Data-Driven
Demand
Forecasting Model

for a Meal Subscription Service

#Data Management
#Data
#WBR
Get free consultation

Duration

The project was completed over
6
months
of continuous collaboration with the client to ensure optimization of the forecasting model.

client

Our client is a
US
based
meal subscription service. They provide a variety of meal options delivered directly to customers' doors.

Tools and Technologies

tools
tools
tools
tools
tools
tools
tools
tools
tools

Challenge

The main challenge was to accurately forecast daily demand for both inventory and operational management for the client. Existing solution, used aggregated pre-order data for each meal and ship date, predicting final demand based on historical data and pre-order curves. The model also considered key features, such as whether a meal was user-selected or automatically added to the cart, to predict specific meal orders.

However, it had significant limitations as it did not account for individual user preferences or user-level features, such as the likelihood of ordering with a promo code, which are crucial for precise forecasting. To efficiently manage inventory and operations, the client needed a highly accurate demand forecasting model for the next 1-28 days.

Solution

1

Customer-Level Predictions Powered by customer data, this stage focused on predicting individual customer behaviors and their likelihood to confirm orders. This granular approach aimed to provide a more detailed and accurate forecast.

2

Aggregate Demand Adjustments Built on the predictions from the first stage, this phase incorporated additional factors such as seasonal patterns, common trends, promotional activities, and other relevant features to refine and adjust the aggregate demand forecast.

solution

Implementation

01

Data Collection:

Gathered and integrated customer data from various sources.

02

Machine Learning
Models:

Developed and trained models for customer-level predictions and aggregate demand adjustments.

03

Data Processing:

Utilized advanced data processing techniques to handle large volumes of data.

04

Forecasting:

Combined the customer-level predictions with aggregate adjustments to produce accurate demand forecasts.

Results

Optimized Stock
Control

Substantial improvements in inventory management, reducing waste and ensuring better alignment with customer demand.

Long-Term Accuracy

34%
improvement for 14-day forecast

Short-Term Accuracy

40%
improvement for the 4-day horizon

Medium-Term Accuracy

30%
improvement for the 7-day horizon

Related case studies

    Get a Free Consultation

    Let's talk about
    your project
    * You can attach up to 3 files, each up to 3MB, in doc, docx, pdf, ppt, or pptx format.