Worst Planning Practices in SAP IBP — WP #1: Applying an Aggressive Forecast Model to a New SKU

Most SAP IBP problems don’t announce themselves. They show up months later — as excess inventory, a write-off, or a board asking why forecast accuracy keeps sliding. Trace it back, and it’s almost always a planning decision nobody challenged.

This is the first post in our WP #1: Worst Planning Practices in SAP IBP series — naming the mistakes that quietly derail even well-funded IBP implementations.

The Scenario

A newly launched SKU shows an early upward movement. Customers are gradually adding it to their assortment. A planner spots the trend and assigns a Trended Forecast Model. It looks like the right call. It isn’t.

Why It Fails

A Trended model applied to a SKU with minimal history has no natural anchor. It latches onto that early trajectory and projects it into the infinite horizon — compounding the slope month after month. The result: massively inflated forecasts, excess inventory build, and potential obsolescence before the SKU even matures.

What to Do Instead

For new SKUs with less than 6–12 months of history, use conservative baseline models — Level or Simple Exponential Smoothing. Apply manual overrides informed by phase-in curves or analogous SKU benchmarks. Only graduate to trend-sensitive models once the demand pattern is statistically established.

This is also where SKU lifecycle management inside IBP matters. Model assignment rules should not be static — they should evolve with the product.

Forecast model selection isn’t a one-time configuration decision. Matching the right model to the right lifecycle stage is a discipline — and getting it wrong on new SKUs is one of the fastest ways to erode inventory health.

This is exactly the kind of model governance detail we address in our SAP IBP usability consulting engagements — because lifecycle-aware model assignment rarely gets configured correctly at go-live.

Follow this series for more real-world SAP IBP configuration failures. And if your team needs a diagnostic on forecast model assignments, let’s talk.

💬 How would you handle this situation? Drop your approach in the comments — what’s worked for new SKU forecasting in your environment?