Using sedApta Demand Management to Optimize Resources and Reduce Waste

Why forecast accuracy, not just shop-floor discipline, decides how much material and capacity you waste

A production line does not waste material because operators are careless. It wastes material because the plan that reached the floor was already wrong. When a demand forecast misses by 10 percent or more, the consequences rarely show up as a line item labeled "forecast error." They show up as rush orders, expedited freight, idle machines between unplanned changeovers, and raw material that sits in a warehouse until it expires or gets written off. For a plant manager or production manager, this waste gets traced to the shop floor because that is where it becomes visible, even though the root cause sits one step upstream, in the plan itself. sedApta's Demand Management software addresses that root cause directly, turning forecast accuracy from a supply chain department's private metric into a resource optimization lever the whole plant can use.

Key takeaways

• Trace material and capacity waste back to forecast accuracy instead of stopping the analysis at shop-floor execution.
• Cut forecast error using statistical models, intelligent forecasting, and collaborative input instead of a static spreadsheet.
• Reduce buffer stock and safety inventory by connecting demand signals directly to resource and capacity planning.
• Free working capital currently tied up in excess raw materials and finished goods.
• Track the impact with inventory days of supply, forecast accuracy, and total supply chain cost as a percentage of revenue.
• Sequence a rollout that starts with the SKUs and production lines generating the most waste today.

Why resource waste starts with a demand problem

Manufacturers have made real progress on waste over the past two decades. Eurostat data shows that manufacturing activity in the EU generated 166 million tonnes of waste in 2022 (excluding major mineral waste), accounting for 10.4 percent of total EU waste generation and down 30.7 percent from 2004 levels (Eurostat, Waste Statistics). That decline reflects real investment in leaner processes, better equipment, and tighter quality control. It does not mean the planning side of the equation has caught up.

Most of what still gets wasted on a plant floor traces back to a decision made weeks earlier: how much to produce, when, and from which materials. When that decision is based on a demand number that turns out to be wrong, the plant absorbs the correction. Overestimate demand and raw material gets purchased, processed, and stored for orders that never arrive, tying up capacity that could have run a different, needed product. Underestimate it and the same plant scrambles: overtime shifts, rush purchase orders at a premium, and expedited freight to cover a gap that better planning would have flagged months in advance.

McKinsey's research on AI-driven forecasting puts a number on the correction available here: applying AI-driven forecasting to supply chain management can reduce forecast errors by 20 to 50 percent, which McKinsey ties to reductions in lost sales and product unavailability of up to 65 percent and warehousing cost reductions of 5 to 10 percent (McKinsey, "Stronger forecasting in operations management, even with weak data"). Those are supply chain numbers on paper, but every one of them has a physical counterpart on a production floor: less material sitting in a warehouse, fewer emergency runs, fewer idle hours waiting for the right components to show up.

Consider a discrete manufacturer running a mid-volume product line. The forecast calls for 12,000 units next month, based on a moving average of the past six months. Actual orders come in at 9,500. The plant has already pulled and staged raw material for 12,000 units, run a changeover to produce them, and allocated two shifts of labor to the schedule. The 2,500-unit gap does not disappear. It becomes finished goods sitting in a warehouse, tying up working capital and physical space, until someone decides to discount them, scrap them, or wait for a demand recovery that may or may not happen at the volume originally planned. Now run the same scenario in the other direction, with actual demand at 15,000 units instead of 9,500, and the plant absorbs it as a rush order: overtime approved on short notice, a supplier called for an expedited shipment at a premium price, and a changeover schedule rebuilt on the fly, at the expense of whatever product was supposed to run next. Neither outcome required a single operator to make a mistake. Both were set in motion by a forecast that missed by roughly the same margin McKinsey cites as the AI-correctable range.

The pattern holds across industries, but it is sharper where the product itself has a shelf life. In food and beverage manufacturing, a demand miss goes beyond tying up warehouse space and turns into spoilage and a direct write-off. Optimizing F&B supply chains with advanced SCM modules for planning and demand management looks at how combining demand and resource planning cuts that specific exposure in perishable categories. The mechanism described there, forecast accuracy feeding directly into resource allocation, is the same one this article covers for discrete and process manufacturing more broadly.

The hidden cost of forecasting errors on the plant floor

Ask a plant manager where waste comes from and the answer usually points at the floor itself: scrap rates, changeover losses, energy spikes during idle periods. Those are real and worth fixing on their own terms. But a large share of what looks like an execution problem is actually a planning problem wearing an execution costume.

