Giving customers more control
Company—Aritzia
Role—UI/UX Design
Tools—Figma
Platforms—Web
Timeframe—Sept-Oct 2022
Testing—User Interviews, A/B Test
OVERVIEW
Recommendations had been staying stagnant over the last 3 years at 7-8% CVR, which by ecommerce standards is pretty good. But we hadn’t experimented or done anything with recommendations in quite some time, and knew this would be low hanging fruit for the Personalization Squad to tackle. Any increase in CVR for recommendations had the potential of generating ~$2mill annually.
PROBLEM
Customers don’t have an easy way to see recommendations by their preferred product attributes (e.g. same colour, fabric, style, size/length or brand) of the parent product they are shopping on the product pages.
Although CVR was considerably performing steadily, engagement with recommendations on our product pages were decreasing ~2.5% over the last year.
Where we started
COMPETITIVE ANALYSIS
We looked at our top competitors in fashion that we felt excel in recommendations/personalization in general (Sephora, Nike, Farfetch and Burberry) and identified areas of opportunity that we can use for inspo.
OLD EXPERIENCE
The old experience included two rows of generic recommendations. One based on similar items to the product page they were originally looking at. The other based on items that other customers frequently purchased with that original product.
Where we explored
DESIGN WORKSHOPS
We conducted 3 workshops where myself and my PM facilitated a brainstorm on how to revamp recommendations on our product page. Each workshop included stakeholders from cross-functional teams: design, product, marketing, merchandising, D&A and engineering.
THEMES
The ideas that surfaced to the top were mentioned by multiple teams, but ultimately focused on providing more tools within recommendations for customers to better control what they see. We took these ideas into account as we entered the design exploration phase of the project.
Where we iterated
CONSIDERATIONS
Given the ideas and themes that emerged from the workshops and learnings from our competitive analysis, we still had a few things to consider:
- How would this experience look to new customers vs existing?
- How would it change from a 1:many/all strategy to a truly personalized 1:1 strategy?
- How many products to show?
- How would they display (carousel, grid, etc.)?
- Which pieces of product info are necessary?
EXPLORATION
I took some time to explore options that provided customers with control and also addressed the above questions, then created a few prototypes to test and get further clarification on the right direction to go in. We intentionally chose to reduce the rows of recommendations down from 2 to 1, to be able to better understand the variables that had the most impact on 1 row, vs measuring for 2 once we went into A/B testing.
Where we learned
USER INTERVIEWS
We held interviews with Aritzia customers to get feedback on the current recommendations experience, and compared that to the design solutions I explored. We really wanted to come out of the interviews answering the following questions
- What product info was the most important to show? Image, price, reviews, etc.
- Do customers care to see more breadth or more specific product recommendations?
- Where on the product page would make the most sense for recommendations?
LEARNINGS
- Multiple participants mentioned not fully being satisfied with recommendations, and often wished to be able to see products based on the specific same colour, size or fabric.
- "I know I only wear Tna Butter when it comes to leggings, it would be awesome if all my recommendations for these leggings just had others in the same fabric."
- Customers actually don't care for the details in recs, they want to see many images of the products all at once.
- "If I'm in a real shopping mood, just show me everything, having to slide to view more products I always found kind of annoying."
- Customers mentioned that sometimes seeing the recommendations further down on the page doesn't catch their attention and mentioned seeing something paired with the "styled with" section or within the image gallery would be preferable.
Where we landed
V1 LAUNCH
With the learnings from our user interviews, we decided to go with the solution of showing less product info and focus more on showing more items/images. We decided to stick to providing options for recommendations that specified the exact colour, fabric, size, etc.
This strategy we recognized would be more of a 1:many/all, where most customers would get this experience. We approached this with the thought that it would get personalized over time based on customer behaviours. If we start to gain signals that you shop within the same colour, fabric, etc. maybe we eventually order the list of options based on your preferences.
We also learned though past analytics that moving recommendations higher up on the page decreased overall engagement and CVR. So we decided to keep the location of recs where they were (below product info). Additionally, this would lessen the variables to fully understand the impact of the changes made once in the A/B experiment.
V2 LAUNCH
As the v1 launch is a 1:many approach to personalization, v2 focuses on what I touched on above—a 1:1 strategy, where the experience is based on that specific customer's behaviour. We explored potential future states where we order the options based on which product attribute that particular customer frequented the most when selecting which type of recommendations to view.
We further explored options to allow the ability to explicitly save this preference, but planned to test this further.
Impact
AB TEST PLAN
- Take v1 launch and test against the current recommendations experience (2 rows)
- Measure success via engagement (CTR) and conversion (CVR)
EARLY RESULTS
- The A/B Experiment launched at the beginning of Oct, early results showed an improvement of ~0.3% on CTR and ~0.5% on CVR.