Global Joker · Delivery Hero · 2019–2024
Helping new restaurants get discovered in 45+ countries
Overview
What is Joker?
Joker was created to help newly onboarded restaurants get their first customers. A new restaurant has no order history, so it’s hard to get noticed. Joker gives these restaurants extra visibility through time-limited discounts, which makes it a kind of advertising product.
When I started working on it, Joker was a relatively small product. Over about five years, we rolled it out to more than 45 markets, and it became a significant revenue stream, also through the extra commission from the new customers we brought to restaurants. I was the product manager.
The problem
How do you scale one product to 45+ markets when every country and city works differently?
Much of our growth came from expansion: launch in a new market, bring in more restaurants, generate more revenue. But markets behaved very differently. Users in Singapore and Argentina don’t order the same way, and neither do people in a capital city and a small town.
Settings clearly mattered. When one market raised its daily offer limit and shortened the wait between offers, its sale rate went up 2.1%. Yet these settings could only be changed for a whole brand at once. Markets needed that control at country and city level, without waiting for a new release.
The principle
Standard where it matters, configurable where it counts
Making everything configurable would have turned Joker into 45 different products, harder to maintain, test and improve. Keeping everything the same would have made it a poor fit for most markets. So we drew a clear line.
- The core offer engine
- How offers are created and served
- Data and reporting
- How many offers users see
- Timing between offers
- Discount levels and minimum order value
- Ranking rules
One shared product for every market, with room for local differences only where they actually changed user behavior.
What I built
Settings for every country and city
I built configuration management in the admin panel, with three levels. Each level inherits from the one above and can override it, so a market only changes what’s different.
Impact
What this made possible
Experiments per country
Markets could A/B test their own settings and follow the results themselves, instead of waiting for a central team.
Discounts that fit each city
Smaller cities got lower minimum order values. Conversion went up 6.29% and Joker revenue 4.41%, while revenue per order dropped 4.9%. A trade-off worth making.
Faster launches in new countries
New markets could start with their own settings from day one, without custom development.
Outcome
Where Joker ended up
Reflection
What I learned
1Standardize the core, configure the edges.
A global product scales when the engine is shared and the local details are adjustable.
2Flexibility is a growth feature.
Letting markets adapt the product themselves helped Joker scale faster than building custom versions would have.
3Every setting is a product decision.
Each new option adds complexity, so a setting should only become configurable when markets really behave differently.