A propensity-to-sell model analyses property data, historical sales records, and local market conditions to predict which homeowners are most likely to list their property. For estate agents, this changes the economics of prospecting entirely. Instead of blanket-mailing an entire postcode and hoping for a response, agents can target the households most likely to instruct, reducing waste and increasing conversion rates.
Prospecting has always been a numbers game for estate agents. Print enough letters, knock on enough doors, and some percentage of homeowners will respond. The problem is that "some percentage" has historically been very small, and the cost of reaching the majority who aren't interested is getting more expensive. Every letter sent to a homeowner with no intention of moving is a waste of your budget.
Propensity-to-sell modelling changes that equation by identifying which homeowners are statistically more likely to list in the near future, before they've contacted any agent.
What data goes into a propensity-to-sell model?
Spectre AI draws on over 1 billion data points across the 29 million residential properties in the UK. The model combines property-level information with historical sales data and market trends to produce a prediction for every property, reflecting the local market conditions right now as well as longer-term patterns.
Ownership length is one of the strongest signals. A homeowner who purchased eight years ago is statistically more likely to sell than someone who moved in eighteen months ago. But the model goes well beyond a single data point.
Hyper-local market conditions are one of the largest factors influencing the probability that any given property will instruct soon. Has the property next door just listed? Are local prices increasing faster than the wider market? Has the property been listed and withdrawn recently? Spectre AI considers all of these factors and many more to generate a highly personalised prediction for each household.
Crucially, the predictions are continually trained on current market data and validated against what actually happened. This ongoing validation maximises accuracy and means the model reflects real market behaviour, not theoretical assumptions.
How does predictive targeting change prospecting economics?
Traditional prospecting in estate agency follows a broadcast model. An agency selects a postcode, prints several thousand letters, and sends them to every residential address. Response rates on untargeted direct mail typically sit below 1 per cent. That means for every thousand letters sent, fewer than ten homeowners respond, and only a fraction of those will result in a valuation appointment or instruction.
Predictive targeting inverts this model. Instead of mailing a thousand homes and hoping ten respond, an agent identifies the households with the highest propensity to sell and directs the budget toward reaching them. The mailing volume drops, the cost per contact drops, but the response rate climbs because every piece of communication reaches someone who is more likely to be interested.
This doesn't just reduce printing and postage costs. It concentrates your team's follow-up time on the most promising leads. When a response comes in from a high-propensity household, the negotiator handling that call knows they're speaking with someone the data suggests is genuinely considering a move. The entire conversation changes.
Spectre AI's own testing shows that adding its predictions to campaign targeting criteria increases ROI by an average of 310%.
How does Spectre Sales use propensity data in campaigns?
The Spectre Sales campaign builder gives agents a wide range of filters to target their direct mail campaigns: estimated property value, time since last sale, property type, location, and many more. Agents use these filters to build a campaign audience based on their local knowledge and commercial priorities.
Once the campaign is built, agents can then apply the Spectre AI model to further refine the recipient list, removing the properties least likely to instruct and keeping those with the highest probability of coming to market. The agent stays in full control of which properties to prospect. Spectre AI's predictions are recommendations that sharpen an already-targeted campaign, not a replacement for local expertise.
This two-step approach is a genuine differentiator. Agents have the granular filtering tools to target exactly the kind of property they want, and can then layer on AI predictions to cut out wasted spend and focus on the doors most worth knocking on.
Spectre Sales also automates 20:20 campaigns for every property an agent has sold, maintaining a consistent presence around past instructions.
Separately, the system includes anti-embarrassment features that are easy to overlook but commercially important. It automatically excludes your agency's current clients, recently completed transactions, and properties already on the market with your firm.
This prevents the awkward scenario of a vendor receiving a prospecting letter from the agent who's already selling their home, or a past client getting a cold approach weeks after you completed their sale. These exclusions run automatically on every campaign, so agents don't need to manually cross-reference their mailing lists against their CRM.
Do propensity models actually work in practice?
Scepticism about predictive models in property is understandable. Estate agency is a relationship business, and the idea that an algorithm can predict human behaviour feels counterintuitive to many experienced agents. The data, however, is difficult to dismiss.
The results a national agency saw with Spectre Sales provide a clear case study at scale. Their targeted letter campaigns, built using propensity data and the Spectre Sales campaign builder, sent 124,527 letters. Those letters generated 167 instructions with a combined listing value of £345 million. The return on investment was 23x the campaign spend.
It's worth noting that propensity scoring doesn't claim to predict individual behaviour with certainty. No model can tell you that a specific homeowner at a specific address will definitely sell in the next three months.
What it does is identify patterns across large datasets that make some households significantly more likely to transact than others. When applied across hundreds or thousands of prospects, those statistical advantages translate directly into more instructions and better ROI.
How does AI prospecting fit into a wider marketing strategy?
Propensity-to-sell models are most powerful when they're part of a connected marketing approach rather than a standalone tool. When Spectre Sales sits alongside Street.co.uk CRM and the wider Spectre marketing suite, the tools share the same data layer, and follow-up communications can be coordinated across channels.
A homeowner who receives a targeted letter and submits a lead on your website, such as an instant valuation request, flows into Street CRM and can be automatically enrolled in a Spectre Email nurture journey with a website lead trigger.
A prospect who doesn't respond to the first letter but whose propensity score increases over the following months can be re-targeted at the optimal moment. This kind of coordinated, data-informed approach is only possible when your prospecting tools, CRM, and marketing platforms share the same data layer.
The shift from broadcast to targeted prospecting is already well underway in estate agency.
Agents who adopt propensity-to-sell models gain a measurable advantage in instruction conversion, and as these tools become more widely available, the agents who don't adopt them will find their untargeted campaigns competing for attention with highly personalised, well-timed communications from competitors who know exactly which doors to knock on.
