Homely Energy

Homely makes heat pump control. A hub in the house talks to the heat pump, and an optimisation engine decides when it runs. On top of them sit a homeowner app, an installer app and Connect, the installer's remote portal.

I own the customer and installer journeys, and I am responsible for growth through product. That covers the homeowner app, the installer app and Connect, and beyond them the manufacturer integrations, Connect's pricing and billing, the installer manual, the packaging and the website.

I write the tickets and work the interactions through with the developers and the designer. I decide the order things get done in. I took the integration launch process I inherited, wrote it down and ran it three times. I've worked both live markets and most of it has gone into the UK.

Connect existed as foundations when I arrived. It was nowhere near market ready, authorisation had not been built, and most of the decisions about what it should be were still open. I specified the paid tier, set the pricing and sold it myself, company by company. Selling it by hand was deliberate, because I wanted to know what signing up actually cost an installer in time before I specified the tooling that would take that time out. Nobody cancelled in the fifty two weeks the paid tier ran, on annual terms with no renewal cycle completed.

Eleven months in I took the paid tier off and folded its price into the product instead. Two things had pointed the same way. Serving the full product to every installed hub cost less than running the paywall that kept people out, and the charge was holding back adoption of the whole product more than it was earning.

Connect now comes with every hub for two years, matched to the UK workmanship warranty, and everyone who had already paid got two extra years. The change moved the cost from the installer to the homeowner, as part of the installation. That took a decision out of the sale, and it gave installers a reason to fit Connect in the first place.

It's now one of the reasons an installer picks the hardware in the first place. A customer rings with a problem and, instead of booking a van, the installer opens Connect and looks. In June 2026 Homely won Ancillary Product of the Year at the National ACR & Heat Pump Awards, on a submission built around Connect.

Installer support tickets per installation fell by 55% between 2024 and 2025, measured against a separate installations pipeline. The installer app flow and the handover to the homeowner were the first things I owned, and tickets about that handover fell 63% year on year by March 2025.

The results came from small, iterative improvements rather than one big fix. Most of them started in research. Over a hundred installers and customers have been through seven research methods with me since 2024, from structured interviews to point of sale walkthroughs to a paid beta.

How the company decides what to build, who gets asked, and what has to be true before anything starts. I proposed it in November 2025 and delivered it that December as four guides written for four different audiences, each with a recorded video. It's still what the company runs on, and it's the piece of work here I'd most want to be judged on.

In the early product, an installer who set something wrong on site had to drive back to the house to change it. The remote route arrived in stages. Connect came online first, then an advanced data page, then remote parameter changes, and flow temperature last, because flow temperature decides whether the house is warm and the bill is sensible.

I specified the flow temperature work, raised the security tickets underneath it and tested it on three live heat pumps before release. It's been in routine use since January 2026. Every change is confirmed to the homeowner in the app and by email the same day, which I asked for deliberately, because a stranger changing the heating in your house should not be something you find out about later. Someone at a desk can now put right in a minute what used to need a van.

Installers were leaving site with the hub never having connected to the heat pump, and the customer only found out when the house went cold. I paid ten to fifteen installers for an interview and built automatic checks around the three faults they described most often. Those were cable polarity the wrong way round, a hub that never joined the network, and commissioning left half finished. The checks run while the installer is still on site, and when one fails an error code appears on screen with a public troubleshooting guide behind it.

They shipped in November 2025 and I pulled them eleven days later. They were making more work than the fault they caught, and a fix took days to write but months to reach installers, because that's how long an installer app update took to ship. I wasn't willing to run a safety net I couldn't mend at the speed it failed, so I fixed the update time instead. The app was rebuilt on Expo and released in April 2026, which cut an update from months to days, and the checks went back out in full on 21 May after two weeks of clean testing.

I can't tell you the before and after on polarity faults, because they were logged as "there is a problem" and the data won't carry the claim.

The installer manual is the one part of the product every installer reads, and nobody thinks of it as product. I rebuilt it. A shelf of brand specific manuals became one universal manual, organised around the order the job happens in. I split water side commissioning from commissioning by hand, because installers had been running them as one and getting caught out. I rewrote the wiring section around what an installer actually sees on a manufacturer's control board, and redrew the DIP switch illustrations after watching installers misread them.

To write any of that I had to learn how a heat pump works, which isn't the same as knowing what our product does.

The box went the same way. I set the completion criteria, worked through rounds of prototypes with the packaging designers and ran the shake and drop tests myself. The rattle I was chasing turned out to be the manual moving inside the box. The box, the accessory bag and the instructions all changed together, and the box got smaller in the same generation, so the whole manual had to fit a smaller format without losing a step.

I built a 24 month study joining the support, installation and call records around the product. Operational data gets collected to run a business, not to be studied, so the same installer or installation can turn up under different names in different systems. Before anything could be counted I matched those records by hand. The matching took longer than the analysis and it's what made the results worth trusting.

The attractive answer was that support friction was driving installers away. The data didn't support it. Tested installation by installation, with new installers left out of the leaving figures until they had been around long enough to count, no link appeared. I recorded it as unproven rather than disproven, because the study could not rule the effect out, and nothing was built on it. The study was accepted and the recommendations were adopted. What it changed is not mine to publish.

The industry treats two zone control as a requirement for ECO4 funding. It isn't one. Leaving it out costs two SAP points, confirmed independently by two funding providers, which makes it a design trade rather than a rule. Pinning that down took research across three certification bodies and the official database of product performance figures, plus three recorded installer interviews, with ticket numbers logged at two of the bodies so those answers can be traced back. It's the kind of question everyone in the trade answers from memory, and the answer everyone gives is worth about as much as the last person who repeated it. The verdict went against the case for our own feature and I published it anyway.

Two zones is not a good answer to the problem either. A house that shares one thermal envelope will not hold two different temperatures cleanly. What you usually end up with is both zones at temperature and one of them warmer than anyone asked for.

We built two zone control regardless, for a different reason. Large installers list it on their specification sheets and some would not onboard without it, so I took a working prototype to market inside a short window. We set a target of twenty dual zone installs a week and had nine three months in, so I funded a paid customer trial to find out what the gap was made of. The answer was that the requirement had been a box on a specification sheet rather than real demand. The project was co-owned, with colleagues carrying delivery, the customer app and go to market.

HomelyLifetime was a Heat Pump Ready innovation project backed by a £465,991 grant from DESNZ. The bid was won before I joined. I ran it as product owner from mid 2024 to close-out in April 2025, signing off milestones seven and eight and the final reports, and checking the claims were accurate before they went in. Grant work has a discipline ordinary product work doesn't, because the commitments are fixed, public and audited.

The government's published case study records the Connect MVP tested on 175 units against a goal of 50, more than 30 in-depth interviews and a survey of over 60 installers and homeowners behind the design, and initial trials showing roughly 30% fewer emergency visits to the home.

• The HomelyLifetime case study, published on gov.uk
• Heat Pump Ready programme, successful projects on gov.uk

Product design came to me in June 2026, and in July I ran engineering prioritisation directly and set the agenda for company wide product strategy. I built the reporting I run the product on and I own the telemetry dashboards behind it.