Reduce dropoffs and make home-shopping easy
Zillow, the leading U.S. real estate website, hosts about 2 million daily users.
Since Zillow launched in 2006, the home shopping experience rarely changed. However, a collection of features and CTA placements began to add complexity to the experience and increase dropoffs.
During my tenure, we launched the biggest redesign in 5 years to address these issues, resulting in a 26% reduction in dropoffs and an increase in customer satisfaction.
As Director of UX for Search, home page, and home details page (HDP), I led a team of 8 designers and 1 manager to make those changes.
Framing the problem to solve
Insights we discovered
When I joined Zillow, there were issues with shopping:
Research revealed that:
Buyers found it hard to locate key details on the lengthy home details page
Call-to-action (CTA) buttons competed with needed information
Most buyers shop for months or years before they’re ready to engage with an agent or buy, yet the Zillow experience was designed for users to take immediate action
Heuristic reviews suggested
The hierarchy on the home details page was confusing
CTAs didn’t align with user goals
Competitive analysis indicated:
Users found it more difficult to discover homes on Zillow than on competing sites
Zillow had substantially more CTAs than competitors
Zillow didn’t offer the lifestyle information that competitors provided
Unintended outcomes: as Product Managers prioritized new features and CTAs for their areas of responsibility, they added them without regard for the overall experience. Design hypothesized some new features and most of the additional CTAs inadvertently made it more difficult for users to perform the core jobs to be done (JTBD) of home shopping.
We needed a holistic approach — it was time for Design to sell the organization on a compelling, effective, and comprehensive solution.
Putting key metrics into perspective
Zillow’s most important CTA is the contact agent button. Users who click it are connected with an agent who pays Zillow for the lead.
Users often clicked the button because they thought they could ask a simple question about the property — in most instances, their current goal wasn’t to connect with an agent who could represent them in a buy transaction.
Of users who click that CTA, 46% abandoned before connecting with an agent. How might we help users accomplish their goals and help reduce dropoffs?
The Saturday Night Live TV show even did a skit about the phenomenon.
The strategy
Suport the shopping experience
Make information easier to find and consume
Provide shopping tools that help buyers process and save information about homes they’re interested in
Reduce funnel drop offs (by addressing user goals)
Improve search (to be more competitive)
Help buyers get a “better sense of home” (a brand promise)
ell the concept to Product and leadership
Before we could get the organization to prioritize our approach over other features, we needed them to fully understand what the proposed solutions were and how they impact users. We also wanted our PEMD (Product, Engineering, Marketing, and Design) partners to collaborate on solutions with us.
We conducted a 2-week design sprint with the cross-functional team, leveraged existing personas and known pain points, and then riffed on narratives of positive shopping experiences that align with Zillow’s company-wide goals and mission.
We didn’t want to host an ideation session where the team talked about a collection of disjointed features, we wanted the conversation to be about the JTBD of our target personas:
Beth, the first-time home buyer who is likely to have no relationship with an agent
Brandon, the repeat homebuyer who is likely to work with his previous agent
Alan, the Zillow agent
We kicked off the ideation session and shared our research and our hypothesis.
Research insights
Users find it difficult to find information on Zillow’s home details page
They’re likely to start their home search on Zillow (because it’s a familiar brand) but subsequently go to competitor sites to get lifestyle information such as climate and crime risk, noise, neighborhood amenities, and schools
In hot markets, users are likely to constantly check for newly listed or reduced homes, for fear of missing out on “the one”
They’re afraid of being spammed with phone calls if they reach out just to get questions answered
Hypotheses
Millions of people “Zillow-surf” to dream about homes months and sometimes years before they’re ready to buy. If we support their complete shopping journey and nurture them with insights and advice along the way, they’re likely to be on Zillow when they’re ready to buy.
If we make it easy for Beth and Brandon to shop together, they’re likely to use Zillow, this differentiates us from the competition and enhances our ability to understand what they’re looking for.
Most shoppers track homes of interest and the attributes they’re looking for on spreadsheets. If we make it easier for Beth and Brandon to share information about homes on Zillow, we’ll become more “sticky” and they’ll likely use us every day of their search.
If they’re on Zillow when they find “the one” they’re more likely to work with a Zillow agent.
If we’re able to recommend homes that meet their lifestyle criteria, Beth and Brandon are more likely to find a home that meets their needs in unexpected locations.
If Beth and Brandon feel supported and get answers to their questions without being forced prematurely into a relationship with an agent, they’re more likely to reach out and let Zillow help.
If we can make search effortless, like having a conversation with a real person, then Beth and Brandon will be more able to articulate their needs and find homes that fulfill their criteria.
Craft a vision and narrative of better experiences
Members of the Product team can tell you what features they want in a product, but that doesn’t often inform:
how compelling the features are or how they make users feel
if a group of features aligns with the user JTBD
if new features create significant gaps in the experience
The power of narratives is that anyone can tell a story. Stories keep us focused on the user and how they accomplish their goals with trust and confidence.
