This post outlines a couple of the lessons I and my co-founder have learnt over the past years, trying to kick-start Plot.dev.
We conceptualised Plot as platform where developers could scan the country using a number of variables, which would output interesting plots, buildings and addresses. Real estate development is a domain I am familiar with through my studies, and in which I have quite a lot of contacts.
We managed to secure multiple pre-sales on the promise of this platform through cold-emailing, and received multiple non-trivial payments (€1K+) before I started building.
Upon launch however we realised it was way too early: the users opened the platform once or twice and then stopped using it. They did not ask for their money back.
By now we had approached about 50 developers, and we realised it’s a sales-heavy process, requiring multiple meetings, both virtual and physical before securing any type of commitments and payments. Together with the fact that the market of real estate developers is limited (< 5000 companies in the Netherlands, of which 80% are one-man operations), and the concrete lack of enthusiasm, we realised this product did not solve a dire need.
We set out to find that angle.
By now we had built up a data infrastructure already collecting and harmonising millions of datapoints daily, from a couple of relevant data sources. Some of these datapoints are very hard to gather for competitors, and we tried to capitalise on this advantage.
Instead of doing customer development however, we made an informed gamble on a micro-product. We set out to build an MVP for this: a daily newsletter summarising properties that just came onto the market that have redevelopment potential according to a relatively standard industry feasibility calculation.
We focused on this product because it served a wider market (developers, constructors, contractors, real estate agents, ambitious individuals), and had a highly scalable sales pipeline through advertising.
We launched the product and sales pipeline, and rather quickly realised there was growth potential. We’d wake up every day with multiple new subscriptions, and saw the advertising costs per conversion decline. Felt good.
Over time however it also became clear that the pond from which we were fishing was too small, the facebook and instagram ads saturated and our conversion rate dropped drastically. Moreover, we learned from customer feedback that the product had a fundamentally problematic aspect: with every new customer, the value of the newsletter declined. It turned out that the most interesting properties we distributed would be fully subscribed to within minutes after sending our daily newsletter at 12:00.
Nonetheless, we managed to scale the product to €3.9K MRR within 3 months, on a fully automated sales pipeline, which would eventually generate about €38K in total revenue from €15k in adspend. We got a taste of the possibilities of scalable saas products, but did not yet quite find the right product.
By now we were convinced that data can bring significant advantages in this domain and that companies and individuals are very willing to pay, given that the data is:
There were also plenty of companies operating in the real estate data space in the Netherlands, so as a company we needed to offer different data.
Through countless conversations with customers from the newsletters, we learned that the majority were looking for opportunities that allowed them to spatially expand a property. We also had learned by now that the process to determine what exactly was possible in terms of spatial expansion, was a highly manual proces. A manual process, fully based on widely available data.
This was a difficult problem. We had to normalize a non trivial amount of ways of describing spatial limitations into accurate and usable data. These were written in legal jargon. We had to handle multiple digital formats, handle hand-written plans, PDFs. We had to find a way to normalize various spatial dimensions and geometries into one or multiple 3D representations, and make those easily understandable, searchable, viewable for customers.
We ended up hiring two junior machine learning engineers, and a senior machine learning engineer to help us with the natural language aspect of the problem.
We managed to get a grip on the problem, and managed to create a system that was identifying thousands upon thousands of sites with significant spatial possibilities in the Netherlands. In a country such as the Netherlands, very dense, where people assume each site has dozens of eyes on it, constantly, this is no small feat. We’d consistently find sites that, after manual validation, did indeed have the possibility to expand current built volume by 200% to 300%.
Making these sites findable through a UI was a trivial task, and to make the spatial possibilities easily understandable to our customers, we built a small demo environment.
This generated a lot of enthusiasm. So far, so good.
Business model aspects
Then we ran into more challenges again.
Firstly, there is a large disparity between the (potential) value of a lead and the reasonable amount you can charge per lead at that stage. Adding hundreds of square metres of area to a building in the Netherlands is worth hundreds of thousands to millions of Euro’s, while you won’t be able to charge more than a few euro’s per lead.
Customers that identified numerous sites also have to consistently invest time to convert “cold” leads into “warm” leads. With real estate - objects which typically make up a big part of companies / peoples lives - that is not a simple feat. It involves finding the current owner and then reaching him/her at the right moment, with the right offer. The conversion on this process is low, even though our trial processes uncovered up to 20% of reached owners were seriously willing to consider an offer.
To capture a more proportional part of the value we uncovered, and to make an irresistible offer for developers, we decided to adapt to the “no cure no pay” axiom of this particular domain.
However, the value of that initial “finding” is only captured once a transaction materialises - which is the only moment developers will accept actually paying you a portion of the transaction. This is how real estate agents already get paid, and as this process excludes the agent, the idea of paying a commission was tolerable for the developers. But: the time between them having access to your data and you getting paid for your data now stretches from weeks to years.
Moreover, you have to fully trust your customer. They have plenty of (fully legal, of course) options to hide a transaction from your eyes.
In terms of scaling this, it became clear that the average developer / constructor is not actively looking for data solutions. As a high stakes domain, many developers are looking for a competitive edge through data, but they do not know exactly what they are looking for. Moreover, time is extremely sparse for developers, most of the time. In “down” times, they will explore alternative ways of generating leads, such as through data. This dynamic requires any data startup in this space to be initially very focused on outbound sales and relationship building.
We spent roughly 18 months working through the issues mentioned above. Some we managed to work through, but without conclusive commercial successes.
While burning through our own capital and our earlier profits, we realised we would not be able to make a sufficiently advanced tech product to fulfil our promise of consistently finding “all potential redevelopment sites”, in real time, for the whole of the Netherlands.
We had had some small successes, and we decided this might be enough to prove the potential of the idea, and raise funds. We spent the seven months from January to July 2023 on fundraising. That whole process merits a post by itself, but to cut a long story short: we had a lot of interaction with relevant investors, a lot of useful feedback, and a handful of interested parties. However, in the end, we did not manage to forge an investment deal.