Benchmarking data from ASCM and PwC's SCORmark research illustrates the gap between average and top-performing supply chains on exactly this point. Median performers ("Parity" tier) carry 63.3 days of inventory on hand; top performers ("Superior" tier) carry just 23.5 days (ASCM/PwC SCORmark benchmarking results). The SCORmark data does not isolate a single cause for that gap, but it is consistent with a broader pattern: supply chains that score higher on planning and forecasting maturity are also associated with carrying substantially less buffer stock, rather than the difference being primarily a warehouse management one. The same benchmark set shows total supply chain management cost ranging from 4.2 percent of product revenue at the median down to 2.7 percent for top performers, a gap that flows straight through to margin.

On the plant floor, that gap translates into concrete, countable events: a changeover run for a product that then sits unsold, an expedited freight invoice to cover a stockout that a better forecast would have flagged three weeks earlier, an overtime shift approved to catch up on a rush order the plan never saw coming. None of these get coded as "forecast error" in a cost report. They get coded as freight, labor, and inventory write-offs, which is precisely why the root cause is so easy to miss and so hard to budget against.

Labor absorbs a version of the same cost that rarely gets attributed correctly. An operator team scheduled for a planned production run is a fixed cost either way. An operator team pulled into overtime because a rush order landed with two days' notice, or held idle because the components for the next scheduled run have not arrived, is an avoidable one. Neither shows up on a forecast-accuracy report, and both show up on a labor cost report every single month, which is part of why the connection between the two so often goes unmade at the plant level.

sedApta's Resource & Supply Planning module addresses the plant-floor side of this gap directly. It plans against actual physical, logical, and financial constraints rather than theoretical capacity, balances demand dynamically across multiple production sites when more than one plant can absorb an order, and validates material readiness before a scheduling bottleneck becomes a live production problem. The result the module is built around is fewer emergency corrections precisely because the plan it produces does not need correcting as often.

How sedApta Demand Management improves forecast accuracy

If waste starts with a demand problem, the fix has to start there too. sedApta Demand Management is built around five capabilities, each targeting a specific way forecasts go wrong before they ever reach the floor.

Statistical forecasting generates a baseline forecast curve using algorithms that adapt to seasonality, trend, and anomalies in historical demand, replacing a flat moving average with a model that actually reflects how a product's demand behaves. Collaborative forecasting brings sales input into the same process instead of leaving it in a separate spreadsheet that planning finds out about after the fact, building one consensus number that production, procurement, and sales all work from. Intelligent Forecasting applies machine-learning models to real-time data and external variables, market shifts, consumer trends, and other demand signals, so the forecast adapts continuously rather than waiting for the next monthly cycle to catch up with reality. Scenario modeling simulates the demand impact of a product launch, a promotion, or a market shift before it happens, giving planners a chance to size resources for it in advance rather than reacting after the fact. Exception management flags forecasts that fall outside expected thresholds automatically, with suggested corrections, so a planner's attention goes to the handful of SKUs actually drifting rather than a full re-review of every line.

sedApta frames the stakes plainly on its own product page: a 10 percent forecasting error is not just a technicality, it leads to excess inventory, rush orders, and eroded margins. Each of the five capabilities above is aimed at closing that specific gap, and each closes it earlier in the process than a shop floor correction ever could.

The five capabilities are meant to work as a set rather than as a menu to pick from individually. A strong statistical baseline improves quickly once collaborative input corrects for the demand shifts a historical model cannot see coming, such as a competitor exiting a category or a new distribution channel opening. Intelligent Forecasting then keeps that improved baseline current between planning cycles, factoring in real-time and external signals so a plant is not still working from a forecast that was accurate three weeks ago but has since drifted. Scenario modeling and exception management sit on top of both, giving planners a way to test a change before committing resources to it and a way to catch the SKUs where the model itself is starting to lose accuracy before that shows up as a shortage or an overstock on the floor.

From forecast to floor: turning demand signals into resource decisions

Better forecast accuracy only reduces waste if it actually changes what gets produced, ordered, and staffed. That link, from a demand number to a resource decision, is where a lot of planning value quietly leaks away. A team can have an excellent forecast and still carry excess buffer stock simply because nobody rebuilt the safety stock policy to reflect the improved accuracy.

This is the specific problem sedApta's Demand-Driven Material Requirements Planning (DDMRP) module is built to close. Rather than sizing safety stock off a static formula that assumes worst-case demand variability everywhere, DDMRP positions strategic, dynamically managed decoupling buffers based on actual demand variability and lead time at each point in the supply chain, so material sits where variability is genuinely high and gets released where it is not. Because those buffer levels are driven by actual demand and supply signals rather than by the forecast number itself, day-to-day execution depends less on forecast precision. The buffer, not a perfectly accurate prediction, absorbs the swings a forecast alone could not catch, which is where the inventory-days gap referenced earlier starts to close.

