Every factory owner plays a guessing game. How much will we sell next month? How much resin, fabric or leather should we buy? Should we run a second shift in December?
Guess too low, and you lose orders and run short of stock. Guess too high, and cash gets stuck in a warehouse. Demand forecasting is a way to make this guess better, using data instead of feeling.
This guide explains how demand forecasting works inside an ERP, which methods suit factories in Bangladesh, and how to start without a big data team.
If you want to see the full flow from order to delivery first, read how ERP works in manufacturing.
What is demand forecasting?
Demand forecasting is an estimate of how much of each product customers will want in a future period. In a factory, this estimate feeds three big decisions:
- What to buy, and when
- What to make, and how much
- How much stock to hold
A good forecast is not perfect. It is close enough to reduce shortages and excess stock.
Why forecasting needs an ERP
You can forecast in Excel. Many people do. But three problems show up fast:
- Sales data sits in one file, stock in another, and open orders in a third
- The forecast does not connect to purchase or production, so someone retypes it
- Nobody checks how wrong the last forecast was
An ERP fixes this because the data is already together. Sales history, open orders, stock, supplier lead times and production capacity all sit in one place. The forecast flows into material planning and production, and the system can show you the forecast error after each month.
The data a forecast needs
You do not need big data. You need clean basics:
- Sales history. At least 12 months, ideally 24 or more, by item, customer group and month.
- Open orders and quotations. What is already confirmed or likely.
- Stock and in-transit. What you hold and what is coming.
- Lead times. How long suppliers take, including import and LC time.
- Calendar events. Eid, Pohela Boishakh, Durga Puja, buyer season, factory holidays.
- Promotions and price changes. These change demand.
If your item codes are messy, fix them first. A forecast on duplicate items is a forecast on noise.
See our guides on inventory management software in Bangladesh and best supply chain management software in Bangladesh.
Forecasting methods in plain English
1. Moving average
Take the average of the last 3 or 6 months and use that as next month’s forecast. Simple, and good for stable products.
Example: Last 3 months of sales: 900, 1,000, 1,100 units. Forecast: 1,000 units.
2. Weighted moving average
Give more weight to recent months. Good when demand is slowly rising or falling.
3. Exponential smoothing
The system gives recent months more weight automatically and updates after each month. A smart choice for many products without a strong season.
4. Seasonal forecast
It compares the same month across past years. If December is always 40% higher, the system applies that lift. Very useful in Bangladesh, where Eid, wedding season, winter and buyer calendars change demand.
5. Order-based forecast
For make-to-order factories, such as garments or footwear, the best forecast is often the buyer’s order book and annual programme. The ERP uses confirmed orders and buyer forecasts, then adds a buffer.
6. Collaborative forecast
Sales, key account managers and sometimes dealers or buyers give input. The ERP records it and compares it with the statistical forecast.
Most factories use a mix. Stable items get statistical forecasts. Big customers get order-based plans.
Make-to-stock vs make-to-order: which forecast fits?
| Type | Example | Forecast approach |
|---|---|---|
| Make-to-stock | Plastic chairs, tiles, packaged food | Statistical plus seasonal; keep safety stock |
| Make-to-order | Garments, special furniture | Order book and buyer programme; capacity planning |
| Assemble-to-order | Bikes with options | Forecast for common parts and sub-assemblies, assemble on order |
For discrete factories, our guide to discrete manufacturing ERP in Bangladesh explains how BOMs link with the plan.
From forecast to action inside the ERP
Here is the typical chain:
- Forecast by item and month
- Net with stock and open orders to find the real need
- Master production schedule sets what to make and when
- MRP calculates raw material needs
- Purchase requests are created with suggested dates
- Capacity check shows if machines and labour are enough
- Review by the planning team each month
This connects with production planning ERP for factories in Bangladesh and top 5 production management ERP.
Safety stock: your cushion against error
No forecast is perfect, so you hold some extra stock. Safety stock depends on two things: how uncertain the demand is, and how long the supplier takes.
