With the growing presence of artificial intelligence, is it surprising that tech may now be capable of undertaking that fundamental and most sacred of skills in real estate – the opinion of market value? Are the machines, in the form of automated valuation models, taking over? Will the algorithm builders usurp the opinion of the valuer?
Whatever the answer, the valuation profession must adapt, accept that technology in its many forms will increasingly determine how we value, and embrace all that AVMs can offer, albeit with some health warnings.
An AVM is essentially an algorithm that can estimate the value of real estate based on a huge range of data points. In some respects, valuers are already using AVMs in the form of proprietary software produced by the likes of Argus, MRI and KEL, which rely on the input of data.
They support a human valuer’s decision, helping to justify why a building is “better” or “poorer” than another without over-reliance on human sentiment and perception. However, the trend towards more flexible and shorter leasing models and the near demise of the 25-year lease with geared uplifts dependent on various inflation indices, and the demand for more complex sensitivity analysis will require more complicated and data-rich modelling.
The algorithms associated with AVMs depend on the quality, quantum and diversity of data available. As new and refined data sets become available, valuations could well be generated as frequently as daily, and indeed, at the press of a button. And in a market starved of real-time evidence, notably for shopping centres and other retail sectors, would AVMs be able to pinpoint more accurately than the human valuer, an opinion of value?
There are clearly many shortfalls still to be overcome. For example, who owns the data which is incorporated into the algorithms? And if it goes wrong was the faulty algorithm, or the valuer to blame? And what if we assimilate a valuation via an algorithm with limited data, when upon incorporation of full data, or after a full inspection, a significantly different valuation is produced?
Consider, also, the requirement for institutional as well as regulatory acceptance over time, and trust by purchasers’ banks. How will the courts view such modelling? Who will regulate the data engineers and the constructors of the algorithms? Akin to “driverless” cars, AVMs will need operators and, currently, the majority still require a driver to operate safely. And what will happen when the algorithm produces a valuation outside the required “value range”? Will an alarm sound? Probably not.
Valuation is an art not a science. But data-driven automated models are a powerful new tool we should embrace, albeit with a healthy dose of human will and empathy
As new valuers are inducted into the profession, we must not lose institutional knowledge of the softer drivers of pricing and valuation. How will the machine put itself in the shoes of a potential purchaser, or of a bank which may potentially lend to a particular borrower for the acquisition of a particular asset, but not to a different borrower or asset?
Machines for empowerment
Valuers must stay engaged and not remain in isolation. They must embrace and explore all that AVMs can offer to move the profession forward.
With greater transparency and more sophisticated models and data available, the entire industry will have to become better at understanding values and explaining what they mean. Human nature is such that clients will always want a human there to advise and offer a second opinion on the data.
I am optimistic for the role of the human valuer. Having debated the subject with peers across lending, development, asset management and tech, I have concluded that automated models are not only a gift to the valuer, but an exciting new way for both landlords and occupiers to understand and make decisions about their space.
Automated models will empower agents to spend more time advising clients, while providing more information than ever before to determine true values. This is something that has become even more important over the past year as structural changes in the UK real estate sector underline the importance of accurate, regular valuations.
Valuation is an art not a science. But data-driven automated models are a powerful new tool we should embrace, albeit with a healthy dose of human will and empathy.
Harry Morten is a partner in the valuation & advisory services division of Knight Frank