ML-driven automation and optimization is transforming Logistics

Logistics

In today’s fast-paced world, logistics and supply chains must adapt to quickly changing consumer demands. Digital transformation in this industry has opened new doors to efficiently manage operations and schedules, and gain end-to-end visibility. Logistics and supply chain operations involve dealing with a huge amount of data, and AI techniques make it easier to analyze these large volumes in a sophisticated and efficient manner.

Learn More

Use Cases

How we can leverage ML in Logistics

Demand forecasting

Route optimization

Peak-hour avoidance

Supply
chain planning

Anticipatory shipping

Automating Email Classification with NLP

About the client

Celistics is a US-based logistics company with operations in various countries across the Latin American region. They mostly cater for the mobile telecommunications market.

Challenge

By their very nature, our client receives a large volume of inquiries via email. Queries must first be classified and registered so that their agents can process them. Our client's staff spent hours manually sorting these emails. In addition, this task must be carried out in real time in order to comply with the SLAs for handling these orders.

The latter is crucial considering that 80% of customers will continue to use a business – and spend over 60% more – when they receive fast and convenient customer service.

Results

To automate this time-consuming task, we designed and implemented a classification and registration system to handle orders and complaints received by our client's account executives. By using a natural language processing engine, we were able to interpret the reason for the request and classify it automatically.

The solution was integrated with Google and Microsoft’s tools. In addition, we were capable of extracting relevant information which was automatically loaded into the CRM.

1000

Companies surveyed

hrs

Average response time

Impact

Teams that can provide quick and effective responses to requests are more likely to retain customers. In a survey of 1000 companies the average response time to handle a customer service request was 12 hours. Automations help reduce response times to minutes, eliminating bottlenecks and reducing human error.

Impact

1000

Companies surveyed

hrs

Average response time

Technologies

Golang, React, AWS, Gsuite integration.

Forecast Demand in Customer Care

About the client

PedidosYa is part of Delivery Hero and is the market leader for food delivery in Latin America. It is located in multiple countries in the region and is expanding by buying competitors like Glovo. PedidosYa currently employs more than 5,000 people and has a market valuation of over 3 billion dollars.

Challenge

The client needed to staff its call center with the necessary number of attendants so as to manage the varying demand of inbound calls, avoiding over and under-staffing.

The rate at which these calls occur depends on several operational variables that had to be analyzed in order to establish which ones were the more determinant and the potential correlation between them.

Proper demand forecasting offers organizations valuable insights regarding customer demand patterns, based on historical records of previous interactions. The critical aspect is to understand product demands from customers and how to fulfill those demands in a timely and efficient manner.

Results

Looking for the best model to forecast the demand, we benchmarked multiple models which ranged from traditional machine learning models including ARIMA models, decision trees and gradient boostings, to deep neural networks based on recurrent neural networks.

Feature engineering included the creation of new variables which were capable of identifying specific user behaviors. An assembled model fine tuned for each country.

The final model is capable of forecasting demand for six hour shifts and reached a RMSE of under 0.72 in the session rate forecast.

Improved performance

Impact

Having an adequate forecast of customer demands allows firms to optimize labor schedules and set targets based on them.

They can significantly reduce forecasting errors compared to manual alternatives. Across the industry, forecasting has been reported to improve its performance by 40-50%.

Impact

Having an adequate forecast of customer demands allows firms to optimize labor schedules and set targets based on them.

They can significantly reduce forecasting errors compared to manual alternatives. Across the industry, forecasting has been reported to improve its performance by 40-50%.

Improved performance

Technologies

Python, Data pipelines using Google’s Dataproc, Google’s Big Query, MongoDB.

Get in touch with one of our specialists.

Let's discover how we can help you

Training, developing and delivering machine learning models into production