“At the end of the day, consumers don’t care about what kind of technology a site is built on,” says Deep Varma, VP of engineering at Trulia. “Consumers care if we’re giving them what they want.”
And what they want, in Trulia’s case, is to not be worn down into a nub by the really exhausting experience of home buying. Complicating that is how emotional an experience home buying can be, he adds.
Data is the secret to tackling these challenges on both the customer side and the development side.
First, there’s the foundational data of their platform: the new listings coming into the system, school data, crime data, public records, and the larger demographic look at who’s buying and selling homes. And on top of that, they layer consumer data, which comes in two flavors: explicit and implicit.
Explicit signals are when you tell the company exactly what you want. For instance, the rare and elusive two bedroom, two baths standalone home in Noe Valley, San Francisco, and within a specific budget. Implicit signals are what can be inferred based on your actions across the site. For example, if you keep coming back and looking into Noe Valley district listings, machine learning and the personalization platform surface the right listings.
“Everyone is different, their needs are different, so the foundation of this platform is the unique preferences and the search criteria of a consumer,” Varma says.
Trulia’s platform requires what he calls the three machine learning pillars: computer vision, the recommender system, and the user engagement prediction model.
Computer vision allows their machines to look at house photos and identify, say, a home with a remodeled kitchen and white granite countertops, with a carpeted family room and hardwood floors in the bedroom. And on the user side, the technology keeps track of dwell time — are you spending lots of time admiring listings that offer white granite countertops? Then it makes sense to show you more homes with remodeled kitchens and granite countertops as you search.
Computer vision also allows the company to offer home collections on the front page, based on your unique attributes — come visit the site and you’ll find a world of white granite countertops to start clicking through, right away.
The recommender system, Varma explains, levels up the company’s game.
“We know what you want, but machines are smarter and can look into that with a much broader view on what other options we can give to you,” he says.
For example, if you’re still on your Noe Valley hunt, admiring a condo with your favorite countertops, and at the same time, similar users are on the same listing, collaborative filtering techniques serve up the next house those folks are looking at too. If you both like one listing, you’re likely to find another listing they spent time on similarly attractive.
The recommendation engine is also able to score new listings with unique user preferences, and then send emails or push notifications in real time when something new and super-desirable pops up.
The third piece, Varma says, is critical: the user engagement prediction model.
In your home-buying journey, they know you want to consume information, but what level of information do you really want? The user engagement prediction model is designed to measure the rate at which you’re opening and responding to emails and push notifications — are you reading them eagerly and clicking through, or are you deleting them because you’re sick of them? The company can throttle its delivery of content to you up and down based on your needs and level of fatigue.
“Because of the Trulia personalization platform, you’ll get an experience that’s enjoyable and meaningful to you,” he says, “and you’ll have no idea we’re providing you this content through this recommender system, which is helping you make your decisions much faster.”
Personalization — and the level that you offer — all rests on respecting explicit signals, Varma cautions. That means you don’t over-personalize. But it is also a delicate balancing act, because it can narrow your search, and your experience, to the point where you aren’t getting what you need.
If the explicit signal you’re giving is that you only want to spend $1M, if there aren’t many listings coming in that range, you’re going to get frustrated. Sometimes, he says, Trulia will want to reach out, start a conversation, and suggest that maybe if you ramped up your budget just a smidge, you might be able to find exactly what you’re looking for. How’s that sound, homebuyer?
“Engaging that emotional connection and taking the feedback — that’s what I believe is the most important thing,” Varma says. “Rather than letting machines decide everything, you need to make this personalization as more of a feedback loop, to understand the consumer needs and put those back in the system to keep delivering value.”
To learn more about how true need-focused personalization can transform a customer journey, how to create emotional connections with your customer that help them trust you with your data, and how companies like Pandora and Trulia are developing killer strategies, don’t miss this VB Live event!
To Read Our Daily News Updates, Please Visit Inventiva Or Subscribe Our Newsletter & Push.