What’s utterly unremarkable is the days on market trend for US home sales in 2020. Despite COVID, its steady downward trend of more than a decade has continued unabated. The explanation is certainly multivariate, probably inconclusive, and a subject for debate for sure. Lot’s of factors are in play, like supply and demand trends, cheap financing, and virtual real estate tech, among others.
We took a look at listing photos.
Secrets in the Image Data
Rather than cracking the “why,” our goal was to simply gather data on the “what happened as a result.” Did the COVID-created sequestering change how sellers presented their homes, specifically the images they posted with their listings?
THERE ARE REMARKABLE EARLY INDICATIONS that our image intelligence reveals insights. More than 2 years and millions of images of machine learning training allows our intelligent tech to recognize not only unique architectural styles but also room types. We analyzed MLS data for California through 2020, including over 170,000 listing images, looking for patterns in seller behavior. Relationships in the data are emerging.
We can say with statistical confidence that the average seller in California posted 15% more photos. Not surprising. What’s much more interesting is the types of photos that gain prevalence. Exterior photos increased 38%. “Other” interior rooms (meaning not bedrooms, baths, kitchens, dining rooms or living rooms), increased by 33%. Hmmm.
You might draw anthropological conclusions about the sequester changing what “home” means and how to present it. The master bedroom, kitchen island, and 3.5 baths are not enough. It’s the other spaces that make life live-able when locked up that need representation. You can’t hang out in restaurants but you can hang in your yard, so buyers need to visualize “getting out” safely.
It’s too early to provide really sexy and reliable correlations linking specific rooms to market valuations or bid-ask ratios. But image intelligence reveals insights that we are only beginning to explore. The image below provides a stimulating hint. Stay tuned.
About twenty years ago automated online home valuations became accessible to everyone. The immediacy of the insights quickly became an expectation bred into the process and the value proposition hasn’t changed since. Pick a home search site, enter an address, get a market value. A branded algorithm matches location-based pricing trends with square footage. Ta da, you have a ballpark estimate, invariably eliciting responses like, “how is our home worth so little?” or, “how is that house worth so much?” and, “what do we do now?”
The statistical vagaries of existing solutions drove us to discover a way to answer these questions by re-inventing the home valuation experience. It’s called StyleExplorer. It mashes up standard valuation (size and location) with intelligent image data of all of the attributes that should contribute a home’s actual value. The new value proposition is confidence, control and guidance in understanding the optimal valuation of home, whether you are selling, renovating or buying. The new approach changes the question from “how” the valuation was calculated, to “what does it mean” and “what can we do with it.”
do the new math
It’s a simple theorem: the sum of the rooms is bigger than whole (house).
Purlin’s StyleExplorer image intelligence allows side-by-side comparison of homes’ individual assets like architectural style, rooms, backyards, and neighborhoods. For buyers it’s a unique experience analogous to crashing multiple open houses simultaneously. For sellers, when done across homes with similar market valuations, the comparison will reveal valuation gaps and opportunities. Some valuations will fall short of a home’s salient features (valuation to gain), some features will fail to look like they should (room to improve). These immediate insights are priceless to sellers and agents nurturing process. It’s truly the automated love child of the open house and the MVP (market value pricing analysis).
Imagine if you are ready for a remodel. You want to keep up with the latest designs and furnishings (and the Jones’s), or you just need some inspiration. Looking at catalogs is helpful but feels staged, if not inaccessible. What if you could look into rooms in the real world, that people live in, and see what your contemporaries have done with their homes? Imagine peeking into your neighbor’s living room, or backyard, or man cave. This tech was custom built for remodelers too.
make it a place to live, not a property to buy
Exploring what’s inside and around homes, the design touches in the photos owners have picked, taps into past or imagined experiences and makes a crude dollar value calculation for a property seem disconnected and even arbitrary. It humanizes home valuation by dimensionalizing it – it gets to what the value of a home really means in personal terms. With this unconscious shift from left brain to right the listing ceases to be a property to buy and becomes a future place to live. Traditional automated appraisal solutions ignore that with a purchase this big, and this important, beyond basic requirements the emotional value will most often prevail over rational appeals. Emotional value can compel buyers to favor one home over another, even with price disparity. Emotional response to image-based immersion, not clever language or negotiation techniques, powers buyer motivation.
take it to eleven
Without your permission, the world has a value in mind for your home. You have no control over whether it’s spot-on, a bit aggressive, or a total low-ball.
