Demand Forecasting: Methods & Steps for Manufacturers
6 min read
By
TranZact Solutions Team
· Published


Demand forecasting is estimating how much of each product your customers will order in a coming period, such as next month or next quarter. Factories base it on past sales, open orders and market signals. They use judgment (sales team, key customers) or simple maths such as a moving average. A good forecast tells you what to make and what raw material to buy, so you avoid both stockouts and dead stock.
This guide covers the main methods, how a forecast becomes a purchase plan, and a worked example.
What demand forecasting means for a factory
For a manufacturer, a forecast is a number per finished item per period. It drives how much you make and how much raw material you buy. It is one part of demand management, which also covers shaping and fulfilling demand.
Demand analysis vs forecasting: Demand analysis looks back at why customers bought. Think price, season, or a big customer's own plans. Forecasting uses those findings to estimate what comes next.
Production forecast: This is how many units you plan to make. It starts from the demand forecast, then subtracts finished stock already in hand.
Types of demand forecasting
Short-term vs long-term: Short-term covers the next 1–3 months and drives production and buying. Long-term covers 1–3 years and drives capacity, machines and hiring.
Passive vs active: Passive assumes next year looks like last year. Active adds known changes, like a new customer, a price rise or a new market.
Internal vs external: Internal uses your own sales and capacity data. External looks at the market, your industry and the economy.
Predictive demand forecasting uses statistical models or machine learning on past sales and outside signals. It works only with clean, item-wise history. Most owner-led manufacturers should start with a moving average plus sales-team input.
Demand forecasting methods
Qualitative methods (judgment)
Sales-team estimate: each salesperson estimates orders from their own customers.
Collective opinion method: the sales team's estimates are added up, then reviewed by senior team members. The review trims over-optimism and adds market knowledge.
Jury of executives: the owner and senior team members agree on one number together.
Delphi method: outside experts answer in rounds, without seeing each other's names, until their views come close.
Customer surveys: you ask key buyers what they plan to order.
Collaborative forecasting: you share forecasts with key customers and suppliers and agree on one number. It suits manufacturers whose few large customers send monthly schedules.
Quantitative methods (maths)
Moving average: the average of the last few months' sales becomes next month's forecast.
Exponential smoothing: recent months count more. New forecast = last forecast + α × (last actual − last forecast), with α between 0 and 1.
Trend projection and regression: you fit a line to past sales, or link sales to a driver such as a customer's output.
Which method fits which manufacturer
Situation | Best method | Why |
|---|---|---|
New product, no history | Sales-team estimate, customer survey | No data to average yet |
Repeat items, steady demand | Moving average | Simple and stable |
Repeat items with a trend or season | Exponential smoothing, trend projection | Picks up the change faster |
Make-to-order, few big customers | Confirmed orders, collaborative forecasting | Demand comes in lumps |
Capacity or new plant | Jury of executives, Delphi | Long horizon, needs judgment |
How to forecast demand in 7 steps
Set the goal. Decide what you forecast (item or item group) and how far ahead.
Collect data. Pull 12–24 months of item-wise sales, open orders and customer schedules.
Clean it. Remove one-off orders, returns and months lost to a shutdown.
Pick a method. Use the table above.
Calculate. Run the numbers, then adjust for known changes.
Compare with actuals. Check last month's forecast against real orders.
Review monthly. Agree the next number with your sales, stores and purchase people, ideally in a sales and operations planning meeting.
From forecast to a buy plan
In the supply chain, the forecast is where planning starts. It feeds the master production schedule, which says what to make and when. The BOM turns that schedule into raw materials. Material requirement planning (MRP) then works out what to buy, which goes out as an indent and a PO.
Two more inputs keep the plan honest. Vendor lead time sets how early you must order. Safety stock covers the gap when demand or delivery slips. A reorder level tells stores when to raise the next indent.
From forecast to purchase order
Forecast demand per item
Set the production schedule
Turn the BOM into materials
Run MRP against stock
Raise indent, then PO
In TranZact, the MRP engine runs through your orders and BOMs, then tells you what to buy and when. The purchase flow runs MRP report → indent → RFQ → PO → GRN.
Worked example: a 3-month moving average
The numbers below are illustrative. Say an engineering unit sold this many units of one item:
Month | Units |
|---|---|
July (actual) | 1,000 |
August (actual) | 1,200 |
September (actual) | 1,100 |
October (forecast) | 1,100 |
The October forecast is (1,000 + 1,200 + 1,100) ÷ 3 = 1,100 units.
Now adjust it for open orders and customer schedules. Say a key customer has confirmed 200 extra units for a launch. October becomes 1,300 units. If August included a one-off order, take it out before you average.
How to check forecast accuracy
Forecast error = actual − forecast. If you forecast 1,100 and sold 1,000, the error is −100, or 10% of actual.
MAPE (mean absolute percentage error) averages that percentage across months or items. MAPE = average of (|actual − forecast| ÷ actual) × 100. WAPE divides total absolute error by total actual sales. Use WAPE when a few items carry most of your volume.
Forecast bias shows direction. Always above actual means you are building dead stock. Always below means stockouts and rush buying.
Where demand forecasting goes wrong in small manufacturing units
One big order skews the average. A one-off order inflates next month's forecast, and you buy material nobody needs.
Lead time is ignored. The forecast is right, but the steel or resin arrives three weeks too late.
Sales and stores work from different Excel sheets. Nobody agrees on the stock number, so the plan starts wrong.
No monthly review. The same error repeats because nobody compares the forecast with actuals.
Excel works well for a few items. It gets hard as items, customers and stores multiply.
Where TranZact fits
To be clear, TranZact doesn't claim a statistical demand-forecasting engine. You set the forecast, or you work from confirmed orders. TranZact handles what comes after.
Its MRP turns orders and BOMs into what to make and buy. Inventory has reorder and AI stock alerts, so low stock gets flagged. Accounts stay in Tally through a one-way sync, with no double entry.
Excel sheets vs TranZact for the buy plan
Excel sheets alone | With TranZact | |
|---|---|---|
What to buy | Worked out by hand | MRP from orders and BOMs |
Low stock | Found at the next count | Reorder and AI stock alerts |
Bill of materials | One sheet per product | Multi-level BOM |
Accounts | Re-typed into Tally | One-way sync to Tally |
The Grow plan also lists sales & purchase planning, inventory ageing & valuation and 65+ reports. See pricing for what each plan includes. TranZact is built on 7 years of learning from 10,000+ manufacturers, and goes live in 2–3 weeks.
Turn your forecast into a buy plan
A forecast only pays off when it becomes the right purchase at the right time. See how TranZact turns your orders and BOMs into what to buy and when. Don't be the bottleneck. Become the AI-Run Factory.




