Matching supply and demand with optimised pricing

ChainCargo is a Dutch scale-up, focusing on digital freight forwarding. Its platform serves as a hub between retailers and logistics, helping retailers to find the best ways of transportation for their goods. ChainCargo has the ambition to optimize logistics through digitizing freight forwarding and reducing environmental impact in the process. They aim to reach this ambition by using data and machine learning (ML) models to automate decision-making. Resulting in an improved user experience and bringing additional value in the market of freight forwarders.

Getting multiple models running in production, ChainCargo asked Enjins to help with the scoping, and setup of a data platform, speeding up the time to production based on industry best practices.

Step 1: Create a plan for development and prioritize the use cases

In a 4-week period, we worked out a plan that contained a thorough assessment of the current data setup, a blueprint of a modern data platform, prioritize of use cases by assessing the business impact and technical feasibility, define the new positions for the data team, and create a roadmap that showed what activities are needed for realizing the use cases.

Operations: automate the order requests

• Start with FTL spot pricing as first use case

• Develop a data analytics platform

• Create first X dashboards to demonstrate value of data to the business

• Hire a data team consisting of an analytics engineer, data engineer, and data scientist with support of a product owner.

• By partnering the aforementioned goals can be achieved in 4 months

Step 2: Co develop the data platform with pricing spot quote

As a follow-up from the audit Enjins and ChainCargo decided to co-develop the data platform and proof its value by bringing the first pricing model to production. I a 4 month period the following 3 components were built.

• Make life easier not harder, choose tools with low infra maintenance and are easy to onboard new people on

• Host models as microservices to use the models in different applications

• Separate your data in 3 schemes to combine the best of both worlds, re-usability and flexibility

Much of the gains to be made in making the transport sector more efficient are in optimizing the spot market for Full Truck Loads (FTL). This is the market for non-recurring or ad-hoc transport requests. Matching these to the right carriers can avoid unnecessary empty return trips. Determining the right price for these requests is challenging, due to their unpredictable nature and dependence on seasonal traffic patterns.

Based on historical pricing data, the spot price model provides ChainCargo employees with a prediction of what price is likely for a particular route on a particular date. This helps in decisions of accepting offers from carriers and providing prompt responses to retailers requesting a spot quote.

With the data platform in place and the first model used in production, the next phase was to scale out. Adding more data, building more use cases, and maturing the platform. By collaborating we aim to speed up the development with 6 months compared to no collaboration. The main milestones are

• Add 4 new major data sources to the data platform that contain customer, shipment information, and marketing performance

• Develop 5 new dashboards to better track route, company, and ML performance over time

• Deploy 4 new use cases (mainly recommendation and pricing use cases)

• Extend the data platform with functions such as experiment tracking, CI/CD pipelines, ML retraining flow