PropertyGuru Taps Deep Analysis to Spruce Up Quality of Listings

Carro urge interoperability

To provide house hunters with good quality listings online, PropertyGuru has turned to data analysis to identify components that should be removed or enhanced and score agents on its platform.

For instance, listed properties that contained more indoor images were found to convert better, with more leads generated from such listings. Images that showed different areas across the entire property, including the kitchen, bedrooms, and living room also had higher conversion, said PropertyGuru CTO Manav Kamboj said in an interview with DTC Daily.

One key challenge the Singapore-based online property marketplace faced was “fake” listings, where an agent would not have a particular property on sale but used it as a way to bait potential buyers and peddle alternative property options.

To address this, PropertyGuru used computer visual tools powered by artificial intelligence (AI) to identify whether a listing contained genuine indoor images of the property and to block images, for instance, that contained the agent’s photo and contact number.

“Through deep analysis, we found consumers tend to give leads to properties with more indoor images,” Kamboj said, adding that agents who used bullet points and maintained a content structure using certain keywords enjoyed better conversion.

Such insights then were used to score the quality of listings, where properties that did not meet such key criteria would not be listed as high on the site, compared to those that did, he noted. The data insights also were fed back to the agents as guidelines on how they could boost the response rate of their listings.

He said his team continuously monitored visual recognition tools for accuracy and worked to improve the AI model.

Founded in 2007, PropertyGuru currently operates in four other markets in the region including Indonesia, Thailand, and Vietnam. It has a headcount of some 1,000, including 120 working on product and technology development in its Singapore and Thailand offices, and 150 employees in Vietnam.

On average, the site has more than 2 million property listings and processes 800,000 enquiries a month. In addition, more than 20 million house hunters across the five markets visit the site each month, including some 5.5 million in Singapore and 8 million in Vietnam.

It also has more than 50,000 agents tapping its online marketplace, with the largest number base in Singapore. More than 80% of these agents renew their subscription on the site each year.

Manav Kamboj, CTO, PropertyGuru

Agent subscription fees formed the bulk of PropertyGuru’s revenue, which also included ad revenue and subscription fees from its FastKey CRM solution, designed for property developers to manage the sale of their properties.

Noting the need for the industry to see greater technology innovation, Kamboj said there had been much development in other market segments such as ride-sharing and e-commerce.

The property sector had lagged because it typically was the biggest investment most people would make in their lives, he said. The industry, though, was due for a major disruption as consumer expectations–fuelled primarily by the growth of online commerce–now had evolved.

Shoppers had become accustomed to better online experience and having their orders shipped to their homes in days. They now expected the same results on property portals, too, and demanded an increasing amount of information, he noted.

For instance, property portals 10 years back would mostly consolidate supply and enable house hunters to search based on certain parameters, such as location, price range, and rent or purchase. They then browse a list of potential properties and set up appointments with the agents to view the property.

Today, they expected the same level of service and user experience they had on other commerce marketplaces on property portals, Kamboj said. PropertyGuru in 2016 realised it needed to enhance its platform–specifically, its search engine–to better meet such customer expectations.

The site incorporated more personalisation and the company began building up its own data science team, so it could better understand its customers and offer more relevant content. This would enable the site to deliver richer user experience as well as actionable insights and capabilities that could help house hunters make more informed decision, he said.

The data science team now is about 20-person strong including full-time staff and interns, he revealed, adding that PropertyGuru also partnered Singapore-based education institutions to work on specific projects. For instance, it worked with Singapore Management University to build a model around pricing.

To ensure its data pool was sufficient from which to glean insights, the property marketplace tapped its own data including demand prices as well as third-party APIs such as Google’s location data and transaction data such as rental yield and property sales from Singapore’s Urban Redevelopment Authority.

Looking beyond product discovery

All commerce platforms thrived on product discovery including PropertyGuru, which aimed to use its data insights to ensure the right properties were shortlisted according to the house hunter’s requirements, Kamboj said. The site, however, differed from other e-commerce marketplaces in terms of fulfilment, since most transactions took place outside the property platform, he noted.

Nonetheless, it was looking to extend its influence beyond product discovery and towards transaction, where it was experimenting with efforts to help property hunters find the right financial service provider.

In Malaysia, for instance, PropertyGuru introduced a home loan pre-approval service where house hunters could obtain their credit score and loan ratio by entering their personal details and national ID number on the platform.

Using an API provided by Malaysia’s credit bureau office, Kamboj explained that the site was able to extract the data and help property hunters with their home loan application. By facilitating this on its site, its users could view properties that fit their approved loan rate and without having to first approach banks to ascertain their credit rating.

PropertyGuru was exploring options to bring similar solutions to other market, he said.

Having focused significant efforts on improving features for consumers, the company now also was shifting some attention to the agent community, he noted. Last year, it kicked off several projects that enabled agents to see, in real-time, how their listings were performance and the next course of actions to take to improve performance, such as replacing the images used.

He said the site’s integration with social media platforms was available in Singapore and piloting in Malaysia. It planned to extend this offering to other markets.

The online site also worked on a pilot in which it analysed its data to identify high demand, for instance, for two-bedroom units in a particular location. It then would provide this data to the agents so they could source for more units to meet this demand.

Kamboj said PropertyGuru was working on a “next-best action framework” aimed at providing recommendations to help agents improve their returns on investment, such as adding more images to increase their conversion rate.

Because data was critical to improve its features and service delivery, he pointed to a data void where the site did not know what transpired after the house hunter ventured beyond the marketplace and interacted, offline, with agents in their property hunt.