Purlin’s image intelligence gives back control, provides guidance, and perhaps most importantly endows confidence. It can illuminate misrepresentations and re-set home value conceptions, from a position of proof-based strength (pictures don’t lie, especially side-by-side). It arms users with evidence of weaknesses and strengths, what’s needed, what’s possible, and what’s truly meaningful. Sellers can feel smart and satisfied that they know clearly the value of what they are selling, glean what to showcase and how to prepare. Buyers can explore and understand (and love) the possibilities of what they can buy. Remodeler’s can learn how best to make their home better.
Our new vision for home valuation based on image intelligence forwards a fundamental belief about home valuations: they provide context not proof; the sum of the rooms, and future experiences in them, is often greater the unexamined appraisal. It allows you to take home value to eleven.
Curb appeal … status, taste or just home? Some say that architectural style of a home says something about you, like your car or your clothes. Others care more about design, and less about outward impressions. It’s not far-fetched to think that either type would love to search homes by style.
Unfortunately, the search for homes for sale by style is weak at best and impacted by considerable misinformation. In other words, you never know what Google’s gonna get you because of faulty text-based descriptions of listings. The principle source of home listing, the MLS, includes architectural style only 33% of the time. What’s more, those descriptions are wrong half of the time. When you search for Modern Homes in California you will most likely get a fraction of what’s truly available. And more than a few mislabelled Mediterranean, Victorian and Craftsman homes.
Images speak better than words
Purlin fixed this using pictures of homes instead of relying on sketchy or omitted text descriptions. As a result, home styles are identified correctly 80% of the time, or 2.5X the accuracy of existing sources.* We trained our AI with millions of images to find homes for sale using a combination of advanced computer vision and machine learning technology. Home buyers (and renters and renovators) can now do home searches by specific architectural styles. They can see what those styles look like in the actual neighborhoods they live in or are hoping to move into.
We say, “a picture is worth a thousand words, but data in an image is far more reliable.” Even a thousand words is a meager facsimile for what your eye knows, consciously or not. We created technology that allow us to skip the awkward verbal translation.
*Based on over 140,000 listings sourced from California MLS
You like what only you like when you see it, even when you can’t put a finger on it. That one piece of fruit or flower in a field stands out and says choose me. Multiply that idea by a thousand when looking at homes in the sea of listing sameness. Imagine if you had a tool for finding a home using style, your own personal style …
Say you are looking for modern homes in Manhattan Beach, CA. Whether you are using home search site or an agent, most often homes for sale are picked for you. Very basic queries that deliver very basic results. You are left to choose which one creates the most tingles (or any?). We think this tradition is backwards. So we reversed it.
We created StyleExplorer to help you figure out what you like before forcing the “what do you want” question. It’s more than just a way for finding a home using style. It’s the first tech tool to allow you to compare the styles and decor of homes side-by-side. You can search every room, feature and view in a specific neighborhood so that there is no question which one to pick.
As you explore your preferences and “Peek” cool images, StyleExplorer keeps track of the details. It learns what you like to make your home searches more productive later. This unique personal style profile gets included in the overall search calculus of Purlin’s matching algorithm. That way when you search for a home to buy, your choices are not strictly defined by rational calculations and externalities. “Finding a home by style” means your consideration set is homes that you will likely like. Your search is through the unique lens of what you like, or, finding a home using style not choosing from options chosen for you.
Opinions may differ on the ultimate impact of artificial intelligence (AI) on humankind, but there is little doubting that it will have a significant affect on real estate. Some may see it as part of the quiet but insidious take over by the Machines. First it was our music. Then our movies. Then what we buy, eat and drive. It’s over when they take control of our homes because we’ll be literally living in a Hollywood cliché of an AI apocalypse …
A less dark scenario is that AI will be the thing that saves the real estate industry. Some companies already use AI as a financial vehicle to arbitrage housing inventory. Others are pursuing AI for process automation, to make the transaction cheaper or agent better or CRM more effective. But despite real estate tech advancement, 2/3 of today’s homebuyers report buyer’s remorse. That’s where the highest and best use of AI comes in: finding the right home. Ironically, it’s the Machines, through intelligent personalization, that will heal the home buying process by restoring to the home buying process a fundamentally human value: trust.
