Replacing the tool that 70+ countries had given up on
Giving 70+ Markets Their Storefront Back.
Problem
Local teams in 70+ countries built their storefront category trees in a tool so broken they’d gone back to raising engineering tickets.
Role
Lead designer. Interviews across brands, shadowing the Berlin team, the redesign principle and every screen.
Scope
Every grocery and retail vendor across all Delivery Hero brands and 70+ countries.
Result
Teams could add and reorder categories with no ticket; released gradually by market; teams asking when the next release was coming.
Problem
The tool was so broken that teams had stopped using it
We knew from data and from the requests coming in: markets were asking to inject updates directly into the database. I ran interviews across brands in 70+ countries, and shadowed the Berlin team using the tool myself.
The choice was binary: either struggle with the tool, or raise a ticket to a backend engineer. Manual injection meant a request to engineering, who'd inject the data, but disabling a category with assigned products created conflicts nobody had visibility into. The back-and-forth to resolve them was long, and local teams had zero independence.
The legacy editor: a name, a badge, a drag handle, and a locale dropdown per country.
The people building the storefront couldn't see the storefront
Three problems, all structural:
- Locale: 70+ countries, each behind its own dropdown. Every edit meant a context switch.
- Hierarchy: Moving a category between levels was hard enough that people gave up.
- Visibility: The tool showed a name and a badge. None of what a shopper actually sees.
Solution
Show them what their shoppers see
Imagine sitting in the Dubai content team reviewing Talabat markets across MENA. Before: download the app and manually switch markets to see what's happening. The tool should have made that unnecessary. It didn't.
The redesign principle was simple: a category should show what's in it. The thumbnail the shopper sees. The number of subcategories under it, the products assigned to it, one click away. The local and English names on the same line. Whether you're managing products assigned to a category or impersonating a specific branch, both are easily accessible. The people building the storefront can now see the storefront.
Tree mapping: the shopper's thumbnail, local and English names, subcategory and product counts.
Products assigned to a category, one click away.
View mode: see the tree exactly as a specific branch serves it.
Give teams their independence back
Teams couldn't add categories without a complex flow, so they didn't, or they asked engineering. The fix: you search inline for existing categories, and if it's brand new, you add it inline, hit save, done. No ticket. No engineer. No waiting.
Search inline, add inline, save. No ticket.
This wasn't a feature; it was giving local teams control over their own catalog for the first time.
Outcome
The full rebuild was too big to ship. So we didn’t ship it as one.
Engineering’s estimate for the full rebuild was longer than the markets could wait. Rather than wait, the PM and I cut it into four releases: the tree with thumbnails, names, and counts first, because that was the finding (the people building the storefront couldn’t see it); inline add second; branch and impersonation view third; products assigned per category last. What I held onto across all four was both languages and the product count on every row: the two things that made a tree legible to someone who’d never opened the app.
Live across all four brands in 70+ countries. Teams could add and reorder categories with no ticket, and asked when the next release was coming.
Named best product release of the year by the assortment team.