A forecast can be calculated from poor history as easily as from reliable history. The resulting supply recommendations may look precise while repeating stockouts, promotions, and incomplete lead-time assumptions. Demand planning readiness therefore starts with the data and decisions behind the calculation.
For a supply-chain director evaluating NetSuite demand planning, a useful proof is a small, representative item sample taken from history through forecast review to a supply decision. This guide explains how to choose that sample, audit its inputs, backtest a forecast, and assign the exceptions before expanding the scope.
Reorder-point planning, demand forecasting, and material requirements planning solve related but different problems. A reorder policy replenishes against an inventory threshold. A demand plan estimates requirements over future periods. MRP translates demand and supply relationships into time-phased material decisions using the relevant planning model.
Choose according to demand patterns, product structure, lead times, and the decisions planners need to make. Do not select a method simply because a feature is visible in the account. Confirm current entitlements, prerequisites, supported item types, and the intended planning workflow with the implementation team.
NetSuite's demand planning capabilities depend on enabled features and item settings. Planning approaches and recommendations also evolve, so validate the intended design in the current account rather than copying an older configuration without review.
Include a steady seller, a seasonal item, an intermittent item, a new product, and an item recently affected by stockouts. Add an assembly if manufacturing is in scope. Select more than the highest-volume products; those may have the cleanest history and conceal problems elsewhere.
Define the planning level. An item can have stable total demand but highly uneven demand by location. Combining locations may hide transfer needs or supplier constraints. Conversely, splitting a thin history too finely can produce unstable forecasts.
For each sample item, record the demand source, period length, unit of measure, item lifecycle stage, replenishment source, order constraints, and responsible planner. Use the same definitions throughout the test.
Check whether the chosen history measures orders, shipments, invoices, or another demand signal. These are not interchangeable. Shipments during a stockout can understate what customers wanted, while orders can contain cancellations and duplicate demand.
Review returns, one-off projects, promotions, product substitutions, item-code changes, and zero-demand periods. Keep the original history alongside any adjusted series. An adjustment should have a reason and an owner rather than being overwritten into an unexplained cleaned dataset.
Consider this hypothetical four-period history for one item: 100, 110, 40, and 120 units shipped. The third period contained a documented supply interruption. Sales records show an additional 60 units of unmet customer demand during that period.
The raw total is 370 and its average is 92.5 units. If the planner approves an adjusted demand value of 100 for the interrupted period, the adjusted total becomes 430 and the average becomes 107.5. The difference matters, but the adjustment is a business judgement supported by evidence, not an automatic forecast improvement.
Break lead time into stages relevant to the replenishment decision: internal approval, supplier production, transport, receiving, and any inspection or release delay. The point at which stock becomes usable may be later than its physical arrival.
Compare quoted lead time with actual performance. Review late receipts, partial deliveries, and orders with exceptional expediting. State whether the measurement uses calendar days or working days and how holidays are handled.
Keep lead-time changes separate from demand changes during the first test. If both are adjusted at once, it becomes difficult to understand which assumption caused the new supply recommendation. A planner should be able to explain why an order moved earlier or increased in quantity.
Choose a historical cutoff and create the forecast using only information available by that date. Then compare it with the following periods. Testing a forecast against the same history used to fit it can make performance appear stronger than a real planning cycle would deliver.
In a hypothetical three-period holdout, actual demand is 110, 100, and 120 units. A simple baseline predicts 100 in each period. Its absolute errors are 10, 0, and 20, totalling 30. Divide 30 by total actual demand of 330 to obtain a weighted absolute percentage error of approximately 9.09%.
A candidate forecast predicts 108, 108, and 108. Its absolute errors are 2, 8, and 12, totalling 22. The same measure is approximately 6.67%. Total forecast demand is 324 against actual demand of 330, an underforecast of 6 units.
The candidate is better on this small error measure, but three periods do not establish future performance. Repeat across representative items and planning cycles. For intermittent items or zero-demand periods, choose metrics carefully and retain the unit errors alongside percentages.
A lower forecast error is useful when it improves a relevant outcome. Test projected stockouts, late orders, excess inventory, and handling effort under the proposed replenishment rules. Include minimum order quantities, pack sizes, usable stock, open supply, and realistic supplier dates.
Check that forecasts and actual orders are not counted twice under the selected demand logic. Review the behaviour around changed customer orders and forecast consumption where applicable. Use explicit examples rather than assuming the planning engine will infer the business's intention.
Set a review threshold for recommendations that depart materially from normal quantities or dates. Thresholds should reflect the item's value, supply risk, and demand profile. A universal percentage can overreact to a low-volume item and miss an expensive absolute change elsewhere.
A planning review should end with decisions, not a large list of coloured alerts. Assign an owner and response time to each category: missing history, suspect demand, stale lead time, supplier constraint, lifecycle change, and unexpected recommendation.
Retain a short decision record with the original recommendation, planner change, explanation, and approval where required. Later, compare the result with the assumption. This creates feedback that can improve item settings and planning practice.
Roll out by item family or location once the sample behaves acceptably. Keep a defined review window and a way to restore the prior settings if the new policy causes unanticipated problems.
It depends on the demand pattern and forecast method. Seasonal behaviour needs enough comparable cycles to evaluate, while new products need other assumptions. Data quality and relevance matter as much as length.
Not automatically. Identify the missing demand signal and document any adjustment. Deleting difficult periods can distort the series and conceal the operational problem the forecast must address.
No. Material planning still depends on demand inputs, product structures, lead times, and supply constraints. Planners remain responsible for reviewing assumptions and exceptions.
Unexplained recommendations, unreliable item settings, duplicate demand, or unresolved ownership of exceptions are reasons to pause. Define acceptance criteria before the pilot so success is measurable.
CuriousRubik can discuss a scoped demand planning readiness workshop using your item history, lead times, and replenishment decisions. The first deliverable to define is a testable sample and an agreed standard for accepting its results.