Automated valuations for homes: Trustworthy or Random Number Generators

I sold my home last year. The UK Land Registry has made that price available as part of its Open Data initiative. Tools like Mouseprice and Zoopla suck this in to give the next person an idea of what this property is worth today.

But is it worth what they say? The property in question was sold for £248k in October 2013 and was previously rented out for £1,300pcm.

Zoopla says this is now (June 2014) worth £261k and the rental value would be £808pcm.

Mouseprice puts the current value at £435k.

I had a local agent value this as well. They said for a quick sale they would price at £295k and take offers at £285k.

This isn’t a slight toward Mouseprice and Zoopla, but surely they need to take a higher duty of care when presenting these random numbers to the world.

It probably isn’t their fault; the meta data on ‘marketed’ property and ‘sold’ prices is neither linked nor accurate. In the same area, you could have a 4 bed property that is a house, a luxury apartment or an ex-council flat. If all are the same size (measured in square feet or square metres), then their values would be substancially different. Online tools just cannot ‘see’ this tangible difference.

The FindProperly ‘Cool Things’ certainly deserves a mention and is a step in the right direction: http://www.findproperly.co.uk/promo.php#.U6wDlPldXDY

Martin Lewis’s MoneySavingExpert probably need to caveat their consumer advice on home valuations to include a warning on the accuracy of online valuation tools: http://www.moneysavingexpert.com/mortgages/free-house-prices

If you are building a valuation tool, marketed prices would be a better place to start, as humans are making irrational decisions when coming up with numbers. After all, it is humans that make offers on homes; and they make those decisions on feeling rather than hard, cold numbers. Invariably, computer says yes, house buyer says no.