The main inventory forecasting methods are the naive forecast, the moving average, exponential smoothing and the seasonal index, with judgement where history is missing. Each turns past sales into next month's demand. Which one fits depends on whether your sales follow a season, such as school terms or the festive season.
A forecast that lands close cuts two costs at once: sales lost to empty shelves, and the holding cost of stock that waits. Below, one product is forecast four ways, so you can see which method misses and why.
The methods at a glance
| Method | Formula | Use it when | What it misses |
|---|---|---|---|
| Naive | Next month = this month | A steady seller with little history | Any seasonal change |
| Moving average | Average of the last n months | Steady demand with random ups and downs | Peaks: it lags behind them |
| Exponential smoothing | α × latest sales + (1 − α) × last forecast | Steady demand, with more weight on recent months | Seasons and trends |
| Seasonal index | This year's level × the month's index | Sales that repeat by season | A pattern you have never seen |
| Judgement | Your estimate from suppliers, customers and events | New products with no history | Your own optimism |
The textbook Forecasting: Principles and Practice, by Rob Hyndman and George Athanasopoulos, sets out the naive and seasonal naive methods. Its chapter on simple exponential smoothing says the method suits data "with no clear trend or seasonal pattern".
One product, forecast four ways
For example, take a stationery shop in Mbarara that sells exercise books. In this example, schools reopen in February, May and September, and sales jump each time.
| Month | Books sold | Seasonal index |
|---|---|---|
| January | 600 | 0.61 |
| February | 2,400 | 2.44 |
| March | 900 | 0.92 |
| April | 500 | 0.51 |
| May | 2,000 | 2.03 |
| June | 800 | 0.81 |
| July | 600 | 0.61 |
| August | 550 | 0.56 |
| September | 1,900 | 1.93 |
| October | 700 | 0.71 |
| November | 450 | 0.46 |
| December | 400 | 0.41 |
| Total | 11,800 | 12.00 |
Each index is the month's sales divided by the average month, 983 books. February's 2.44 means February sells almost two and a half times an average month.
This January the shop sold 660 books, 10% more than last January. Here is February's forecast by each method.
| Method | Working | February forecast |
|---|---|---|
| Naive | January's sales | 660 |
| 3-month moving average | (450 + 400 + 660) ÷ 3 | 503 |
| Exponential smoothing, α = 0.3 | Run month by month from last January | 697 |
| Seasonal index | (660 ÷ 0.61) × 2.44 | 2,640 |
February's sales came in at 2,700 in this example. The first three methods missed by about 2,000 books, because none of them knows that schools reopen. The seasonal index missed by 60.
The seasonal working has two steps. First, 660 ÷ 0.61 = 1,082 is this year's level for an average month. Then 1,082 × 2.44 = 2,640 is February's share of it.
Turn the forecast into an order
A forecast counts pieces, but your wholesaler sells cartons. Order quantity = forecast + safety stock − stock on hand, rounded up to full cartons.
| Step | Working | Result |
|---|---|---|
| Forecast for February | Seasonal index | 2,640 books |
| Add safety stock | The shop's chosen buffer | + 400 |
| Less stock on hand | Counted at the end of January | − 300 |
| Books needed | 2,640 + 400 − 300 | 2,740 |
| Cartons of 200 | 2,740 ÷ 200 = 13.7, rounded up | 14 cartons, 2,800 books |
Keep stock in pieces and let your system convert cartons on the way in. Then the forecast, the shelf and the till all count the same thing.
The trap in your sales history
Sales history records what you sold, not what customers wanted. If the books ran out on 20 February last year, the 2,400 understates February's demand.
Mark every month that had a stockout before you forecast, and raise it by judgement. Otherwise each year's shortage becomes next year's forecast.
From forecast to reorder point
The forecast says how much to buy; the reorder point says when. AccountingTools gives the basic reorder level as average daily usage × lead time, and warns that it assumes usage stays steady.
| Formula | Example | Result |
|---|---|---|
| Lead time demand = average daily sales × average lead time | 20 × 10 days | 200 units |
| Safety stock = (maximum daily sales × maximum lead time) − (average daily sales × average lead time) | (30 × 12) − (20 × 10) | 160 units |
| Reorder point = lead time demand + safety stock | 200 + 160 | 360 units |
An earlier version of this page gave 140 for that safety stock; 160 is correct. For maximum lead time, use your slowest recent delivery, not the supplier's promise.
The max-and-average rule is a rule of thumb. The full method is in how to avoid overstocking and stockouts and the reorder point formula.
Which method to use
- A new product with no history: judgement. Ask the supplier how it sells elsewhere, order small, and review every week.
- A steady seller such as sugar or soap: a 3-month moving average, or exponential smoothing if recent months matter more.
- A seasonal line such as school supplies, festive goods or seed before the rains: a seasonal index built from last year.
- A line that is growing or shrinking: compare this month with the same month last year, then apply that growth as the January step did.
Forecast your A items with care and the rest roughly. The 80/20 rule in inventory shows which items are A. Forecasting is one part of inventory management for a small business; for customer demand beyond stock, see demand forecasting.
How to forecast with ArmPOS
ArmPOS doesn't forecast for you. It keeps the sales history a forecast needs and turns your reorder point into a warning.
- Export the history. Open Reports › Product sales report for last year, search the product, and click EXCEL. Each line has the date and the Qty sold. See Sales reports.
- Find the seasonal groups. Run Sales by Category one month at a time to see which groups swing.
- Build the index in a spreadsheet, laid out like the tables above.
- Enter the reorder point as the product's Stock Alert. When stock falls to it, the product shows as low stock. See Add your first products.
- Check the list every day. The Product Quantity Alerts report lists everything at or below its alert. See Reports to check every day, and turn on alerts on your phone.
- Buy by the carton, sell by the piece. A unit built on Piece, such as a box of 12, takes 12 pieces off when one box sells. See Categories, brands and units.
Frequently asked questions
What is the simplest inventory forecasting method?
The naive forecast: next month equals this month. It works for steady sellers and fails at every seasonal turn.
How much sales history do I need?
At least one full year, so every season appears once. Two years give a steadier seasonal index.
How do I forecast a new product?
Use judgement: a similar product's sales, the supplier's figures from other shops, and a small first order. Switch to a moving average once you have three months of sales.
How often should I update the forecast?
Every month, and again before each known peak such as term opening or the festive season.
Your forecast is only as good as your sales record. Create a free ArmPOS account and every sale is kept by product and date, ready to export. Shops in Uganda can start from ArmPOS in Uganda.
