Real estate is a data-driven industry. Every investor and stakeholders like banks, developers and insurance companies rely on data to make crucial decisions.
Automated Valuation Models (AVMs), initially hailed as a revolutionary tech solution, are proving challenging in an unpredictable market where two adjacent properties can have drastically different values. What will it really take to not only automate real estate reports and valuations but also make them faster and more accessible?
This week, we take a deep dive into the world of property valuation to understand where we are techwise, besides this, data scientists will also guide us on where we should be.
Who needs accurate valuations?
Alex Moseti, a Valuer and the Director, Cambrian Valuers Ltd, has been in the valuation industry for well over a decade and a half. Through the years, Moseti has witnessed the industry transform in many ways. Valuations are continuously taking a shorter time to deliver, from seven days to two or three days.
Beyond the finance industry where valuations help banks determine the value of a property before taking it up as collateral, accurate property values are also important in mortgage institutions, hospitals (where titles are used as security), courts, accounting, and insurance companies.
Alex Moseti is a Valuer and the Director at Cambrian Valuers LTD.
Photo credit: Pool | Nation
Real estate developers and buyers need valuations to ensure they don’t sell or buy below market value.
As such, accurate and timely valuations are becoming increasingly important, and while valuers have adopted a few tech advancements to improve efficiency, there are still challenges.
One of the simplest and most accessible tech tools adopted by valuers is Google maps, which are also used by surveyors. Moseti explains these maps have become more accurate and are updated daily.
They make it easy to track changes in a location without visiting a property physically. Automated Valuation Models (AVMs) are another promising tech innovation.
“AVMs are like an AI tool which gathers data and provides value estimates. While AVMS work perfectly for the automobile industry which has a straightforward valuation system, they are not easy to apply in property markets,” says Moseti, explaining that there are a lot of dynamics in real estate that make it difficult to use static data in determining property values.
“Real estate is a commodity like any other and it responds to supply and demand curves. If all the land in Nairobi was sold at a standard rate, regardless of the exact location, everyone will rush to buy upmarket properties and no one will consider downtown properties. Naturally, downtown property owners will have to lower their prices in response to the supply and demand curves”.
Unique challenges for Kenya’s unique property market
Beyond supply and demand dynamics, Moseti explains, zoning and building codes are respected in other countries and properties can be identical; with similar materials, amenities and features within a development or location.
This uniformity makes their values almost similar and AVMs can work in such countries. In Kenya, however, zoning is often dismissed, which impacts property values even when they are in the exact same location.
“For example, you can have a mega block of apartments in a prime location. One side of the apartment is facing a slum, other units in the apartment face a forest and others face the highway. The values for these units cannot be the same. This presents a challenge for tech solutions like AVMs.”
Ultimately, Moseti says, you’d need enormous amounts of data, collected and updated frequently to achieve accuracy with an AVM model in Kenya. And although valuers have plenty of data in their hands, Data Protection laws in Kenya are also very strict. Ethically, once a valuer submits a report and they have been paid for it, the data is not theirs to monetise.
Land fraud and the impact of misleading market reports
On whether it’s possible for the industry to come up with an accessible automated model that provides instant insights, Moseti explains that the most one can get from such a platform is general data, for now.
However, the property values and market reports would not be entirely accurate or applicable to individual properties. A valuer would still need to provide professional input.
“When data is not accurate, it can easily mislead and lead to costly consequences. For instance, a location like Upperhill had a lot of hype for years, but the demand has gradually decreased over time. Developments that came up based on misleading data have struggled to make a return on their investment,” Moseti cautions.
Besides, real estate data is not static. Sometimes property values stagnate and in other cases they drop. And even when a machine settles on a specific value, the buyer holds the ultimate power to determine property values and market trends.
Fraud in Kenya’s land market is also difficult for tech models to navigate. Illegal land transfers are sometimes masked so well that it is difficult, even for experienced investors, to detect. Valuers analyse all these factors within seconds and determine the property’s true value.
Despite these challenges, Moseti says, things are changing. Platforms like Ardhisasa are a great milestone in potentially cubing fraud. Additionally, virtual valuations are also an exciting development to look forward to. For instance, if you are in Malindi and you want a valuation, it’s possible to record the house virtually, and share the content with a valuer.
Abinah Gekara Onyambu is a Data Scientist and a Business Analyst at Nakala Analytics LTD.
