Postmortem of a British Startup: Tract

Between May 2023 and March 2025, Tract attempted to build a venture-backed company to address Britain’s housing crisis by improving the planning permission process. After raising a £744,000 pre-seed round in April 2024, we explored several business models: a site-sourcing tool for developers (Tract Source), a free land-appraisal tool for landowners (Attract), becoming tech-enabled land promoters ourselves, and finally, an AI-powered platform to assist with drafting planning documents (Tract Editor).

Despite significant technical progress, building tools like Scout and the well-received Tract Editor, our journey taught us critical lessons from failing to secure a viable, venture-scale business model in the British property market. We learned the difficulty of selling software into a conservative sector, experienced the operational complexities and timelines of land promotion, and encountered a low willingness to pay for useful tools. Furthermore, we came to understand how the market's conservatism and fragmentation limited its potential for venture-backed disruption. After nearly two years without revenue or committed paying customers, we realised we lacked a clear path to the necessary scale and returns. This prompted the decision to cease operations and return capital, sharing our experience as a case study.

In retrospect, it's easy to see ways we could have approached things differently. This document is a post mortem, explaining what happened and why it went wrong. Our aim in writing this is:

We want to stress from the start that the ultimate failure of the company lies with us. Some issues were within our control and some were beyond it. At times, we’ll describe external factors in ways that sound negative. This isn’t an attempt to pass blame. We want to tell the story as matter-of-factly as possible. Most importantly, we’re extremely grateful to everyone who supported us over the last couple of years, in money and in time.

Jamie Rumbelow and Henry Dashwood April 2025, London

This document is long; this table of contents also functions as a bullet-point summary. The Advice for Founders section contains the main lessons we hope people take from our experience, so feel free to only read that and skip the rest.

You can also summarise this page with ChatGPT .

Many young people have had to delay forming families and often take poorly paid, insecure jobs that can barely cover rent and living costs as the price for living in culturally attractive cities. They see opportunity as limited and growth as barely perceptible. Meanwhile, older generations sit on housing property worth many times what they paid and, stuck in a zero-sum mindset, often prioritise the protection of their own neighbourhoods over the need to build more homes. Can you blame young people who resent older people, and the West’s economic system itself, when this is what it offers them?

Myers, Bowman and Southwood, The Housing Theory of Everything

Housing in Britain is expensive because of artificial supply constraints - specifically, the difficulty of obtaining planning permission to build more. Planning permission is Britain’s regulatory approval that landowners must secure before developing land or changing a building's use. When permission is granted to build homes on agricultural land, its value increases dramatically (often 140x or more), creating enormous wealth for landowners.

This looked like a business opportunity: if we could help more sites get planning permission, we could capture some of that uplift. The market size looked compelling - billions in Britain alone, the political winds seemed favourable, and there seemed to be few modern software solutions.

We were driven by a mix of moral outrage and technological optimism. It seemed absurd that bureaucratic obstacles were holding back so much potential value, especially when millions were being spent on documentation that could be automated. With the rise of LLMs, we believed we could transform this inefficient process.

Strategic land teams are groups within property development companies responsible for identifying and acquiring land with long-term development potential. They focus on finding sites that don’t yet have permission for development but possess characteristics that suggest it could be obtained in the future. The earlier they can identify and secure control of these sites, the greater their competitive advantage and profit potential.

This felt like a software-shaped problem and could form the basis of a larger platform for planning, moving elsewhere along the value chain. When Jamie began working on Tract, this site sourcing problem was the first thing he looked at.

Some sourcing tools existed, but the popular options seemed poorly built: badly designed, built atop untrustworthy data sources, and not built around the sourcing workflow.

Jamie had ideas to improve the product experience and began building them out. He started to lean on industry advisors for introductions. He read The Mom Test (more on which later) and had dozens of calls with whoever would take them.

During these calls, he realised his intuitions about the existing products’ problems were correct (“I don’t trust it”, “I need to check the numbers, it’s annoying”). This was exciting: he thought he was onto something.

But he didn’t ask why users tolerated it (“It’s fine”, “Does what I need”, “We went with [competitor] because, to be honest, they’re nice guys”). It was unlikely whatever he built could be 10-100x better at solving this problem without radically changing their workflow, which was tough for a solo developer working on savings – Henry began helping in his free time from late summer but wouldn’t come on full time until mid-December.

Jamie also began to find more competitors, some well-resourced, all of whom appeared to be competing on price. These were already cheap contracts: a few hundred pounds per month for a small/medium-sized team. And they were getting cheaper.

