no image

User Research: Yelp & Directory Websites


The Problem

Let’s start by talking about DexKnows. DK (DexKnows) is a directory website exactly like Yelp, Foursquare, and TripAdvisor.
In early 2015, while working at Dex Media, the stakeholders saw a dip in DexKnows’ performance in terms of Bounce Rate and monthly UVs. This directly correlated to a decrease in revenue for one reason: DexKnows derives part of his revenue from paid actions that users take and clients pay for. For example, if a user clicks through a business website from the business profile page on DK (DexKnows), said business is charged a small amount.

They reached out to my team to help to fix this issue and I took charge of the UX research.

The Discovery

The first thing I did was looking into Google Analytics. Yes, the numbers were down compared to just a few months ago. This could have been the result of a big internal migration to a different system, so I wanted to find out exactly what users thought of the website.

First, I ran 5 user tests on to gather some feedback on DK. This first round of interviews was very productive and shined a light on some major discoveries:
1. DK doesn’t look friendly.
2. DK shows irrelevant results.
3. Why would I use DK when I can use Yelp?

These 3 discoveries touched on 3 different aspects of the business:
1. UX/UI – make the website easier to use and more engaging.
2. Technology (backend) – find out what “relevant results” means to the users and show only those results.
3. Marketing – give the users a “compelling reason to buy” (borrowed term from “Crossing the Chasm”, by Geoffrey Moore).

In order to dive deep into all 3 issues, I had to get a better understanding of the user needs. What were we lacking compared to similar websites? What did the user think of the competition? Why were we lagging?

The Process

I set out to start doing some moderated user testing across the major known competitors.
To start, I checked on the budget I had to do so, which I found out, was very low.
This left me with one option, really: going with In-person tests are the best for this situation, but also require some sort of reward for the users, which the company was not willing to pay for.

I decided to create 3 major tests for the top 2 DK categories: Restaurants and Dentists. By doing this I would make sure to cover up to 55% of the users using the search feature.
After that, I created a small process loop to get the most out of my tests:
1. Create a list of 5 tasks the users have to go through followed by 3 open questions.
2. Choose 3 major competitors.
3. Create User Test with 15 users.
4. Have the users go through each of the 3 competitors and perform the tasks + answer the questions.
5. Watch the 15 videos, record takeaways, ratings, and answers.

I went through step 1-5, 3 times, for a total of 45 users and 11+ hours of videos (average 15 minutes per video).

The Solution

The result of the user testing re-confirmed what I already discovered with the first (5-people) one, but gave me more in-depth insights.
I know had the key to the solution:

1. UI/UX: Do not reinvent the wheel. Not yet at least. Reduce the learning curve by getting inspired by Yelp and Google Maps. Let it be expensive for them and cheap for us. Keep enough differentiation to brand it DK, but keep a similar flow.

2. Technology: Relevant results are the once 10± miles away from the searched area. Also, do not add on a bunch of similar categories just to fill out the page. Instead, display a better empty page listing nearby area with results. I.E. a user searches for Restaurants. There are just 3 restaurants nearby, so the page, instead of displaying results for catering, restaurant appliances and such, displays a link to restaurants close to the area searched (along the line of “No results here, but we found 10 restaurants in [city]. Want to take a look?).

3. Marketing: To differentiate ourselves, I listened to what the users were complaining about the most. I came up with some features that competitors did not have at the time:
• partnering with Uber/Lyft to being able to drive to the restaurant w/o a car (only mobile website feature).
• partnering with parking detection apps to being able to give the users more info on parking.
• use google maps API to detect traffic and route and tell the users how long it would take them to get there (without them having to do that).
• cut in half the amount of ads or users will get annoyed and leave the website.
• following the recommendation for relevancy would have been a big differentiation item.

With the results in hand, I met with the above-mentioned teams to discuss improvements, roadmap, and tasks.
We were able to set out a plan to slowly implement all of the improvements.


This extensive user testing was an eye-opening experience. I was able to understand DK on a much deeper level.
Not only that, I now had hundreds of takeaways recorded and ready to be delivered to the UI team (myself) and the IT and Marketing teams.
See Report

svgPrev Post
svgOnline Learning Platform: eUni