Let’s step back. The below-scribbled chart, the inspiration for this piece, describes the inverse relationship of trust and remorse in the real estate industry, overlaid on our relationship with data over time. It frames how we got here and where we could go. Despite its rough and humble appearance, this chart is not water-cooler whimsy, but a working theory based on real data (much of it from National Association of Realtors surveys, 2015-2000).
Trust, Remorse and our relationship with Data
Prior to and through the Data Access Era (which began with the advent of the WWW), trust in the real estate space was still placed in the real estate agent. He or she proffered access to a proprietary database, knowledge of a confounding process and wisdom about neighborhoods and listings. The agent would show the best options to the buyer who would ultimately be satisfied with picking the best of those options. Trust was high, remorse low.
When Zillow cracked open the Multiple Listing Service (MLS) in mid-2000s, granting the world access to every listing, we entered the Data Transparency Era. The old script was flipped, “the customer doesn’t know what they want, until you show them,” became, “the more you show the customer, the more likely they’ll know what they want.” More listings, more pictures, and more agents meant more control, a better experience, and a better home. That’s still the lure, and over 90% of homebuyers use online search services like Zillow. Others like Redfin have joined the conversation with promises of using technology to make the process cheaper and easier. Despite these promises and market penetration of these companies, buyer’s remorse is reaching new heights. Trust is diving. What gives?
Too much too soon
Transparency solutions have eroded the agent’s core value to consumers, which is still trust. Zillow’s old-school disintermediation model turned agent selection into referral bidding. The typical brokerage business model changed to one built on high lead volume and low closure rates. The agent role was commoditized, diminishing the relationship, experience and service levels provided to buyers.
Buyers want their agents to “help find the right home to purchase,” four times (yes, 4x) as much as they want help with price negotiations or process. This stat flies in the face of two decades of real estate marketing and tech innovation. Agents are now actually unsuccessful at meeting their clients’ top need, failing to find their homes over 2/3 of the time. Interestingly, when choosing an agent homebuyers care most about trust, five times (5x) whether the agent uses the latest technology. So it’s safe to say that over 2/3 of the time, none of the new fangled tech matters. Trust has been betrayed.
The Transparency Era has left buyers on their own sorting through countless poorly-suited listings and images, at the mercy of the Paradox of Choice (see “Homebuyers have a lot of choices and that’s not a good thing”). Technology’s gift: an average of 20 gigabytes of data and images, and mind-numbing trade-off analyses. Over 80% of buyers report having to compromise. Trust in the process is lost.
There’s no data like your own
The new Data IntelligenceEra offers hope by way of trust through intelligent personalization. We’ve begun to trust AI in many areas of our daily lives implicitly. An AI’s recommendations are more than just convenient, they are comforting on another level because they allow us to let go of decision-making anxiety and the accompanying self-doubt that comes with too many options. Today’s convergence is not around a device or platform, but the individual. The perception, however real, that something was made or found for us specifically is unavoidably powerful. In fact, 80% of consumers are more likely to purchase a brand if it’s personalized (Epsilon report, 2018).
What the smartest AI woulddoto find the “right” home
It would get to know everything about the home buyer first. Start with what styles and floor plans and furnishings they like and what features, windows, and neighborhoods catch their attention. Track of their interactions with images, uploaded photos and listings. Learn where they are in life and expect from their lifestyle, where they like to spend time, where they need to be, and what they like to do. Use machine learning to constantly fine tune the relationship and make it stronger. Take all of memories from this shared experience and perform a ridiculous multivariate trade-off analysis that considers both needs and preferences. The smartest AI real estate solution would be the ultimate antidote to too many, ill-fitting, options, and try its hardest to prevent discontent. The buyer, having invested themselves at every interval will have no choice but to trust the end results — they would be the best reflection who the buyer is.
The way AI will re-invent the real estate industry not by magically cracking the code and fixing an antiquated and anxiety-ridden process. It will be by aligning technology’s priorities with those of the buyer, and first and foremost with a focus on finding the best possible home for the individual buyer. It will do this treating the process as a relationship rather than the transaction.