.
Case Study:
Demand Forecasting / Global Automotive Components Supplier

Driving Accurate Demand Projections for 1000s of SKUs in Parallel
Teranalytics Forecast.AI automated engine delivers best-in-class predictions
Challenge:
Inaccurate forecasting impacting profitability and inventory planning
Demand forecasting is often inaccurate because of the inherent complexity of the problem: first, it is not easy to directly forecast a metric that can be highly volatile; second, that metric is affected by multiple parameters whose influence itself varies over time. Producing a demand forecast that addresses either of these two challenges individually is bound to inaccuracies: a straight time series analysis would lose the causality/correlation effects of influential parameters, while a straight AI approach would lose the time dimensionality of the problem. Merging the two is therefore key, but is time consuming to implement ab initio. As a result, lots of the forecast produced by companies are currently created manually by a human analyst. This is both time consuming and prone to errors.
Solution:
Automated data-driven forecasting engine that automatically selects the best forecasting method for each SKU
-
- Near just-in-time inventory control slashes carrying costs: Accurage forecast allows for leaner operation both at the inventory level as well as labor planning. Subsequent add-ons can penalize over- and under- inventory differently in order to deliver optimized tactical recommendations
- Teranalytics Forecast.AI hides the complex analytics in the backend and provides the user with an intuitive and simple to use interface, with cloud-based computation for speed efficiency
- The two step approach ensures high accuracy at the desired aggregation level (at the SKU level or more broadly by geographies, by customer types, etc).
-
- Step 1: Project. Important historical parameters are projected in time using advanced time series analysis
- Step 2: Combine. The projections are incorporated into machine learning models that are self-correcting and error bound
Results:
Reduced Stock Inventory by 5% and Provided 95% Accurate Forecasts
- Improved accuracy: study cases consistently show better matching to actual data in post-forecast analysis
- Drastic reduction in processing time: forecast that use to take hours can be produced in real time
- Significant cost reductions: savings in labor time and better utilization of industry experts
- Reduction of errors: manual processes are error prone and contribute so forecast inaccuracies
- More strategic discussions: run scenarios in real time and make informed decisions with executive team
- Increased competitive advantage: better forecast, better strategic decisions, cost reduction
- Competitive advantage: organizations become more agile and more responsive to customer demand
Client Feedback:
“Thank you – this tool has quickly and significantly improved our operation.”
— Global Region Director