Supply chains & complex forecasting problems

Connect demand forecasts to inventory, purchasing and capacity decisions.

Your challenge

Purchasing and capacity planning need forecasts over different time horizons. Seasonal variation and changes in the product range can make prediction difficult. Overestimation and underestimation have different consequences for inventory and product availability.

How IAP approaches it

IAP analyses temporal structures and compares suitable methods on held-out periods. Where appropriate, these include multiresolution and wavelet methods. Evaluation considers the planning horizon and the cost of incorrect decisions.

Project deliverables

You receive forecasts with an analysis of errors over the required planning horizon. If implemented as an application, staff can select products, investigate unusual forecasts and contribute additional information. Interfaces and updates are configured for your planning workflow.

Read the technical detail ↓

Methods and technical detail

The decision defines the forecast

Purchasing, inventory and staffing require different time horizons. Overestimation and underestimation can also have different costs. IAP aligns evaluation with the planning context and connects forecasts to the decisions they are intended to support. In a sales-forecasting prototype, forecasting methods and error measures were selected by product group and production cycle. Five error grades and interactive filters helped users examine important products or unusual forecasts. They could adjust models manually, for example to account for known marketing campaigns.

Temporal structure and changing seasonality

Time-series analysis examines seasonal patterns, changes and the available history. Wavelet methods consider several temporal resolutions. The 2021 Processes paper studies multiresolution forecasting on four datasets covering call-centre planning, electricity demand, prices and stocks.

Evaluate in chronological order

Rolling-origin comparisons separate the available past from later observations. Forecast horizon, error distributions and simple baselines help assess whether additional development effort is justified for the planning problem. The forecasting prototype developed for SW-MOTECH was delivered as a Docker container with a browser interface for local use at the customer’s site. Standardised input and output formats and parallel R computation supported repeated use without programming knowledge.

Methodological sources

Multiresolution Forecasting · 2021 ↗ mrf · R software ↗

Projects related to this service

Commissioned development · MVP

Forecast demand and support supply-chain decisions

Future product demand needed to inform planning in a supply chain with an automated warehouse.

Project details →

Project work · method evaluation

Evaluate forecasting methods for sales planning

Viessmann needed an assessment of sales-forecasting methods for strategic planning. The task was to determine which methods suited that planning problem.

Project details →

Commissioned development · MVP

Order picking and pallet requirements in warehouse logistics

Orders need to be grouped into suitable picking batches. Estimating pallet requirements also supports dispatch planning. Both tasks must account for the relevant logistical constraints.

Project details →

The right form of handover

Option 1

Prototype for your IT team

IAP investigates the problem, develops and evaluates the approach, and hands over the agreed prototype with documentation. Your IT team takes responsibility for further product development and operation.

Option 2

Application through integration

IAP develops the agreed application, including its interface and integrations, connects it to the intended workflow, and completes acceptance and handover. Your organisation or operating partner takes responsibility for subsequent operation.

Project stages and indicative timing →

Which decision do you want to improve?

In a free, non-binding 20-minute introductory discussion, we clarify your objective, starting point and a useful next step.

Request an introductory discussion