Towards faster food delivery with dark stores

22 Mar 2022

Towards faster food delivery with dark stores

In the current situation of the pandemic, there is a great demand for food delivery services. 

Dark Stores are local warehouses with basic products in stock to provide faster delivery of daily consumer products, like food, drinks, medicine, etc. Inside the dark stores, employees work on assembling products into carts based on the incoming orders from the customers and passing them down for delivery. The main difference from the regular grocery store is that a dark store is not open for general public, as there are no common shelves, cashier desks and product section signs. 

Fast delivery of products is a very new and high-tech service. Reducing the total delivery time is the main aim of delivery service. For such a competitive market, even a slight decrease of the order assembly time gives a great competitive advantage. As many companies guarantee the delivery in 30-60 minutes, the order assembly should onl take minutes. In most cases, companies do not have a ready-made solution for warehouse and delivery service management. 

We are the Navigine team, with the main competences in indoor navigation and 9+ years of R&D experience in indoor positioning services. We decided to experiment and understand how our expertise can add value to this market. We believe that order fulfillment can be improved not only during the last mile delivery, but at the collection stage.

This year we launched a pilot project with a local retail center and provided a route optimization solution for order assembly. We added a tool to the existing application of the retail center’s delivery service.

As a result, we have significantly improved the average time of order assembly using smart routing. decreasing the overall order collection by 30%. In this article we are sharing how and what we did to achieve such results.

Our vision

Nowadays most warehouses are equiped with machinery, devices and software that improve efficiency and operations. For order fulfillment, usually there is an application with a catalog of SKU`s and product arrangement inside the warehouse. Order pickers carry around portable terminals with the list of products to collect, they walk all over the facility and pick up necessary products.

Despite spending less time on locating an SKU using the warehouse app rather than doing a manual search, pickers still waste time and do not perform as eddectively as they could. 

Ideally, workers shouldn’t search for the product at all. Thus the application for pickers should contain all necessary information for faster order fulfillment.

The problem of route optimization is common for delivery and cars, but it is yet to be solved for staff routing management in the new growing food delivery market. 

Our solution

Our solution is a software kit that optimizes the work of order pickers in the dark stores and warehouses.

The server processes the order and builds the optimal route for its assembly. The picker receives instructions with the optimal route on the map or exact positions of SKUs. This approach allows pickers to complete orders faster.

The planner takes into account the distance and categories of goods. With the software kit, it is possible to predict order collection time, optimize pickers work and forecast the workload of the store. 

The solution can be modified and customized by adding optimization constraints. For example, it is possible to set the picking order in such a way that frozen and heavy products get into it last, which generally improves the quality and speed of order picking.

Top benefits for the order collection process:

The benefits for the Dark Stores:

Pilot project, route optimization

We tested Navigine route optimizer software kit in collaboration with one of the major retailers in Eastern Europe. The pilot was launched in 3 dark stores as an added feature to the existing warehouse app. 

During the POC, we collected data on the movement trajectories of employees and the differences in the chosen routes of both new and experienced workers. The data showed that experienced pickers spend 15% less time for an order than the new pickers. Although not even the most experienced pickers choose the most optimal routes. Truth is, that most pickers moves are unpredictable and ineffective. After route optimization software kit was added to the pickers devices, we managed to shorten the paths of employees by 20%, for experienced pickers we achieved 10-15% improvement.

Pilot results

The solution was tested on historical data from several fulfillment centers. 


store 5005

store 4801

Number of orders (per week)



Total length of routes

1380 km

904 km

How much shorter Navigine routes on average order compared to actual



Time required to build 1 route

<0.05 sec

<0.04 sec

The most noticeable improvements happen to the orders with more than 15 items. Our system significantly reduces the length of the route, which at the same time increases the speed of assembly, and allows to work better with perishable products.

For different locations, routes were optimized by 19% (st. 5005), 23% (st. 4795) on average.

What’s next

We added a routing feature to the existing warehouse app. The solution to optimize the order assembly route showed successful results for certain categories of orders - 23% time decrease for one picker during single order completion (order up to 15 items).

We plan to modify our solution for order picking optimization, taking into account the lessons we learned. In the new version of the algorithm, we plan to add the possibility of parallel assembly of several orders. For orders with a small number of products, this will make the order assembly process two times faster.

Route management can be improved with zoning of the facility. For instance, if data shows that milk and bread are more frequently bought together in an area, it means those sections shall be placed closer to each other. Zoning can be done with the same management tool as used for routing, as it has the same optimization principles and metrics inside.

We plan to continue our piloting for several more stores, test new features and improve the solution performance.

If you want the details of the experiment, contact us on our website.

Other news