The Planning Hierarchy Problem

The Modern Demand Planning Playbook  ·  Post 1 of 4

Most forecast failures are structural, not statistical.

Ask any supply chain leader about their biggest operational headache, forecasting will often make the list. It could be too much inventory in the wrong location, not enough product where demand is surging, or revenue missed because the plan didn’t reflect reality.

The instinct may be to blame the model, but in most cases, the model isn’t the root problem. The problem is often structural: the plan is built at the wrong level of detail and is disconnected from how the business actually operates.

Modern supply chains are inherently hierarchical. A specific product exists in a specific warehouse, sold through a specific channel, to a specific account, in a specific region. Demand behaves differently at every intersection of those dimensions. A product that moves steadily at a national level might be highly seasonal in the Northeast and purely promotional-driven in the Southeast. Treating those regional conditions the same way inside a forecast model produces a forecast that’s wrong everywhere.

Planning at the Right Level

Effective demand planning requires the ability to work at multiple plan granularities simultaneously. At the top level, leaders may need to make decisions around brand portfolios and regional performance, whereas in order to execute, planners may need to see detailed individual product-warehouse forecasts. Neither planning perspective is wrong; however, they need to be consistent and synchronized with each other.

The modern forecast development approach is dynamic aggregation: the planning system maintains a single source of truth at the most granular level and automatically aggregates upward to whatever planning level is required. Any change made at a higher level, such as adjusting a brand-level forecast, automatically disaggregates back down to the detail level using rules that reflect actual demand proportions.

In practice, the planning system should allow editing at higher levels of aggregation and automatically distribute adjustments across individual warehouse-product combinations weighted to each combination’s historical share of demand.

This approach also solves a common data quality problem: sparse data. Individual product-location combinations often don’t have sufficient sales history to develop a reliable statistical forecast. By grouping similar forecast detail combinations by product type, geography, or channel behavior, planners can leverage shared sales patterns and produce more stable forecasts at the detail level required.

Combination Lifecycle Management

Every unique intersection of planning dimensions is a planning combination — for example, a specific product in a specific warehouse sold through a specific channel. The lifecycle of such combinations is one of the most underappreciated aspects of demand planning.

When a product launches in a new region, a new combination is created. When a product is discontinued, many combinations need to be deactivated, though their historical data remains valuable for informing forecasts of successor products or for category-level planning. When a distribution center closes, dependent combinations need to be transitioned carefully to avoid corrupting the forecast for downstream periods.

Organizations that treat combination management as an afterthought end up with “ghost forecasts” for discontinued products, missing forecasts for new launches, and accuracy metrics that don’t reflect reality. Tracking where there are forecasts with no sales, or sales with no forecasts, helps to find these anomalies.

Every Structural Decision Has a Cost

Here’s another planning dimension that often gets overlooked: the manner in which you structure your planning hierarchy isn’t just a technical choice. It’s a financial one. When you plan at a level of aggregation that is too high, you may obscure the inventory and margin implications of being incorrect at the detail level. A forecast that looks reasonable at the brand level can mask significant over-stock in one region and stock-outs in another. Those aren’t forecast errors in the abstract: they’re carrying costs, lost sales, and working capital tied up in the wrong places.

This is why the most sophisticated planning organizations are moving toward a total system view: rather than asking “how accurate was the forecast,” they ask “what did it cost to support this plan, and what would it have cost under different assumptions?” A demand shift, a supply delay, and promotional underperformance are forecast drivers with a quantifiable impact on the full system. Making cost implications for events visible at the right level of the hierarchy is what turns planning from a forecasting exercise into a decision system.

The Practical Implication for Leaders

Before investing in better forecasting models or more advanced analytics, you should determine if your planning hierarchy reflects the structure of the business and if the granularity allows you to act on the output. If not, then the structural foundation of your forecasting process isn’t solid, and sophisticated models will not correct it.