We concluded that this was a difficult market to sell into, with race-to-the-bottom pricing dynamics, and no obvious way to make a SaaS tool that helped strategic land teams do site selection so well that these factors wouldn’t matter. It was time to pivot.

However, it took too long to reach the pivot point. A more thorough competitor analysis would have revealed the market problems earlier, rather than being distracted by writing code. None of what we needed to learn required writing code to get there.

The biggest lesson here: getting time-to-validation as low as possible matters more than anything else . If there are ways faster to get there, you should take them.

Our site appraisal tool, Attract, emerged from discussions with our design partner Paul and from Tract Source's challenges. The concept was simple but promising: instead of selling land information to developers, we’d provide it to landowners for free. The greatest constraint for any promoter or developer is access to developable land, so we imagined a tool encouraging landowners to reveal their openness to development would be valuable. The generated appraisals would allow us to rapidly qualify sites and identify potential opportunities.

Initially, we considered selling this tool to developers and land promoters seeking strategic land opportunities. However, we quickly recognised two fatal flaws in this approach:

Despite this, Jamie completed the conversion work and launched a white-labelled version on Paul’s company’s website in early November. It delivered impressive results: he received more submissions and higher-quality leads. This quick win validated our technical approach, but the underlying business model remained insufficient to build our company.

This experience taught us something valuable: we could create genuinely useful tools that solved real problems. However, we still hadn't cracked how to transform that utility into a venture-scale business model.

With the white-labelled version of Attract online, we entered a period of exploration and uncertainty. We had a call with a major surveyor where we showed them a mocked-up tool for writing a planning application using an LLM and the information about a site in our database. But after the demo, we were ghosted.

Market concerns loomed large. The pricing dynamics made us question our ability to raise capital with our current approach. In retrospect, we may have been closer to viable products than we realised. Multiple British prop-tech companies have been funded, including in the US, during the lifetime, suggesting alternative paths we could have taken.

Despite challenges, we continued experimenting. In December, we created a ‘Tinder for Buildings’ demo with inpainting models that received positive feedback. We continued scoping ways to build a business on our planning data work. We were awarded an Emergent Ventures grant , validating our mission if not our approach.

At this stage, we were clearly concerned about the market, but I don’t think we did enough – now or indeed later – to precisify those concerns. We gave up on the platform idea because of a couple of crappy meetings, rather than work out how to pitch better. It feels like we weren’t trying hard enough to be an organisation that learned from its mistakes.

In January, we began fundraising while working on extracting Local Planning Authority validation checklists—technical work to support our evolving vision.

A February visit to a developer proved illuminating, though not as hoped. Their skeptical in-house planner admitted: "To be honest, the system being broken helps us." This comment crystallised a tension in our market: many established players benefited from the inefficiencies we aimed to solve.

This realisation prompted us to revisit our fundamental problem statement. The facts were compelling: land with planning permission becomes dramatically more valuable, and this value creation stems from the friction and uncertainty in the planning process—friction that good software could reduce.

We faced a dilemma. Selling software into this market would be tough, and it seemed wasteful to capture only a small slice of the value we could create. If we could facilitate 100x value increases in land, why sell this capability to others for modest SaaS fees? What B2B model could compete with capturing that uplift directly?

We considered a radical pivot: becoming land promoters ourselves. We’d partner with landowners or acquire land, secure planning permission using our technology, and sell with permission granted—capturing the value uplift directly.

We found another opportunity: promotion costs don't scale linearly with site size, making many smaller sites economically unviable for traditional promoters. We could counterposition by targeting these ignored sites and sell them to SME developers.

This model offered significant advantages:

Implementation would be challenging—some aspects like site visits couldn't be fully automated—but we believed we could create a durable advantage by controlling the critical fulcrum point where planning permission is granted.

During the fundraising process however, we identified a bottleneck: access to land. Without reliable land access, our growth would be constrained. How would we:

One investor – who ultimately passed – described this challenge as requiring extensive "hand-cranking"—an apt analogy. But at the time, we thought Attract could solve our top of funnel problem.

Rather than white-labelling the tool for strategic land teams, we could own it, market aggressively to landowners, keep the sites we want to develop ourselves and pass the rest to strategic land teams for a referral fee. Appraising land and advising on its development potential is part of what a land agent does. We would be automating that.

Why would a landowner use our tool? Partly out of curiosity like taking personality quizzes, but also because our information was genuinely useful.