A simple way to think about it:
- Stable demand and short lead time: low safety stock
- Wild demand and long import lead time: higher safety stock
For imported raw material with LC time, plan more cushion. See LC management ERP and ERP for import-export.
A good ERP calculates reorder points and suggests safety stock levels by item. Review them every quarter.
Bangladesh-specific factors to build in
- Eid cycles. Demand for many consumer goods rises before Eid, and factory output drops during the long holiday. Plan stock and labour well before.
- Export season. Garments and footwear follow buyer seasons. Use buyer programmes, not only last year’s numbers.
- Monsoon and construction season. Building materials, tiles and plywood see swings. See hardware and building materials ERP.
- Currency and import cost. A rate change can shift demand and purchase timing.
- Gas and power supply. Capacity can change, so link the forecast to realistic output.
- Lead time risk at port. Add buffer for customs and transport.
How to measure forecast accuracy
You cannot improve what you do not measure. Use two simple numbers.
Forecast error = Actual sales minus Forecast
MAPE (mean absolute percentage error) = Average of (absolute error divided by actual sales), shown as a percentage
Example:
| Month | Forecast | Actual | Error | Error % |
|---|---|---|---|---|
| Jan | 1,000 | 900 | 100 | 11% |
| Feb | 1,000 | 1,200 | 200 | 17% |
| Mar | 1,100 | 1,050 | 50 | 5% |
MAPE for these months is about 11%. A lower number is better. Track it by item group and look at the worst ones first.
Also watch bias. If you always forecast too high, you carry excess stock. If always too low, you lose sales.
A simple 90-day plan to start
Month 1: Clean and collect
- Fix item codes and units
- Load 12 to 24 months of sales history
- Pick your top 20 items, which often give 70% to 80% of revenue
Month 2: Forecast and compare
- Run a moving average or seasonal forecast for those items
- Compare with your current planning method
- Add known events such as Eid
Month 3: Connect to planning
- Link the forecast to MRP and purchase suggestions
- Set safety stock and reorder points for the top items
- Start a monthly forecast review meeting with sales, production, stores and accounts
Then add more items every month.
Common mistakes
- Forecasting by feeling only. Use data, then add judgement.
- Using sales as demand. If you ran out of stock, real demand was higher than sales.
- Ignoring one-off events. Mark big one-time orders so they do not distort the average.
- No review. A forecast that nobody checks stays wrong.
- Too many items at once. Start with the top items.
- Sales and production do not talk. Hold one monthly planning meeting.
How HR planning connects to demand
Demand also drives labour. Peak months need more shifts and overtime, and slow months need leave planning and training. When forecast and workforce data sit together, you can plan hiring and overtime early. See HR software features and modules in Bangladesh and our HRM software Bangladesh hub.
What to ask an ERP vendor about forecasting
- Which forecast methods are built in?
- Can I handle seasonality and events like Eid?
- Can I forecast by item, customer group and branch?
- Does the forecast feed MRP and purchase suggestions automatically?
- Can I see forecast accuracy reports?
- Can I override the forecast and keep a record of why?
- Does it work with my BOM levels?
For wider evaluation, use how to choose the right ERP software.
Frequently asked questions
Do I need AI for demand forecasting? Not at the start. Simple methods with clean data beat complex tools with messy data. Add advanced methods later if the basics work.
How much history do I need? Twelve months is the minimum. Twenty-four months or more helps you see seasonal patterns.
Can small factories use forecasting? Yes. Even a top-20 item forecast in a simple ERP can cut stock-outs and cash locked in stock. See SME ERP software in Bangladesh.
What if my demand is very unstable? Use shorter forecast periods, higher safety stock and more reliance on confirmed orders.
How often should I update the forecast? Monthly is a good start. Fast-moving consumer goods may need weekly updates.
Does forecasting reduce cost? It can, by reducing stock-outs, rush purchases, idle machines and excess stock. See how ERP software saves cost.
Final word
A forecast does not need to be perfect. It needs to be better than a guess, reviewed every month and connected to what you buy and make. Start with your top items and clean data, then build up. Explore our manufacturing ERP software hub and our ERP software for the manufacturing industry guide to see how planning, stock and production work together.