Independent demand is demand you forecast because it comes from outside, like customer orders for a finished product. Dependent demand is demand you calculate because it is derived from something else, like the parts needed to build that product. The first drives reorder-point logic; the second drives MRP.
One of the oldest mistakes in a storeroom is treating every part the same way: set a reorder point, watch it, refill it. That works fine for the finished goods a customer buys on their own schedule. It works badly for the bolts and boards that only get used when you build those finished goods. The difference is not the part, it is where the demand comes from, and that single distinction decides which planning method keeps you from either stocking out or drowning in inventory. This post explains the split and why it drives two different planning engines.
What is the difference between dependent and independent demand?
The difference is the source of the demand. Independent demand originates outside the plant and is not derived from the demand for any other item, so you cannot calculate it, you have to forecast it. Dependent demand originates inside the plant and is derived directly from the demand for a higher-level item through the bill of materials, so you do not forecast it, you calculate it. A finished good is independent. Every component, subassembly, and raw material that goes into it is dependent.
Take a simple example. A shop sells finished pumps. How many pumps customers will order next month is independent demand, nobody inside the plant controls it, so the shop forecasts it from history and market signals. But once the shop commits to building 500 pumps, the demand for 500 housings, 1,000 seals, and 2,000 bolts is not a forecast at all, it is arithmetic. Multiply the build quantity by the bill of materials and you have the exact dependent demand. Forecasting the bolts would be silly; you already know.
Why does the distinction matter for planning?
It matters because forecasting something you could have calculated adds error, and calculating something you should have forecast is impossible. If you set reorder points on components, you are forecasting demand that is actually derived, which means you carry safety stock against uncertainty that does not really exist, the demand was knowable, you just chose not to compute it. Worse, component demand is lumpy: it arrives in big batches whenever a build kicks off and is zero in between. A reorder point assumes smooth, steady usage, so it either triggers late and starves the build or triggers early and piles up stock.
The right method follows the source. Independent demand, being uncertain and continuous, calls for statistical forecasting plus a buffer, the safety stock and reorder-point world. Dependent demand, being derived and lumpy, calls for material requirements planning (MRP): explode the build schedule through the bill of materials, offset by lead time, and order exactly what each build needs when it needs it. Using the wrong engine is the single most common cause of a storeroom that is simultaneously overstocked and short.
Notice what each path assumes. The forecast path assumes you genuinely cannot know the number in advance, which is true for a customer order and false for a component. The MRP path assumes you can compute the number exactly from the build plan, which is true for a component and impossible for a customer order. Put an item on the wrong path and you fight its nature every week: buffering a bolt you could have calculated, or trying to calculate a customer whim you should have forecast.
How does MRP handle dependent demand?
MRP calculates dependent demand by working down from the finished-goods plan, level by level, through the product structure. It starts from the master production schedule the independent-demand plan for finished goods, and does a gross-to-net calculation at each level: figure the gross requirement from the parent, subtract what is already on hand or on order, and time-phase the remaining net requirement backward by the item's lead time so the order lands exactly when the build needs it.
Run as steps, the logic is short:
- Start from the finished-goods plan. Take the master production schedule as the independent-demand driver, how many finished units are planned and when.
- Explode the bill of materials. Multiply each parent quantity by its components to get the gross dependent requirement at the next level down.
- Net against inventory. Subtract on-hand stock and open orders to find the true net requirement for each item.
- Offset by lead time. Time-phase each net requirement backward by its purchase or build lead time, so the order or work order releases early enough to arrive on time.
- Repeat down every level. Feed each level's planned orders as demand into the level below, all the way to raw materials, so one finished-goods plan drives the whole structure.
The output is a time-phased plan of exactly what to order and build, and when, with no forecast and no reorder point on any dependent item. Because the demand is calculated, MRP can order lean and still hit the build, which is why it replaced reorder points for components when it spread in the 1970s.
When should you still use a reorder point?
Use a reorder point for independent-demand items and for cheap, high-volume dependent items where the arithmetic is not worth the effort. Finished goods sold to customers, spare parts, and service items are genuinely independent, their demand comes from outside, so a forecast plus a reorder point is the right tool. And for very low-value dependent parts consumed steadily, washers, labels, common fasteners, the usage is smooth enough and the cost low enough that a simple reorder point or a two-bin system beats the overhead of MRP. This is where dependent-demand theory meets practical inventory sense: not every bolt deserves a calculation.
| Attribute | Independent demand | Dependent demand |
|---|---|---|
| Source | Outside the plant (customers, market) | Inside the plant (derived from a parent item) |
| How you get the number | Forecast it | Calculate it from the BOM |
| Typical items | Finished goods, spare parts, service items | Components, subassemblies, raw materials |
| Demand pattern | Continuous, uncertain | Lumpy, tied to build schedule |
| Planning method | Reorder point, safety stock | MRP, time-phased and netted |
What do the standards and data say?
Context from standards bodies and primary data:
- The dependent-versus-independent demand distinction and its link to MRP are defined in the supply-chain body of knowledge maintained by the Association for Supply Chain Management (ASCM/APICS) which frames MRP as the calculation engine for dependent demand and forecasting as the method for independent demand.
- MRP became the dominant approach to dependent-demand planning through the educational efforts of APICS in the 1970s, when computerized bill-of-materials explosion replaced reorder points for manufactured components across much of industry.
- The inventory this logic governs is a large, live number: the U.S. Census Bureau's Manufacturing and Trade Inventories and Sales series tracks business inventories in the trillions of dollars, and using the wrong demand model, forecasting what should be calculated, is a direct driver of the excess and shortages inside that figure.
The practical point: MRP exists precisely because reorder points fail on dependent demand, and the distinction is old, settled, and still routinely ignored on the floor.
How does this connect to a manufacturing operating system?
The whole chain, forecast the independent demand, calculate the dependent demand, only works if the finished-goods plan and the bill of materials are both current and connected. A stale BOM explodes into the wrong component demand; a disconnected schedule feeds MRP numbers the floor has already overtaken. This is one reason the distinction sits at the core of a manufacturing operating system: the system has to know the real build plan and the real product structure to calculate dependent demand correctly. It is also the foundation the balancing work in demand-supply balancing and the network view in distribution requirements planning both build on, since each starts from a clean split between what is forecast and what is derived.
Where the distinction breaks in practice
In theory the split is clean; in practice it blurs whenever the data is scattered. If the bill of materials lives in one system, on-hand inventory in another, and the real build schedule in a supervisor's notebook, MRP calculates dependent demand from numbers that no longer match the floor, and planners quietly fall back to reorder points on everything because at least those they can see. The failure is not the theory, it is the disconnect between the plan, the structure, and the stockroom. Harmony is an AI-native layer that connects machines, software, and paperwork into one operational layer, with no rip-and-replace, so the build schedule, the bill of materials, and real inventory become one live record instead of three that disagree. AI search returns cited answers across those records, so a planner can ask which components a change to next week's build will short and get a grounded answer, and Harmony's digital workflows route each order and work order to the right place. It is the same paper-to-digital move Harmony makes elsewhere in the plant (see the CLS case study), and it also sharpens disciplined inventory work like ABC analysis by keeping the usage data behind it honest.