Project: 2-sided e-commerce experience
Role: Lead product designer, Supply division
Catawiki employs hundreds of Experts who review items submitted to the platform by sellers, and depending on the type of object, they can each get through a hundred of more submissions per day.
We knew our UI was old and increasingly showing the mounting technical and design debt, and our user testing had confirmed this was leading to issues with usability and accuracy.
At the same time, the way that Catawiki's category management worked was making things extra challenging. Each category manager had previously been allowed to set their own options for describing items, leading to inconsistencies like those below where each category asks about an item's condition differently.
The impact of making changes to the submission flow ran far and wide, and so we knew we needed to involve different people from across the business in first understanding the problems and then in developing solutions.
We brought heads together by running different workshops online, first to simply listen, then to gather pain points in a more structured way, and later to capture and develop ideas and solutions.
One of the best ways I've found to help make progress on complex issues involving multiple different people and departments, is to find simple ways to involve everyone from the start – to understand problems together and and build solutions collaboratively.
Using the Value Proposition Canvas to help capture pain points and relievers.
Crazy 8s to help bring ideas out of the different people involved.
Many of the solutions to our problems lay in humble form fields – but these are only deceptively simple, and the success of them lies in small details and micro-interactions. In order to develop these successfully therefore, we needed to be able to use them in real time, with real data.
To do this we moved pretty quickly from wireframes and static UI designs into Framer, where we could use React to build functional prototypes that we could actually use, and importantly, test with customers.
User tests helped us build confidence in the new interactions, and importantly helped us to convince the numerous stakeholders that it was ok to go ahead and implement the new field types.
User tests helped us build confidence in the new interactions, and importantly helped us to convince the numerous stakeholders that it was ok to go ahead and implement the new field types.
Some of the new fieldtypes we introduced
The changes included new visuals, copy and error feedback and modernised interactions
On a platform of this scale, any small movement in conversion up or down could mean a big financial impact, and so it was fundamental we monitor changes and work carefully to ensure any difference is indeed a result of the changes we introduced.
Working closely with product managers, front end engineers and data scientists, we were able to ascertain some small changes in metrics resulting from the revised UI.
Success had been defined at the start of the initiative as being measurable in terms of:
We were successful in two of them. Especially helpful was the reduction in the use of the 'other' field, when a customer can't find the relevant pre-defined value for their item and has to use the wild card. This makes our data messy and adds work for the Expert. A 14% reduction here was very significant.
The last one was tricky – we hadn't expected more rejected lots to result from this work, and it was difficult to theorise why this could be happening. Perhaps an increase in accuracy had led to the correct rejection of lots that might have otherwise made it though? The decision was taken to carry on monitoring this one. To be continued!
In any project like this there are ideas that don't make the final cut, in this case largely because having solved the most problematic cases the team was soon pulled onto other more pressing challenges.
Here are a few of the field types that didn't make it!
Date ranges in drop -downs were always a challenge, being different for every category of object.
Who can ever remember which is depth and which is width??
Condition ratings are also dependent on object type, so a visual device could help make sense of these for the less experienced.