Bottom-up forecast

A bottom-up forecast is built from the ground up — at SKU, channel, and location level, from each product's actual sales history — and then aggregated upward into category and business totals. The number starts where the demand actually happens and rolls up, rather than starting at the top and being divided down.

It's an unconstrained view. It estimates what demand would be if you could meet it — before stock on hand, budget limits, or supply constraints are applied. That's deliberate: you want to know true demand first, then decide separately what you can actually fulfil.

Bottom-up vs top-down forecasting

The two approaches start from opposite ends.

Bottom-up starts from item-level history and rolls up. Each SKU gets its own forecast based on how it has actually sold, and those add up to the category and total.

Top-down starts from a revenue or category target — the plan or budget — and breaks it down to lower levels. The number is set at the top and cascaded.

Most brands need both. The bottom-up forecast tells you what's likely given how products are selling; the top-line plan tells you what's required to hit the business's ambitions. The gap between them is the planning conversation: if the bottom-up lands short of the plan, that gap is the list of launches, promotions, and pushes you need to close it. Neither view alone is enough — one is grounded but unambitious, the other ambitious but ungrounded.

Types of demand forecast

It helps to be clear about which kind of forecast you're talking about, because the word covers several different things:

• Bottom-up vs top-down — who starts the number. Bottom-up starts from item history; top-down starts from the business target. This is about direction of travel.
• Consensus — the agreed forecast that lands after sales, ops, and finance reconcile their different views into a single set of numbers everyone plans against.
• Constrained vs unconstrained — whether supply, stock, and budget limits have been applied. An unconstrained forecast is pure demand; a constrained one is what you can actually deliver given your limits.

This page covers the unconstrained bottom-up view — true demand estimated from history, before any constraints are layered on.

How a bottom-up forecast can be built

There's no single method, because different products behave differently. These are common approaches — not a fixed recipe:

• Seasonal products — growth-adjusted seasonality. Take last year's weekly demand shape and adjust it for the recent growth trend, so the seasonal curve repeats but scaled to where the product is now. Cross-reference seasonality.
• Stable products — rolling averages. Where demand is steady and doesn't have a strong seasonal shape, a rolling average of recent sales is often the most reliable base.
• Uplifts layered on top. Promotions, events, and launches are added on top of the base demand, so a known spike is planned in rather than left to distort the underlying read.
• New lines with no history — similar-product forecasting. A brand-new product has nothing to roll up from, so you borrow the demand profile of a comparable existing product as a starting point until it builds its own history.
• Custom calculations. Where standard methods don't fit a particular product, a bespoke calculation can be used instead.

The right method depends on the product, and a good bottom-up forecast usually mixes several across the range. The shape of demand variability in each product is a big part of which method suits it — stable lines forecast cleanly, erratic ones need more cover and judgement.

Why build from the bottom up

Forecasting only at the total level is faster, but it hides the thing that actually trips brands up: mix. Your total can look exactly on plan while the products underneath shift — your bestseller fading, a new line surging, a channel growing faster than the rest. A bottom-up forecast catches those shifts because every product carries its own number, and the total is simply the sum of the parts. That detail is what makes the forecast usable for ordering, allocation, and forecast accuracy tracking at the level you actually plan at.