Overview

As a lead designer for the Lead Growth team, I delivered design work and strategy focusing on conversion, engagement and monetization growth.

Redesigning realtor.com’s photo experience was one of the team’s first projects. Started as a small feature update, it was turned to big opportunities, followed by multiple projects that benefit the business’ ecosystem. The features were later branded as Photo First by the company, as media reported about us being a market differentiator: a photo centric experience that uses artificial intelligence to provide relevance for home-shoppers in their searching.

Photo First was one of the company’s best performing features in the fiscal year. Here are a few key metrics:

• 60%+ unique users engaged with the new features

• Lifted 20%+ lead generation

• Lifted 20%+ unique users for existing photo features 

• Lifted 10%+ conversions from Search Result Page to Listing Detail Page

The project has gone through a few iterative phases:

• Vertical photo gallery

• Photo categorization by AI to enable personalized experience

• More AI-powered features to facilitate user’s searching; opening up more project opportunities for future development

Background

Realtor.com connects home-shoppers and agents

As part of the business runs on lead generation, realtor.com delivers different experiences towards home-shoppers and agents. My team, on the consumer side, helps home-shoppers search and discover properties of interest; and then seamlessly connect with agents who will hand-hold through the home buying processes.

A new team formed to focus on lead growth & monetization

The company’s new structure has a few orgs, one of which focuses on the growth of the core business, where I took the lead for the design, and be the advocate for user.

Close collaborations across functions has led to the success of this team, including design, product, engineering, research, data analysis, data science and so on. We constantly validated through A/B testings, significantly lifted conversion and user engagement through a wide range of projects from small to large scale. Eventually we had the highest quarterly velocity among feature development teams, and our improved agile process set an example for other teams.

My team focused on mobile platform and there was an equivalent team for desktop. Following the “mobile first” principles, we were usually a step ahead for experiments, while the designers across teams were constantly communicating and collaborating. 

Photo is the top used feature, but not the top lead driver.

We sat down with data analysts, looking through extensive metrics on all the functions in the app, and found that Photo is the most used feature, possibly because it is one of the initial access point for Home-shoppers to decide if he/she likes about a property listing before contacting agents. However, it didn’t generate the most leads as expected.

Why?

Our researchers were doing weekly benchmarking, they analyzed the apps in the vertical market such as Zillow, Trulia and Redfin. I also looked at different market’s apps handling photos, such as Yelp, Google Photos, Instagram and so on.

With two goals in mind, we had a sketching section for solutions

• Help home-shoppers find the right photo more efficiently

• Continue delighting home-shoppers with immersive photo experience

There were concepts as simple as adding a layer of vertical gallery, and also as complex as totally redesigning the Listing Detail Page. PM and I looked at the estimated efforts and impact to make decisions on what concepts to pursue.

Some concepts did not make to the final based on user testing results

I created interactive prototypes for a few concepts, so our researcher took them for user testing.

One of examples is showed below: the redesigned default view of Listing Detail Page would land user on the photo gallery by simply scrolling. User would naturally immerse in the listing’s photographic story right away.

But we didn’t pursue this concept because user felt more confused and difficult to see key information right below the fold. Other concepts performed better in solving the identified usability issue.

Remarkable results from A/B testings exceeding our expectation prove our hypothesis and direction.

From small feature to bigger picture

“But is this good enough for us to meet the ambitious goal?” Even though we had achieved great numbers, I couldn’t help but think. I attempted to look at the bigger picture again.

“What is the real problem?”

We had personas and extensive user journeys. They are very helpful for answering the question. I took out part of the user flow and deconstructed it, which helped us understand that it could possibly go wrong in almost every steps.

For example, it could be: the hero image does not interest the user; gallery is hard to use; repetitive photos push user away; difficult to locate photos of the kitchen and bedroom; or when there is something missing either in photos or property facts, going back and forth takes a lot of effort and user forgot where they were, etc.

Putting together the flows helped us see why the experience is broken

When realtor.com got the data, we didn’t do enough optimization, most of the time we show it as it is. That’s why the heavy-lifting is on user’s side. The usability issue we identified earlier was the surface of the systematic problem. Those gaps are the root of the user’s need, if we can fill the gap, we can dramatically enhance the user’s engagement with us.

Updated problem: the system fails to deliver information relevant to home-shoppers’ interests timely, which stops them from moving forward.

In order to find a better solution, there were some key questions I asked. For example these HOW MIGHT WE questions are closely targeting at the root problem, also taking our long-term ecosystem into account.

Ideas supported by machine learning

I found artificial intelligence could be a great answer to the HOW MIGHT WE questions. So I sat down with data scientist to evaluate some ideas I had. They confirmed, after doing a spike, that “with our in house deep learning model, AI could recognize photos with accuracy over 97%.” 

What does that mean? It means that we can do various things with a technique called Advanced Image Tagging, for example we can:

• Breakdown photos from raw data into categories

• Connect photos with property facts, such as style or materials etc.

• Provide personalized content and better recommendation for listings,

• Enable searching through photos

Phase 2 solution: photo category with AI

The photo categories are right away in the listing detail page, also in the photo gallery. As user continue use the app, they could be personalized. If kitchen is important for the user, then they will see the kitchen first.

Phase 3 solution: image/text mutual understanding

It is commonly seen that Home-shopper forgets about the listing’s key facts (e.g. room size, yard size, floor material etc.), while he/she enjoys the immersive moment with photos. With the same AI technique, we can help user connect the dots.

This design solution helps user quickly learns about the textual facts of the property from the photos they are looking at. For example, user can see right away the kitchen is 200 sqft, it has an electric oven, and a marble countertop. 

Meanwhile, I took the chance to also launch the redesign of Property Details. The long detail page has not been touched for years and was difficult to search information. With Photo First, user now can seamlessly navigate between photos and property information.

Interactive Prototype

I created this interactive prototype for various purposes from quick testing to development specs.

When applied to this particular project, the vertical gallery is the Good; the categories and image/text mutual understanding are the Better; lastly the match score is the Best. As time goes by, the Good will be done, the Better will become the Good, and the Best will become the Better.

Key result

We contributed 76% of company’s quarterly lead growth, along with lifting 30% for most user engagement metrics.

The Good design gave us lifted 21% leads, making the photo become the most used feature AND the best lead generator throughout the app. As for the Better design, it gave us 60% users engaging with it, and 23% leads lifted.

These outstanding numbers have made the feature the best performing one for the company, thus it was rebranded as Photo First and allowed for more resources for future developments.