Placing an order
Project Profile
Forecast total customer demand and orders for change based on operational data for several months for 5000 locations. Within the supply chain, all possible orders for customer requirements are predicted. The ordering process is reduced to reporting for the customer and allows the logistician dispositive freedom.
At a glance - essential project data
| Duration | From 8/1/2019 to 11/29/2019 with about 4 months of full engagement | |
|---|---|---|
| Data and Tools | Market - Logistics Sources • Order database • Customer information • Transportation database (MS-SQL) of 8 years • Weather • GEO information • Holiday calendar • Bank holidays | |
| Integration | •Web API for metadata enrichment with model generated data • Filling the order store with the predictions • Quality Reporting: Power BI • ML-Ops: Coupling with the customer systems for continuous transfer to and from our models | |
| AI Methods • Feature Engineering • Feature Importance • Deep Learning |
Engagement Use-Case
Forecast customer demand/orders change based on operational data for several months for 5000 locations.
Kundenmotivation / Lösungsansätze
Increase customer loyalty
Degree of automation for OneStop-Shop
Increase integrated planning of orders
Forcasting clarity for customer meetings/budgeting sessions
Know-your-customer
AI Approach
| AI key technology used in our solution | Processing (capturing, cleansing, unbundling) of raw data and complex data structure. Very good correlation of cyclic components and delivery frequency. | |
|---|---|---|
| Solution Approach | 3-layer Deep Learning Network with cyclic features Categorical Embeddings | |
| Project Approach | Simply agile | |
| Project Type | Operations/ML Ops | |
| ML Integration and ML Operations | • Operations-Integration API • VIsualizationPower BI and Webshop Customer One-Stop Shop • ML-Ops at Customer Infrastructure onPremise |
Insights and Details
Accurate prediction from operational data with the Deep Learning approach!
Accuracy of the approach as a check of the ACTUAL order (if necessary outliers, vacations etc...)
The forecast covers a whole year. This makes budgeting for the coming year possible
Error evaluation is essential for a correct approach. Feedback to the customer for the ideal package size becomes possible.
Recognition of a vacation coverage by order is recognizable.
