In the industry there is often criticism about the supposed accuracy  of ‘instant valuation’ tools as found on property portals.

Now a proptech company is promising that its new method of AVM, which adds computerised analysis of property photographs to their proprietary date and the more usual mix of data sources, will provide valuations that are 95% as accurate as those done in-person.

It’s a very bold claim and as yet we have not seen the evidence to support it, but if it really can deliver on its promise then could the days of the Valuer in estate agency be numbered?

No doubt EYE readers will have a view on that, as they also will about whether 95% is close enough to 100%.

 

Houzen’s Brand Director, Zuzanna Chmielewska, explains the background to their new product:

Property valuations, as a service, have offered pretty much the same thing over the last 10 years. The “run instant valuation” button is found on most estate agency websites in the UK and all it takes is a postcode, a click of a button and voila: you’ve got a quite generic, not very helpful, valuation figure.

However, the websites always claim that “an on-site visit by a professional is required for a specific valuation” and after you’ve completed your online instant valuation you’ll receive a report via email and a local agent will get in touch to arrange an appointment. This involves an allegedly accredited and qualified property expert coming to your home to give a more detailed and accurate valuation. An agent will consider the most recent market dynamics and property rates, also taking into account the unique features of your home as well as advice on how to potentially heighten its value.

But in 2021, having just lived through the times of social distancing and remote working, can we make property valuations work as they are supposed to work – online and on point? Indeed, with most products and services undergoing their digital improvements nowadays, why can’t we get a better way to value our homes online?

An Automated Valuation Model (AVM) is the name given to a service that can provide real estate property valuations using mathematical modelling combined with a database. Most AVMs calculate a property’s value at a specific point in time by analyzing values of comparable properties. Some also take into account current asking prices, previous surveyor valuations, historical house price movements and user inputs (e.g. number of bedrooms, property improvements, etc.).

That’s the Wikipedia summary but in plain terms, it’s your usual “instant property valuation”. It’s a report of the average market price for homes that are similar to yours, within your local area. The information is based on data from HM Land Registry and recent property sales. Most widely used AVMs in the UK are produced by Hometrack, a Zoopla company.

Why aren’t instant valuations or AVMs very accurate?

Simply put – because the majority of instant valuation tools use only publicly available historical data (e.g. HM Land Registry, Registers of Scotland, Royal Mail and Ordnance Survey) to create valuations. Zoopla says their estimations are based on region, property type and asking prices of similar properties in your local area.

This sounds like a decent idea, but in reality, property prices can be determined by hundreds of factors: from current buyer and tenant demand on the market, budgets, number of shops in an area, street safety, sustainability improvements in the building etc. Each of these factors could theoretically be assigned a value and measured but so far it has proven to be quite a challenging task to perform by any tool available in the market right now.

Many AVMs are also using transactional data, which may lag anywhere from three to six months, although surveyors are similarly restricted in terms of data recency. Transactional data is a good data source but still does not account for changes in current market conditions.

On top of this, automated valuations can only assess the data they have access to. Which means the stats are limited to short and limited information available on Zoopla or a marketplace listing as well as official documents. They can’t take into consideration the decor, view, parking, quality of exterior etc.

Aware of all these issues, the tech-driven team at London-based startup houzen decided to tackle this issue.

Saurabh Saxena, Founder of houzen said: “With social distancing measures and switching to work-from-home, it was obvious that the classic visual inspection done by an estate agent is becoming more and more inconvenient. With every other service being moved online, it was time to deliver a data-driven tool for online property valuations, which could deliver results as specific as an in-person visit.”

In building the cutting-edge tool, houzen used hundreds of data points to assess each property’s value and took not only public historical data but also their own proprietary data, gathered from agents and buyers over the past 4 years.

The company’s tech team, accompanied by data scientists and analysts, worked together to deliver a tool which promises to deliver valuations 95% as accurate as those done in-person, 100% online and in a matter of seconds.

houzen’s tool uses Computer Vision – a new type of technology which allows the computer to “see” and assess pictures.

So even if no additional explanation is provided, the computer knows for example that it’s looking at a medium-sized bedroom with 1 large wooden wardrobe, simple Scandinavian bed that looks like new, soft rug on the floor – and also, there’s a crack on the wall so it would drive the price a little lower.

Assumptions like these would traditionally only be done by humans – but not anymore.