How to optimize the delivery of pet supplies using simulation: a Zooplus case

 Simulating parcel distribution for the delivery optimization of pet supplies

Zooplus is the leading e-commerce pet retailer in Europe. Founded as a German start-up in 1999, Zooplus has over 8 million active customers in more than 30 European countries. In 2021, its revenue was 2.39 billion euros.

In this blog post, I will tell you about how we minimize delivery time and costs using simulation modeling.

Zooplus business model

Daria Iakovleva, an associate product owner at Zooplus
Daria Iakovleva,
an associate product owner
at Zooplus

Our business model is based on shared 3PL (Third Party Logistics) services. Firstly, fulfillment centers store goods and pack parcels. Then line-haul providers ship them to hubs. And finally, last-mile services deliver parcels to customers. Zooplus doesn't physically "touch" the parcel at any step of the process.

Zooplus is an internet platform with its own branded goods that manages all product flows, starting from the supplier and finishing at the final delivery point. The usage of 3PL at each stage of the process has both advantages and obligations. These advantages are flexibility, scalability, and lower complexity. At the same time, obligations include contractual volume agreements and a more strict planning process.

Pet supplies are quite a specific category. On the one hand, customers have a strong loyalty to the brand, while on the other hand, marginality is relatively low. At the same time, the logistic costs are high due to parcel features: high volume and weight.

Besides, Zooplus guarantees 2—3 days’ delivery to most of the customers. To achieve this lead time, the company operates with a low backlog. When the fulfillment center starts production in the morning, the items for the parcel that it will pack in the evening haven't been ordered yet.

The prediction of the delivery system behavior

To find the balance between saving costs and reducing lead time, we apply several approaches. They are Gurobi-based optimization, blended (mixed) forecasting, simulation modeling, and a combination of these. Our decision on what technology to use depends on the level of planning and operating activity.

One of the most important procedures in the company is a “parcel release.” As soon as the order is in our system, we need to make dynamic decisions about where the parcel will be produced and how it will be shipped. It depends on item availability, total cost, lead time, and all other constraints.

We have created a comprehensive product at Zooplus that distributes parcels between fulfillment centers. One of its applications is an optimization engine. Its objective function is the minimization of costs, where delivery time is translated into money value. The decision is based not on the indication of a single parcel’s optimal life cycle but on the full picture.

We also need to find the balance between all fulfillment centers, delivery providers’ load stability, and real-life business contracts. The challenge is to predict how the system will behave in case of changes. These might be different costs, capacity restrictions, emergencies, higher or lower intake, etc.

The parcel distribution simulation for accurate planning

Proper and accurate prediction can't be done analytically. Only the simulation modeling approach allows us to consider the required level of detail. That's why we created a digital twin for a digital product. We simulate system behavior with different parameters and observe how the parcel distribution should change in an optimal way.

Knowing the optimal set-up, we prepare our physical entities beforehand. For example, we will book the proper amount of capacity with the second carrier in case the first increases prices. Also, we will secure additional production capacity in one fulfillment center if we know a strike will happen in another.

A “relaxed delivery” case study

Lead time is a key success factor in e-commerce. Previously, we also spread the “deliver as fast as possible” rule to all our customers. However, research made by our marketing analytics team shows that a significant share of customers don’t value fast delivery.

That’s why we introduced “relaxed delivery.” How does it work?

  1. The customer consents to a longer delivery time (up to 3 additional days) when placing the order.
  2. The order is deprioritized in the queue.
  3. "Relaxed parcels" aren’t packed and shipped as soon as possible and wait for higher truck utilization. Having more flexibility over the delivery timeline gives us more flexibility for optimization and better resource utilization. Hence, a reduced carbon footprint.

The decision of how to distribute orders between fulfillment centers is made not only on actual supply chain characteristics but also on the “steering” parameters.

This is the list of different bonuses and penalties which is based on agreements with service providers and business rules. Our “what if” analysis of how “relaxed parcels” should behave was based on the variation of 1 particular bonus.

3 options of how the “relaxed” bonus can be applied

3 options of how the “relaxed” bonus can be applied

We wanted our Digital Twin to answer the following questions:

  • How should we balance time in the queue with the delivery time?
  • How does it affect costs, lead time, and fulfillment center capacity?
  • Which countries would benefit the most?

The simulation was based on a real weekly parcel intake. Such a time frame allows for a balance between considering a range of possible incomes and model performance. The share of "relaxed parcels" was 5—20%, depending on the country.

We created several scenarios with different "steering" parameters logic and chose the model that gave the highest cost and lead-time benefit without having a negative effect on our 3PL providers.

The best simulated scenario predicted the following outcomes:

  • Release of “relaxed parcels” partially migrated to our cheapest fulfillment centers.
  • We have a cost-benefit +0.6 ppt on the network level in the model and 1—3% in real life (depending on the delivery country).
  • We observe a lead-time KPI improvement for “fast delivery” of 0.33 ppt in the model and 5—10 ppt in real life.
  • No disbalance in fulfillment center capacity according to our contractual limits.


Accurate planning plays a crucial role in the company. Cost and lead-time balance issues drive us to invest in the search for new technologies and improve existing products. Let’s see how it will affect our business in 2023.

Related case studies:

Subscribe to our monthly newsletter so we can bring our latest blog posts directly to your inbox.

Verwandte Posts