Narratives are memorable — they help the PEMD team understand and empathize with the user’s experience. During feature development, we can refer back to the narrative to make smarter choices about the impact that our choices have on users.
The team riffed on several narratives. The UX and content designers refined the language and added illustrations for the top ideas. The result was a clear representation of how Zillow might impact Beth and Brandon’s lives in a way that results in a win-win for buyers and Zillow.
Highlights of the final narrative
Beth and Brandon shop for a home
Zillow's new experience makes it easy for them to share what’s important for their next home and the homes each wants the other to consider. Each night when they check for new homes, they can easily see what’s new and see key attributes so they quickly triage homes.
They can also get a “deeper sense” of each home because there are high-fidelity photos. Instead of looking at a collection of disparate photos, they see a layout they can use to move around the house. Lifestyle information like flood risk, neighborhood amenities, and more are also highlighted.
To help them understand what they can afford, they are encouraged to engage with a Zillow mortgage advisor early in the process so they can have confidence in how much they can spend and be ready when they “find the one”.
The narrative takes them through months of Zillow-surfing and exchanging information until they’re ready to take a tour. They meet Alan the agent, share their search criteria, and continue to evolve their requirements as Alan exposes them to even more homes and areas.
The story ends when they find a home that meets most of their needs and they make an offer.
During the experience, they didn’t feel rushed — instead, they were provided the tools, time, and support needed to make a confident home-buying decision.
Get leadership buy-in
The team presented the illustrated narrative to Product and company leadership who approved the concepts. Over the following months, we worked with our Product partners and adjacent teams to implement parts of the vision.
The goal was to deploy capabilities in a way that would “do no harm” to key metrics. We designed MVP experiences that provided enough user value so that we could learn about needed enhancements before investing more in any feature.
Define, build, and track the MVPs
For each feature, we used the narratives to explore the least amount of functionality, or minimum viable product (MVP), we could deliver and still provide user and business value.
This helps the PEMD team think through tradeoffs — we don’t want to release more than is needed because extra features make it more difficult to determine which capabilities impact outcomes and metrics.
The next section describes outcomes for several of the narrative’s features.
Note: during the project, the company introduced a new business goal, the Housing SuperApp, and re-aligned resources to that initiative. The work and learnings from this work provide a foundation for future work.
Executing the vision…incrementally
Triage helps you quickly decide if a home meets your needs
A home shopping experience often involves identifying tens or hundreds of potential homes and then quickly triaging a few to learn more.
When key attributes are located at the top of the page, buyers can quickly determine if they need to spend more time on a home.
To find out which attributes are most important, we conducted a card sort exercise and then tested a few options to determine what information might work for most users. We then user-tested the top candidates and A/B tested them to determine which one to deploy.
Key moments and learnings
Conducted card sort exercises
User-tested 4 approaches
Discovered users want to customize the triage list (not implemented)
Learned users who viewed the page top then abandoned increased (they saw what they needed then moved on to next property)
Note: it would have been preferable to leverage AI to synthesize user goals and customize home information for each user, but we were still early in our exploration of AI.
Hub and spoke helps you find and focus on what’s important
To reduce page length, we explored a page layout that organized home details into a few key sections or “hubs”. Once a user determines they’re interested in more detail, they can find them on the spoke.
Key moments and learnings
User-tested 2 approaches
Discovered users who moved past triage section (interested shoppers) continued to page bottom more often than previously
Design performed well on mobile
Next best action CTAs guide you confidently to your goal
Leveraging our card sort and JTBD insights, we positioned the most important CTAs for touring, financing, and contracting so that they’re always visible in a static panel.
We placed other CTAs on the spokes that they’re available in those moments when the user has enough information or interest to benefit from them. For example, if a user views the spoke related to the home’s cost, the most relevant CTAs to place on the spoke are those related to financing.
We collaborated with other Product teams to define a framework for CTAs so they know where on the home details page they can experiment with conversation moments without negatively impacting the shopping experience.
Key moments and learnings
Because we provided context, research shows users were better informed about why they might click the CTA at the end of a spoke
A/B tests showed 10% reduction in post-CTA abandonment
Comparison tool helps you identify “the one”
We designed a feature that allows users to compare homes. This initial release was well received by customers and differentiated Zillow from competitors.
It was not the robust, customizable version that we initially envisioned, but it set the stage for true collaboration between co-shoppers and ultimately Alan.
Key moments and learnings
Customer surveys indicated users appreciated this feature
Discoverability was difficult; surveyed users didn’t know the feature existed
Initiated a project to understand how we might emphasize the feature (not prioritized due to SuperApp pivot which might deprecate the effort).
