Case Study 1: Onboarding
Conception, Wireframing, Design & UX, Testing


One of the first areas I tackled when I got to Havenly was the on-boarding process. Havenly is a marketplace for interior design, and as a marketplace, the on-boarding process is critical. It’s role was to facilitate clients and designers by matching interior designers to clients.

The Problem

The previous iteration suffered from numerous problems; It had a low conversion rate, it was asking the wrong questions, it lacked a cohesive style, and it was poor at matching potential clients to designer. It also lacked the ability for the client to pick a designer, and instead went with a closed box matching system. The whole on-boarding needed to be re-conceptualized, so we started from the ground up to design an experience that was fun, engaging, and effective.

Thought Process

Primarily, the we were concerned with increasing conversion to paid design fees. Instrumental in this was displaying designers that matched a users style. The better we analyzed the clients style, the more likely they were to find a designer they liked, and the more likely they were to sign up. Thus, we focused a lot of attention on collecting the right type of information to display the best designers for a client.

The secondary goal was to create a fun and engaging experience. The more enjoyable the survey was, the more likely a user was to complete it, and eventually sign up. We focused on gamifying the style survey by asking questions in unique ways.

New Survey 1— Pick Room

I moved the room selection page from end of the survey to the front of the survey. This gave users an easy question for the first part of the survey to ease them into the process. Also, we changed from using real images to icons for the room selection, because real images come with a connotation of style, and we needed this page to be style agnostic.

New Survey 2— Style Questions

We knew from our testing that asking a user to self-identify their style was problematic. Most users don’t know their preferred interior design style and were being swayed with how much they liked the example picture, instead of how much they identified with a particular style. So instead of asking them to self-identify, we added a few questions that would build a profile for the user in order to predict their style. Our questions we’re changed to approximate different elements of a users style (modern v. traditional, messy v. clean, light v. dark), so even if they were swayed by an individual picture, we could begin to build a profile for their style type. I developed this question set via input from our interior design team. Additionally we felt the image toggle was a fun experience for users.

New Survey 3 — Store Selection

From consulting with the interior design team, I learned that one of the best analogues for a user’s style is the stores that they shop at. If you’re a fan of modern and Scandinavian designs, you might shop at Ikea, while if you’re more traditional you might shop at Restoration Hardware or Pottery Barn. Since, again, we can’t trust users to self identify their style, we used store preferences as part of our algorithm for building a user’s style profile.

New Survey 4— Pinterest Game

Finally, the best analogue for a users style is the interior design images they like. Thus, we included a fun “pinterest” like game which asked the users to choose images that appealed to them. This offered two advantages over the old method of the user choosing a single image — We got a more accurate assessment of the users style, and by collecting the images the users liked our designers instantly had a set of inspiration images to work off of, a huge win for our operations team.

Style Page

After a user signed up, we derived a style profile for them using their image selections and a basic algorithm. The algorithm was based mainly on image tags from the the users image choices. With each image that the user selected, the styles of those images were added up, and then we proportionally divided them up to display to the user. 

We used the results of the user’s “style survey” as a hook to increase signups, which was relatively effective. The problem with this was that we ended up with a lot of unqualified leads. Because of our sales cycle, however, we had the highest conversion to paid via email outreach, so this was still important.

Designer selection

Everything builds up to this — the designer selection page. The new survey was so much more accurate at determining a users style that we were able to surface matched designers for the first time ever. The designers were listed by quality of match and then randomized, not experience or qualitative rankings, because we wanted to give each designer the ability to attract clients. 

In order to give clients a sense of the designers own personal style, I developed the idea of a “moodboard” which was a collection of products in each style showing the individuality of each designer, without requiring new designers to have examples of previous work. This allowed new designers to compete for clients on a relatively even playing field with more experienced designers.


The survey overhaul was a massive project, but was also a significant success. The final result was far more functional, fun, and engaging.

The changes we made to determining the users style helped surface more accurate images and style results to the user, increasing their happiness with the process. Because we could collect images from each user, designers were now more prepared to design for a client the second a payment was received.

Most importantly, the changes I made to the survey increased conversion to paid from 1.5% to a whopping 3.5%!

See it Live