs a house. Guests in residence get no light, no heat (a gas boiler still needs mains for its pump and controls), no hot water once the cylinder cools, and no Wi‑Fi. That is a refund and a review. Guests arriving find a dark, cold building. And an empty property in a January outage has no frost protection at all, which is the one that ends in a burst pipe and an insurance claim.
Not everything — that is the mistake that makes backup unaffordable. Oven, hob, shower, immersion and any future EV point stay on the normal board and go dark. What is worth keeping alive draws very little:
| Backed-up circuit | Load |
|---|---|
| Comms cabinet — router, switch, controller, one AP | 70 W |
| Boiler, pump and heating controls | 210 W |
| Fridge / freezer, cycling average | 45 W |
| Lighting, LED throughout | 130 W |
| One kitchen socket circuit | 120 W |
| Diversified continuous | 414 W · 9.9 kWh/day |
| Battery | Alone | +8 kWp, January | +8 kWp, March | +8 kWp, July |
|---|---|---|---|---|
| 5 kWh | 11 h | 10 h | 10 h | indefinite |
| 10 kWh | 22 h | 29 h | indefinite | indefinite |
| 15 kWh | 33 h | 39 h | indefinite | indefinite |
| 20 kWh | 43 h | 58 h | indefinite | indefinite |
From dusk, essential loads only. Note how little January sun contributes — 8 kWp buys just seven extra hours in midwinter. From March to October the array carries the essential load indefinitely and the outage simply stops mattering.
A 10 kWh battery with the switchgear to island properly is about £5,400 installed, or £450 a year over a twelve‑year life. Solar self‑consumption returns £91 of that and a time‑of‑use tariff another £229, so resilience has to be worth £130 a year for the whole thing to wash its face. At typical four‑bed rates that is less than one avoided lost night per year — before counting a single burst pipe.
Fit 8–10 kWp on the SSE pitch and a 10 kWh battery with EPS backup, on a time‑of‑use tariff, with the hot tub heated under solar and off‑peak control. With the tub in the picture the battery pays back in 6.7 years on energy alone, and the outage resilience comes free on top of that.