Property listing data that survives contact with reality
Property data looks structured until you try to use it. The same house appears on three portals under two agents with different photographs, a price that changed last week, and a description that describes a different number of bedrooms. Anyone building valuation models, investment screens, or market reports discovers quickly that the collection was the easy part and the reconciliation is the product.
- Starting at
- from £1,200
- Refresh
- Daily, with hourly checks on a watchlist
- Access
- public web
What comes out of it
- Address, coordinates, and property type
- Asking price with change history and dates
- Bedrooms, bathrooms, floor area, and tenure
- Listing and withdrawal dates for time on market
- Agent details and listing source
What makes Property listing portals awkward
Deduplication has no reliable identifier to work from. Addresses are written inconsistently and sometimes deliberately vaguely, coordinates are often approximate to protect privacy, and the same property is photographed differently by each agent. Matching combines normalised address, geographic proximity, and attribute similarity, with image hashing as a tiebreaker when a property is listed by two agents with the same photographs.
Price history is the valuable signal and portals actively obscure it. A listing that is withdrawn and relisted at a lower price often shows as a new listing with no history, which hides exactly the motivated seller a buyer wants to find. Detecting relisting means fingerprinting properties by address, coordinates, and photograph similarity, then stitching the episodes into one timeline.
What the data supports
Market analysis is the most defensible use: asking prices, time on market, and price reductions across an area over time, which is a useful picture that no single portal presents. Aggregate trends are also less sensitive to individual matching errors, so the analysis holds up against the messiness described above.
Investment screening is the second use, filtering for properties matching yield criteria, reduction patterns, or time-on-market thresholds. This one depends heavily on the relisting detection, since a property quietly relisted twice is the signal you most want and the one most easily missed.
Valuation model inputs are the third, where the requirement is completeness and consistency over time rather than real-time freshness, which changes the design towards careful historical collection rather than aggressive polling.
Building the property timeline
Each observed listing is fingerprinted from address, coordinates, and core attributes, then matched against existing property records. Where a match is confident the listing becomes another episode in that property's timeline. Where it is uncertain it goes to review rather than being merged.
Price changes are recorded as events rather than overwriting a field, which is what makes reduction analysis possible later. A property that has been on the market for six months across two agents with two reductions reads very differently from a fresh listing at the same price, and only an event history shows you that.
Withdrawal and relisting are inferred from disappearance and reappearance within a window, with the timeline stitched together so time-on-market reflects reality rather than the portal's presentation of it.
Cost and the sourcing conversation
Workflow automation starts from £1,200 for a bounded property-data extraction from one permitted source into an agreed dataset. Scheduled refresh, cross-source deduplication, price history and reporting are scoped separately.
Before either, there is a conversation about sourcing that I would rather have early. If your business depends on data from a portal whose terms forbid collection, you are building on ground someone else can remove. Sometimes the right answer is a commercial licence, sometimes it is restricting scope to agent sites and public registers, and occasionally it is that the project is not viable as imagined.
Portals change layout often and several deploy escalating bot protection, so maintenance from £400 a month is realistic for anything running continuously.
Where the line is
Publicly accessible pages, robots.txt and terms respected.
Major portals generally forbid automated collection in their terms, and I will tell you that plainly rather than quietly building it anyway. Where a portal offers a commercial data licence, that is the route I recommend even though it costs more, because a business built on a data source that can be cut off is a business with a single point of failure someone else controls. Agent websites and public records such as sold price registers sit on better footing. Listings also contain agent contact details, which is personal data with the usual obligations.
MathersLabs is not affiliated with or endorsed by Property listing portals. Work is scoped to publicly accessible data and to compliance with applicable terms and law, and feasibility is assessed per engagement before any fixed price is agreed.
Related work
SignalSauce
250 qualified leads a month, fully automated
Built an end-to-end intelligence pipeline that ingests from 5 curated news sources twice daily, scores every announcement against a custom algorithm, and enriches the ones that pass with 21 structured data points per company.
Questions people ask first
Is your source on this list?
Tell me what data you need and where it lives. I will tell you whether it is feasible, what it would take, and whether it is worth building before you commit to anything. Working hours overlap with US Eastern and Pacific.