The Problem
Photos are super vital to Dia&Co, influencing all aspects of the customer experience and business. The box service model depends on a “back and forth” relationship with the product, so that over time we can send better boxes and hone in on the tricky, evolving aspects of a customer's style.
In the Fall of 2017 the product team decided it was time to improve our user photo submission and management system. We had launched a user gallery the prior year and had been thinking about additional community features we could build, but it became apparent that we had to fix how we managed photos site-wide before we could build infrastructure on top. At the time, only 10.7% of all converted users had uploaded a photo.
The current experience left a lot to be desired. It especially felt like the wrong approach for onboarding, when it's our first and best chance to get her comfortable using and contributing to the service. An uploaded photo early on means the stylist could better style her box, resulting in a positive experience, meaning a repeat customer. I uncovered some obvious flaws:
Single page across lead, converted, return users. There was no distinction between what the she saw at different touch points along her experience.
Poor/ opaque messaging around value/ privacy. We were not being transparent about how we were using her photos, why we needed the photos, and what controls she had to manage her privacy.
Hard to find after on-boarding. The photo page was linked from the top navigation under Your Style profile, a place where the user would not likely return after signing up, thus creating a big discovery problem.
Touch points and messaging do not align. The copy was all directed at getting her to contribute photos, but we weren't explaining what value it would provide her.
Not mobile first. Over 80% of our traffic was on mobile, this page needed to be simplified and designed for a small screen.
Research
So I started to do research and uncover why only 11% of users were uploading photos. I had a general idea about the barriers facing this demographic of women in general when it comes to photos and sharing online. Phone conversations were best because nearly all of our customers live outside of NYC. I also looked at various other software products to get a sense of best UI practices.
10 customers, 15 minute phone interviews with customers who had uploaded at least one photo:
• Determine a clear timeline of events around their photo upload behavior. ( when, what, how )
• Determine the customer’s motivations and obstacles around photos: why they uploaded them and what value they get for themselves.
• Determine customer’s knowledge of the gallery.
Comparative analysis of various products where photos are to product usage, to study UI:
• Bumble, Instagram, Facebook, Modcloth, Pinterest
INSIGHTS
One of the main points I learned was that we needed to take a more tailored, nuanced approach to the different kinds of photos we were asking her to upload. This meant the difference between a lead/ converted experience, as well as context around photos for stylists vs a photo for the gallery. ( A gallery photo needed to be approved, meet requirements, where a stylist image did not ) I also decided on a stepped approach to the UI, as to not bog them down cognitively with lots of options at once when attempting to upload.
I identified 3 main touch points when we asked for a photo, or when a user might encounter photo upload:
Share with your stylist- this is the moment during on-boarding where we ask for a photo to share with your stylist. She is new to the product, had just gone through a long and complex survey and checkout. Messaging and value are paramount to getting her to participate.
Dia Gallery- after she has ordered her first box, she might discover the gallery through our own marketing messaging, or on her own exploration. We must capitalize on this moment and leverage the community to get her comfortable and open to her own participation.
Once she gets her 1st box- after she has become a converted customer, she has several opportunities to upload more photos to both her stylist and the community. She's more likely to participate of the location of photos, upload process, and photo management are easy to use and find.
This creates 4 different photo upload types with different behaviors:
Profile photo- simple headshot, used to identify user in their profile and throughout product.
Personal photo for stylist- A private photo sent to a stylist prior to receiving a box, asked for during on-boarding.
Box item photo for stylist- A selfie image taken in one of their Dia pieces. Due to its low photo quality, it is shared privately with a stylist to show fit preferences.
Gallery photo- A selfie image taken in one of their Dia pieces. It is shared on the gallery with the broader community as a celebration of style and belonging.
Motivations and Obstacles
I broke down the user’s interview comments into motivations and obstacles. I discovered there was a good understanding about the value of photos, a desire to share them with stylists and other women, and a sense of benevolent behavior. ( they’d show others how they wore it if it helped them ) The biggest obstacle was around privacy: they did not understand who could see it. Other issues were finding the UI difficult to find and navigate, having the “right place right time” moment to take and share it, and not having someone to help get a good shot.
Recommendations
I categorized my final list of recommended changes to the current design:
Educate on use of photos. Clearly message use, value, and privacy of photos. Educate about use of photos, why we need them. Display examples of stylist-type photos by women to increase confidence/ decrease photo standard
Prompt user to upload the right photo at the right time. Really focus on making that first ask for a photo count. Simplify the experience, and get them to do it once, its the hardest. Distinct messaging for both post conversion / post box moments. Make each photo touch point unique to its situation.
Flexible but consistent UI. Simplify and systemize the UI across all photo uploads. Make it different and tailored to the photo request, but always familiar. The diagram below lays out the organization across each touchpoint.
Hypothesis
So by re-designing the experience from the ground up, we could impact several corners of the business, and essentially deepen the product. When a stylist can see a picture of who they are styling, its much easier to glean important information about that person’s shape, age, and even cultural point of view. With more photos we could start applying machine learning to identifying inventory, improving styling, consumer experience, merchandise purchasing, and more. More photos would allow for more community features: showing how others wore it, styling tips, letting her see “how it really looks on someone like me”.
Dia&Co is really careful about its AB testing and not disrupting its conversion numbers, so because this was part of the on-boarding experience we had to prove positive moves in metrics. We know that users who upload photos stay engaged longer.
By redesigning the photo upload process through better education, prompts, and UI, the number of customers uploading photos will increase,
• Better boxes result when stylists can visually see what the customer looks like and likes to wear today.
• Data can leverage machine learning to ID the style of items she’s uploaded, and ID similar products in our inventory for stylists to put in her box.
• Product can use photos to create engaging user experiences for her to explore the style of other users and increase retention.
• This should lead to an increase in: number of photos per user, and % of users uploading.
Designs
LEARNINGS
18% increase in % of users to upload a photo/ day
34% increase in new users to upload a photo
This came with a tradeoff, however. There was a 37% decrease in shareable photos, and 27% in tagged photos for the new photos being uploaded. The bulk of new photos were both private and un-tagged. I felt though that this was an acceptable result due to the fact that we were getting many more users to upload their first photo, and getting greater participation on the platform.
Next steps would be to improve our inventory tagging system site wide. The team could start to explore machine learning options for photo tagging, since its a heavy lift for users to both do it, and do it accurately. Lastly, we could improve all user messaging around photo uploads outside of the site itself to drive better engagement.