Problem:
The company needed to address inefficiencies in meal kit production that emerged following expansion in the market.
Solution:
Gousto used factory flow simulation to improve efficiency, uphold quality standards, reduce waste, and meet customer demand more effectively.
Results:
- Development timelines for routing algorithms were cut from months to days.
- Optimized order routing increased station utilization by 13%.
- Over 10 million recipe boxes were simulated with 97.5% real-world accuracy.
Introduction: from start-up to scale-up
Gousto, a leader in customizable meal kit delivery, sets itself apart from other competitors thanks to the concept of recipe boxes with precise ingredient portions. The company offers over 250 different recipes monthly.
In 2023, Gousto generated £308 million in revenue, reflecting its rapid growth from a start-up to a scale-up. With such expansion came the need to optimize its production processes, leveraging factory flow simulation. The company wanted to streamline operations to maintain quality, reduce waste, and meet customer demand efficiently.
Problem: the need for factory optimization in a complex system
Gousto's operations involve a complex production system. Custom recipe boxes are assembled through dynamic processes, including ingredient picking, box routing, and station management.
Key challenges of managing and trying to optimize such a system included:
- Station utilization inefficiencies: Underutilized stations and manual adjustments led to delays and resource wastage.
- Lengthy testing cycles: Traditional development methods for new routing algorithms were slow, requiring extensive manual validation before implementation.
- Stock and parameter mismanagement: Assumptions of perfect stock and static factory parameters often resulted in inefficiencies.
These challenges limited Gousto’s ability to scale efficiently and maintain its competitive edge in the meal kit delivery market.
Solution: factory flow simulation for production process optimization
To address these challenges, Gousto built a dynamic, self-serve simulation model in AnyLogic. This approach allowed for:
- Prominent testing and validation: Simulation enabled developers to test hundreds of scenarios safely and rapidly.
- Integration of live data: Gousto could optimize station workloads and ensure stock availability during simulations by incorporating historical and real-time data.
- Quicker release to production: By reducing development cycles from months to days, it became possible to implement approved simulation plans in real life, get feedback, and make iterations when needed.
Read a blog post about improving warehouse efficiency with simulation modeling. Discover strategies and real-world examples of companies that show the application of advanced technologies for warehouse operations optimization.
Gousto adopted AnyLogic for its factory flow simulation due to multi-platform availability (Mac, Windows, Linux) and seamless integration with Python algorithms, which are core to the company’s data-driven workflows.
Developing a complex model by a big team also required a tool that integrated well with version control software. AnyLogic Professional enabled distributed model development with features such as code review, pull requests, and continuous integration.
With an established factory simulation model, Gousto’s team could simulate and optimize the following processes:
- The movement of boxes through the pick tower to maximize throughput.
- Dynamic replenishment and consumption of stock.
- Processing times and downtimes.
with a factory simulation model in AnyLogic (click to enlarge)
Read also a white paper that introduces digital twins, their characteristics, and construction. Explore how this innovative technology is unlocking new possibilities, combining real-time data from the subject with its simulation model.
Results: a 13% improvement in station utilization
Gousto’s project serves as a perfect production optimization example. Using AnyLogic, over 10 million recipe boxes were simulated across 5,000 runs, with results validating real-world accuracy at 97.5%.
The company has already used a simulation model for over 30 different use cases aimed at factory optimization. For example, Gousto achieved a 13% improvement in station utilization by optimizing the order routing algorithm within the factory.
Looking ahead, Gousto plans to enhance its factory flow simulation modeling by extending advanced 3D animation for the end-to-end flow of stock and boxes. The company also aims to increase the granularity of events for more precise optimizations. Moreover, with such promising results in one factory, the business intends to build new simulation models for other factories.
The case study was presented by Dr. Kean Dequeant from Gousto at the AnyLogic Conference 2024.
The slides are available as a PDF.
