The wrong incentive at the wrong time
A delivery marketplace needs couriers when customers order — not simply more couriers on paper.
When I was leading Wolt Hungary’s operations, the courier compensation system included sizeable progressive bonuses for couriers who completed a high number of orders in a day. There was a logic to it: encourage commitment and reward couriers who spent long hours with us. But it wasn’t particularly good at directing supply to lunch and dinner peaks.
That mismatch had very practical consequences. During peaks, weak supply could lead to slower delivery, tighter delivery areas or more expensive last-minute interventions. Outside those windows, paying for extra supply did not necessarily improve the customer experience. Adding more couriers wasn’t a sufficient answer.
Could we match supply more closely to the hours when demand actually arrived — while keeping courier earnings understandable and the economics under control?
Start with the behaviour, not the spreadsheet
We first looked at the existing payment mix, delivery patterns and how individual courier groups would be affected. The aim was to move more compensation into predictable base and distance components while introducing higher pay during selected peak windows.
I worked with the team on modelling the alternatives in Looker and Excel, testing what the same couriers would have earned under the proposed structure. Average results alone were not enough. A model that looked good for the business could still feel like a pay cut for its most active couriers.
Use clear lunch and dinner windows rather than presenting a technically sophisticated but hard-to-explain real-time pricing engine.
Simulate earnings across courier cohorts and assess hours on duty as well as completed orders.
Identify the highly active couriers who might lose out and design transitional support rather than pretending the change affected everyone equally.
Brief support teams, explain the new rules to couriers, and collect feedback through surveys and existing community channels.
This was surge pricing by pre-defined peak periods, not a continuously adapting algorithm. That choice was deliberate: it could be implemented, understood and operated with the tools we had.
Make it work on a Thursday
Pricing changes do not become real when the formula is agreed. They become real when couriers understand the change, pick different hours, and the marketplace actually performs better.
We set target cost per task, defined peak windows and fee levels, ran earnings simulations, ensured enough courier supply was available for the transition and prepared the communication. We also used feedback channels — including support conversations, surveys and courier communities — to spot confusion quickly.
Then we watched the operational data. The original review compared comparable weekday performance before and after the introduction, using live marketplace dashboards. It wasn't a controlled trial, and I wouldn't describe it as one. But it gave us an immediate view of whether the new incentive structure was moving the right indicators.
After the Budapest implementation, I helped bring the approach to other Wolt markets. That required more than sending over an Excel file: the mechanics had to be understood and adapted to different courier economics and local operating realities.
The numbers, and what they mean
In the Budapest post-implementation review, deliveries per courier hour increased from 2.3 to 2.7, while cost per delivery fell from HUF 825 to HUF 800.
Pre/post figures from an internal Budapest project review. They are observed outcomes, not a causal estimate of the pricing change in isolation.
More importantly for me, the project showed how a local operations team could identify a systemic problem, develop a workable intervention and help turn it into a repeatable approach for a larger organisation.
What I would do again
I would still start with the incentives and the people affected by them, not with a more complicated model. We were changing the economics of people’s working days. If the logic was too clever to explain, it was probably too clever to operate well.
The other lesson was that a successful local pilot is only half a project. The hard part of scaling is making the idea understandable, measurable and adaptable enough for teams with different constraints to use it.
Data-backed operational design, commercial trade-offs, stakeholder communication and the ability to scale an intervention beyond the market where it was first built.