unthinkable
Case study · A London-based fintech (unnamed in the source) · digital valuation of UK residential property

UK lenders and estate agents get a property's value in moments, with its neighbours and its market alongside.

The company wanted lenders, financial institutions and estate agents to value UK residential property quickly and to capital markets standards, without paperwork, inspections and manual data entry.

London, United KingdomDigital lending & loan management

What changed

93%valuation accuracy (as reported)
70M+training records (as reported)
7M+missing values filled (as reported)
100+model experiments run (as reported)
What stood in the way

What stood in the way.

What stood in the way
  • Valuations relied on paperwork, physical inspections and manual data entry, which caused delays and errors and could not scale.
  • The company had the data and the market knowledge, but not the technical infrastructure to automate valuation.
What changed
  • 93%valuation accuracy (as reported)
  • 70M+training records (as reported)
  • 7M+missing values filled (as reported)
  • 100+model experiments run (as reported)
What we built

What we built.

  • Data from more than 10 sources, including listings, transaction histories, leases and market trends, cleaned into about 70 million training instances.
  • Imputation that filled more than 7 million missing values, and outlier detection across over 3 million data points.
  • A valuation model chosen from over 100 experiments, from regression to deep learning, built with Keras, TensorFlow and Python and tested on location, neighbourhood and market conditions.
  • Market, Property Intelligence and Neighbourhood features for agents' performance, lenders' risk appetite and comparison with nearby homes.
  • Cloud deployment integrated with the client's systems, with automated monitoring and alerts.
  • Python
  • Keras
  • TensorFlow
  • Cloud deployment
What it shows

What this work shows about us.

We know what a lender needs from a valuation

Price, stock and liquidity trends sit next to the number, so a lender can check it against its risk appetite.

Data work before model work

Most of the effort went into joining sources, filling gaps and removing outliers, which is what made the model accurate.

Want to build a similar automated property valuation model? Tell us what you want to change. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with the right team.

Start your plan ↗
If we built it today

If we built it today, with AI.

What we would add now, moment by moment.

The decision

An agent would prepare the valuation pack for a mortgage case with the model's figure, comparable sales and the reasons it differs from the asking price. The lender's underwriter still decides.

Documents

An agent would read the lease and title documents for a property and flag anything that changes its value, such as a short lease. A valuer checks every flagged case.

Ask Unthinkable

Want this for your business?

Tell us what you run or what you're building. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with a senior AI engineer.

Send us a note instead

Required

Thanks. A senior AI engineer will reply to you.
u
Ask UnthinkableClarity first, then the right team
Your conversation

An AI agent. This prototype uses prepared answers and sends nothing.