Once the planning layer is right, the remaining waste tends to concentrate at the execution layer, energy consumed during an unplanned changeover, scrap generated by a rushed production run, both of which are easier to fix once they are not compounded by a bad plan underneath them. Energy and waste reduction on the shop floor with sedApta MES covers that execution-layer half of the problem in detail, correlating real-time production data to energy and scrap events as they happen. The two layers are meant to work together: planning-stage accuracy from Demand Management and Resource & Supply Planning helps reduce the waste caused by a wrong plan, while shop-floor visibility helps reduce the waste caused by how a correct plan gets executed.

Measuring the impact: KPIs that matter for resource optimization

None of this is worth pursuing without a way to track whether it is working. A plant or production manager evaluating a demand-driven resource optimization program should watch a small set of metrics rather than a large dashboard nobody reviews consistently.

Forecast accuracy, measured at the SKU and family level, is the leading indicator: it moves first and predicts most of what follows. Inventory days of supply, split by raw material, work in process, and finished goods, is the metric that most directly reflects whether resource decisions are catching up to demand reality or still hedging against uncertainty with extra stock. Total supply chain management cost as a percentage of product revenue captures the financial bottom line of all of the above in one number executives already track. Scrap and waste rate by product line connects the planning-side improvements back to what actually leaves the plant as unsellable material. Overtime and expedited freight spend, tracked specifically as a cost category rather than absorbed into general logistics spend, exposes the reactive costs that better forecasting is meant to eliminate.

The timing to start tracking these numbers is not neutral. Gartner projects that 70 percent of large-scale organizations will adopt AI-based forecasting to predict demand by 2030 (Gartner press release, September 2025), which means the competitive baseline for forecast accuracy is moving up industry-wide over the next several years, not staying flat. At the same time, Deloitte's 2026 manufacturing outlook reports that manufacturers expect input costs to rise by an average of 5.4 percent over the next year, with 80 percent of surveyed executives planning to direct at least 20 percent of their improvement budgets toward smart manufacturing initiatives (Deloitte, 2026 Manufacturing Industry Outlook). Rising input costs make every tonne of wasted raw material more expensive at exactly the moment competitors are investing to waste less of it.

Getting started: a phased approach to demand-driven resource optimization

A demand-driven resource optimization program fails most often when it launches everywhere at once: every SKU, every plant, every planner, on the same day. That approach spreads the improvement so thin that nobody can point to a result six months later, and it makes the inevitable rough edges of a new forecasting process visible across the entire catalog instead of a contained pilot. A phased rollout, starting with the products causing the most waste today, gives the program a chance to show a measurable result before asking for a wider commitment.

• Pull twelve months of forecast-versus-actual data at the SKU level and identify which products carry the largest, most consistent forecast error. These are the highest-waste candidates and the right place to start, not the highest-volume products by default.
• Quantify the current cost of that error in concrete terms: expedited freight invoices, overtime hours approved for rush orders, and inventory write-offs tied to expired or obsolete raw material. This becomes the baseline the program has to beat.
• Introduce statistical and collaborative forecasting for the identified product set before expanding company-wide, so sales, planning, and production all work from one consensus number for that subset first.
• Rebuild safety stock and buffer policy for the same product set once forecast accuracy improves, rather than leaving legacy buffer levels in place by default.
• Connect the improved forecast to resource and capacity planning so material allocation and scheduling decisions actually change in response to the better number, rather than staying confined to the demand plan on paper.
• Track forecast accuracy, inventory days of supply, and scrap or waste rate for the pilot product set on a fixed cadence, and compare against the baseline from step 2.
• Expand to the next tier of high-waste products once the pilot shows a measurable reduction, using the same sequence rather than a full-catalog rollout on day one.

Conclusion

Waste reduction programs that start on the shop floor and stop there address a real problem, but only half of it. The other half starts weeks earlier, in a forecast number that either reflects genuine demand or does not. Closing that gap changes what gets ordered, produced, and staffed before a single unit reaches the line, which is a cheaper place to fix a waste problem than after the material has already been converted into something nobody needs.

Manufacturers that connect forecast accuracy directly to resource and capacity decisions are doing more than cutting a supply chain metric. They are helping reduce a cost that would otherwise show up as freight, overtime, and scrap, at a moment when input costs are rising and competitors are actively investing in the same fix.

What to look at next

For the execution-layer half of this problem, real-time energy and scrap reduction on the shop floor itself, see Energy and waste reduction on the shop floor with sedApta MES. For a closer look at how the same demand and resource planning combination performs in a perishable-goods environment, see Optimizing F&B supply chains with advanced SCM modules for planning and demand management.