Carrefour Spain’s new HR solution, Quijote, is reshaping resource planning for 500 stores and 42,000 associates. Using advanced solutions and genetic algorithms, Quijote is simplifying workforce management and improving operations. Explore its innovative design and impact.

Resource planning in stores is like playing a giant game of Tetris, where aligning staff resources with needed workload is the daily challenge. Just as in Tetris, it’s all about adapting to unexpected blocks falling on the board. Yet, at Carrefour Spain, we’ve crafted our own cheat code, enabling us to anticipate these blocks and ensure our store teams are adequately staffed. Its name: “Quijote”.

Quijote is our new HR application, a human resources planner for integrated stores, that offers time planning for all our associates in 15-minute intervals. Powered by tailored-made algorithms, it provides legal scheduling for all store employees, complying with Spanish collective agreements and ensuring the best staffing option.

The name Quijote is a reference to Miguel de Cervantes’ famous novel, “Don Quixote”, in which the protagonist mistakes windmills for giants and decides to fight them. This story symbolizes taking on impossible challenges, much like what we felt when we began the project in 2020… but we did it! And let’s see why and how.

The birth of Quijote: Turning obsolescence into innovation

Quijote was born from teamwork between our HR department and hypermarket and supermarket operations teams. Managing staff resources in the retail sector is part of the daily life of each store. So we aimed to develop a tool that could cater to all our integrated stores. While our hypermarket teams had one, it was becoming outdated. So, we turned this liability into an opportunity.

We decided to create an in-house scheduling solution capable of efficiently managing large volumes of data. Quijote was designed to enhance the availability of shop personnel, optimize resource allocation, and streamline operations across all store formats.

Moreover, Quijote addressed the need for precise resource allocation by enabling us to understand shop volumes and allocate resources effectively in 15-minute intervals for each task. This level of precision was essential for maximizing efficiency and ensuring smooth store operations.

Quijote’s print-to-app shift: Driving change for 42,000 users

The Quijote project kicked off in 2021 with a trial run in six stores, expanding to all 500 stores by mid-2022. This rollout involved 42,000 associates, generating over 170,000 schedules and digitizing 62 million hours of demand. This marked a substantial shift in our approach to resource management.

One of the biggest changes came with digitizing calendars, freeing up time for other tasks. While some employees were unsure about digital calendars at first, accessing them through our HR App and Intranet meant no more printing and calendars were always available. This led to quick adoption of Quijote by everyone.

Quijote users appreciated its user-friendly interface and the ability to easily visualize schedules and access comprehensive reports. With over 91% of calendars sent on time and improved employee scheduling, less time was wasted adjusting plans, resulting in increased productivity. Everyone agreed Quijote was faster and more efficient, providing a clear view of employee scheduling and demand coverage.

Quijote’s data allies: Powering precision in scheduling

Crafting effective schedules within Quijote requires orchestrating various data streams to ensure precision and efficacy. To simplify, let’s categorize these parameters into three distinct categories within the scheduling framework.

Firstly, Quijote analyzes resource requirements, examining historical sales data, customer behavior patterns, and shop-specific details like operational hours and closures. This comprehensive approach enables more accurate predictions of future demand.

Secondly, Quijote incorporates employee data, including employment type, annual working hours, availability, and skills, to create schedules that meet operational needs and employee preferences. This granular approach ensures optimal workforce utilization while prioritizing employee well-being.

Finally, Quijote navigates legal restrictions per employee group, ensuring compliance with regulations governing rest structures, holiday compensation, and other labor-related matters.

Quijote’s strengths: Forecasting and adaptability

Quijote uses all this data to plan store resources at different stages. It all starts by uploading data and calculating out how many staff are needed for each task, store, and 15-minute slots (phase 1). This includes details like employee info, agreements, holidays, and skills.

Then, Quijote creates monthly or quarterly schedules according to collective agreements (phase 2). It considers things like sales forecasts, historical data, and operational requirements to ensure stores have enough staff and flexibility to meet changing needs. A monthly work calendar is legally required here. 

After schedules are made, they’re sent out via email (phase 3). Managers can tweak them by making manual adjustments before they’re finalized.

Quijote also keeps track of real-time data from store staff every day. This helps it adjust schedules as needed to make sure everything runs smoothly (phase 4).

Quijote significantly enhances store operations, empowering managers with data-driven insights for better decision-making. Whether it’s for quarterly schedules or daily adjustments, digitization is the key to simplifying the process.

Now, buckle up and get ready to dive under the hood to uncover the secrets of Quijote!

From literature to genetic algorithms: when Quijote meets Darwin

Quijote uses genetic algorithms to evolve and adapt, inspired by nature’s own methods of change. Like Darwin’s ideas, these algorithms help Quijote improve its solutions over time to overcome challenges.

Genetic algorithms operate starting with different solutions, akin to ‘individuals’ in a population. They compete, just like species in nature. Through crossover and mutation, these solutions combine and change, creating better versions. This ongoing process makes Quijote’s solutions better over time, like nature’s constant evolution.

Feeling a bit lost? Well, fear not! Just remember, Quijote employs a two-step optimization algorithm inspired by natural selection. First, it defines work days for each employee, then it outlines the hours and tasks to be performed. Let’s delve into this!

Step 1: Tackling working days

To define working days, Quijote crafts a solution while considering various constraints, such as employee availability, holidays, work contract duration, and legal terms. However, Quijote doesn’t settle for just one solution; it explores multiple solutions, known as ‘generations’, each representing a different stage of solution evolution.

In order to achieve the best result, every generation evaluates, selects, and enhances solutions using techniques such as ‘crossovers‘ and ‘mutations‘. This iterative process persists for up to 10 generations until the stop criteria are met (6 hours of execution), all with the aim of maximizing coverage.

Crossovers are like mixing genes from parents to create descendants. We select two parent solutions and exchange some of their genetic information. This process produces new ‘children’ solutions that inherit traits from both parents. For example, if one schedule has a morning shift for Employee A and an evening shift for Employee B, and another schedule has the opposite, a crossover operation might create two new schedules: one with morning shifts for both employees and the other with evening shifts for both.

Mutations introduce random changes into a solution’s genetic makeup, exploring new possibilities and preventing the algorithm from getting stuck. Just as genetic mutations add diversity to a species, mutations in Quijote’s algorithms add diversity to the population, enabling adaptation and improvement over time. For example, a mutation might involve moving the shifts of two employees on a given day to accommodate unforeseen changes in availability or workload.

Step 2: Juggling with working hours

If you understand how we handle working days, the next step is similar but focuses on working hours. The difference lies in the data we use. Quijote needs information like how many hours each person is hired for in a year, the hours they work according to their collective agreement, and the tasks they’re skilled at.

The generation process aims to meet the demands of the stores. The ultimate outcome yields the optimal work schedule, provided to each associate.

Quijote next windmills

Looking ahead, Quijote is working on some important improvements to make things run smoother. This year, we’re focusing on making necessary legal changes for the new collective agreement and adding Key Performance Indicators (KPIs) to keep track of how things are going.

For example, Quijote will give very soon:

• Information related to digitalized, delivered, displayed and pending calendars;
• Details about any changes to employees’ schedules and where those changes were made;
• Comparisons between the hours employees are supposed to work and what they actually work;
• Lists of employees who’ve had their hours extended.

We’ve also been trying to make things easier by automating inventory day planning, so there’s less manual work to do.

These initiatives show how Quijote keeps improving and finding new ideas to make things better for everyone. By constantly making things smoother and more efficient, Quijote stays ahead in organizing work and helping our stores run smoothly at Carrefour Spain.