Search is as easy as a conversation with an agent
In my early days with Zillow, the team worked to evolve natural language search that enables users to enter multiple terms and phrases to specify what they’re looking for. However, the strings users entered had little to no context (within the same session or over multiple ones).
In the previous year, we designed and abandoned conversational AI search over concerns that it might provide responses that violate Fair Housing rules.
We conducted a second design sprint and produced a narrative that focused on a new search experience. The interactions didn’t provide conversational responses, it only delivered a list of properties. We were able to leverage the user’s conversation to enrich what we know about them and their changing requirements; AI synthesized the multi-session information, property photos and descriptions, and neighborhood information and then deliver better search outcomes.
Our new search narrative describes an experience where Beth and Brandon speak or type phrases to describe what they’re looking for. For example, in our narrative, Beth says:
“I want a 4 bedroom, 3 bath, single family, mid-century modern home, with the primary bedroom on the main level, a large back yard, and within walking distance of a park and restaurants. I’d also like a pool, mature trees, and a 1 acre lot.”
Natural language search would have been unable to put much of that criteria into proper context. It also would not have been able to use the current search query to clarify previously entered search criteria.
Key moments and learnings
Allowing users to conversationally specify search terms aligned with what they thought natural language search did, however the outcomes were more accurate
The percentage of homes returned (i.e. search quality) increased
Climate, school, and neighborhood data help you find the right lifestyle
Research insights indicated that when home buyers shop, they not only shop for a house, they shop for a lifestyle. Details that users expect include climate risk, noise, schools, and local amenities.
Our narrative helped highlight the importance of this information and how previously users went to other websites to get information about neighborhoods of interest.
Zillow leadership previously resisted adding this information because of legal and ethical concerns. However, our narrative helped paint a picture of how critical this information is to home shopping.
Key moments and learnings
We found that showing climate risk with map colors in an accessible way was challenging. Eventually, we were able to discover color schemes that work with visually impaired users.
All of these features involved working with 3rd party content providers; discovering one that aligned with company goals and cost parameters was the biggest challenge.
The competitive analysis provided the incentive we need to convince leadership to provide the content.
A deeper sense of home
We collaborated with another design team that focused on Showcase, an interactive experience to deliver a high-fidelity, immersive experience on the home details page.
Showcase helps address some of the simple questions buyers might click contact agent to answer. Now they could see the overall layout of a home, and change their views to see what rooms look like.
Key moments and learnings
While getting users to tour is a key goal for Zillow, helping users virtually see a home didn’t replace the need to tour — it did give users more confidence when choosing which homes to tour.
Get financial information on your terms
One of the biggest obstacles new buyers face is understanding the true cost of ownership and the landing process.
We explored a finance chatbot that allowed buyers to ask finance questions so they could be confident they were searching in the right price range. When buyers reach the end of the advice they can get from the bot, they are encouraged to connect with a Zillow advisor.
The Product, Design, and Compliance teams trained the AI for months to uncover and mitigate responses that might pose compliance risks.
The AI finance chatbot, AI-powered conversational search, and Zillow’s ChatGPT plug-in were either not released or pulled after release due to compliance concerns.
Pick up where you left off and never miss out
Zillow’s home page performed extremely well in terms of SEO and driving traffic from the web. However, it provided no features for returning visitors to resume previous searches.
Research indicated that users land on the home page, then enter a new search term to start over again. Many were unaware that Zillow had a feature that allowed them to save search criteria from one session to the next. However, even with saved searches, there was no way to update legacy search criteria which resulted in multiple search lists that generated unnecessary update emails from the Marketing team.
Leveraging our narrative again, we explored various options that allow Beth to see a list of homes she hadn’t seen from her previous search criteria. This helps her avoid seeing homes from her previous session, and provides a way to resume and evolve her search.
Key moments and learnings
The search term had been exploring ways to make it possible to pick up where you left off within that experience, but adding more features there increased the complexity of that surface.
Displaying this feature on the home page provided the context needed by users who simply go to Zillow.com to resume their search.
The home page also gave us the space to pair search information with recommendations.
Results
“The new design delivers a fun and efficient way to browse homes on the Zillow website, making it easier for home shoppers to navigate and process information," said Jenny Arden, chief design officer at Zillow.
(PR Newswire article, Oct 2023)
Key takeaways
Be a strategic partner to Product and the business
Use narratives, which are the most efficient low-fidelity artifact we can create, so teams can quickly iterate on complex ideas
Complement business metrics with qualitative insights to help the team see new ways to evaluate outcomes
Led vision exercises so the team has a unified vision of a cohesive, functional experience
Include PEMD partners so everyone has a voice and buys into the vision
Have a hypothesis as to the efficacy of new experiences
Think lean — release only those capabilities that are needed to validate your hypothesis, then incrementally improve based on user behaviors and feedback