Photo credit: Pool | Nation
Data science to the rescue
While the real estate industry is catching up, big tech is revolutionising other industries and data scientists say it is possible for real estate to tap into existing technologies. DN2 Property also spoke to Abinah Gekara Onyambu, a Data Scientist and a Business Analyst, as well as Enock Keya, an AI expert from Nakala Analytics. They paint a picture of a fast-moving world out there.
In the finance industry, credit officers simply key in your information into an algorithm and in seconds they generate your credit rating. They can also analyse market conditions and determine whether it’s a good time to lend you.
The e-commerce industry is also using data to improve sales by using data to recommend products to people, thus increasing the probability of selling. In governance, data has been used to identify sources of misinformation, especially during crisis periods like the pandemic era and quell such with well-informed public statements and campaigns.
Although using data to make decisions is not entirely new in businesses, the scale and techniques used are on another level today.
“Previously, people used traditional statistical approaches like “increases and cuts” to analyse market trends. Later we moved to analytics which relied on previous data to make decisions, and now decisions are made predictively. Predictive analytics is about predicting where an industry is headed instead of relying on past data,” explains Onyambu.
An example of this is the digital twin in retail. “AI can simulate an individual in an ideal world and anticipate what they need not just presently, but in the future too, and businesses can leverage the digital twin’s data to determine the person’s future needs. As a result, retailers are making decisions based on surety, not what is likely to sell, rather, what will sell,” says Keya.
AI in real estate valuations
Despite the challenges in real estate, both Onyambu and Keya agree there is hope for the industry to tap into these exciting technologies.
“There is plenty of real estate data in property listings, land registries, county governments, google maps and even social media that can be used to train machines. The only challenge for real estate companies is the capacity to harness and use the data.”
To harness such data, there has to be trust and a middle ground that will convince all data holders to bring the data together and share it even if they adopt aspects of data monetisation (selling data to each other) within prescribed data laws and ethics.
To make valuations faster and more accurate, the industry needs to embrace a data-driven revolution.
Photo credit: Shutterstock
Beyond gathering data and using AI to analyse it, Onyambu suggests rethinking how property values are arrived at.
“With proper data analytics, you try to single out different features in or around a property, for instance, and determine how they affect the price. If a house has a sunrise view, valuers can agree that such a feature affects the value of the house by 5 per cent or so. Factors such as distance from the road, floor level and others can have definite percentages that determine price increase or drop. With proper data analytics we are able to create a structure in valuations. Such an approach will standardise the industry and make it easier to apply big tech solutions,” Onyambu explains.
This kind of approach would perhaps make Automated Valuation Models, more applicable. To add to this, Keya says, “traditional valuation uses what we call regression techniques or hedonic pricing models. Statistically, these methods have an assumption of linearity. For example, the assumption that when a road is tarmacked next to your property the values increase is linear, yet a tarmacked road does not always lead to the expected property price increments”
A lot of the linear assumptions in real estate are simply speculative, as such, the traditional valuation models may need to be over-hauled to adopt new dynamics that will be supported by artificial intelligence and machine learning.
The traditional models also make it difficult to adopt progressive technologies which are readily available.
A good example is spatial data sets. Using satellite imagery over a period of 10 years, you can easily tell how the physical land mass is changing over time and therefore be able to predict future market trends.
Virtual tours, mentioned above, would also be useful to AI models. “Once you put that data into a well-trained AI model it will extract all the nitty gritty details, and produce market reports faster than a human valuer,” says Keya, adding this does not means valuers would be replaced. Instead, they should be leading in this revolution and running these models.
With other industries becoming more efficient through tech, Onyambu wonders if there is a push from the consumers to improve real estate valuations and market reports.
For instance, banks, which are very proactive in adopting technology, are likely to start pushing for tech solutions, noting, however, that ordinary consumers might be unwilling to take up tech in real estate because people are not likely to trust computer generation valuations or reports. “People trust physical research and talking to people,” explains Onyambu.
Cost is always a big concern when it comes to tech transitions, but Keya advices, “In data science, we don’t look at the cost, rather, the need. The cost of not having improved systems is always higher. For instance, what is the cost of an overvalued property to a bank? In many industries, small companies are able to adopt data technologies and even serve thousands of clients without a single employee because beginner level tech is very affordable.”
Beyond data science, both Onyambu and Keya recommend expanding the scope when it comes to tech in real estate.
Keya suggests adopting digital twins before developing housing projects to ensure units address buyers’ real needs and wants. Onyambu concludes by saying that data science would also be helpful in analysing units that have been in the market for far too long to figure out why they are not moving and what can be improved in future developments.
