Problem and task

Of course, a company of this size has an extensive warehouse management system. In addition to production, the company also has twenty warehouses for spare parts, tools, maintenance and overhead materials with a total value of 2.5 billion CZK and 250,000 items.

However, warehouses this big cause problems and Škoda Auto needed to identify duplicate items and determine the optimal stock levels so that money wasn’t “stuck” in the warehouses, eliminate problems with storage capacity, etc.

Naturally, taking on such a task is not easy: internal capacities were insufficient, and as they didn’t have enough experience with it, they were looking for someone to help them out.

Idea – hackathon

Škoda Auto is a very progressive company and they thought that they could find the right team using a hackathon – a competition where they gave companies assignments and data samples to see who performed the task best and achieved the best results.

Many data analytics companies participated in the hackathon, as no-one wanted to miss the opportunity to showcase themselves to such an important company. Revolt BI won the hackathon by a country mile when we detected 6,000 duplicate spare parts (2.4%) during the 48 hours of the event and prepared an MVP of a web app that would allow workers to subsequently work with duplicates.

Revolt BI’s convincing results subsequently prompted Škoda Auto to enter into collaboration with our company – one that has continued and expanded.

By the way – this wasn’t our only successful hackathon at Škoda Auto; we followed it up the next year with victory in a hackathon for the analysis of financial deviations.

Revolt BI’s diploma for winning the ŠKODA Smart Warehouse Hackathon.

Solution

The basis of the solution for the identification of duplicates was the NLP analysis of descriptions and classification according to the similarity of texts and other parameters. Items are deduplicated using a scoring algorithm that pre-calculates the similarity of the material to the database, while at the same time new material requirements can be compared using REST API.

In line with other requirements, we added another evaluation that works with turnover, with the aim of getting rid of items that aren’t “shifting” and, if appropriate, leaving only the minimum necessary quantity for the next period.

In this, of course, we evaluate data on current inventory and consumption by month over the last few years and classify items according to how often and over what length of time they are consumed.

The goal is to calculate how long the warehouses have stocks for and which items they can get rid of so they don’t tie up money and storage space

Depending on the type of item, they are then quantified in pieces, metres, kilogrammes and then, of course, according to price.

In addition to the calculations themselves, we, together with our partner company Cookielab, also prepared an independent web app that is integrated into the internal system for Škoda Auto’s logistics employees. It is customised in terms of both the design and the logic of the internal app control.

Result

Škoda Auto is extremely satisfied with the result. Of course, our cooperation doesn’t end there, as individual parameters are fine-tuned and additional functions are added.

Thanks to the integration of the app, there was no need to train employees. This was confirmed to us immediately after deployment, when everyone started using it straight away and praised its benefits.

In addition to the deduplication of current inventory items, the inventory purchasing process, in which procurement employees are automatically notified of the existence of the same or similar items at a warehouse when purchasing, was also improved and streamlined.

The system also gradually learns about the movements of items during the year, meaning that it can suggest to buyers the optimal order size to ensure future warehousing needs.

All of this, of course, brings additional savings by preventing the purchase of excess inventory items.