How do you learn business forecasting quickly and correctly without pretending that anyone can predict the future perfectly? Start with one question: what might happen next, what evidence supports the estimate and which decision will the estimate help? Business forecasting uses past observations, current information and explicit assumptions to estimate possible future outcomes. This eduKateSG Business Quick Notes lesson teaches you how to forecast sales, calculate a simple average, learn from forecast errors and test a different scenario. A useful forecast is not a promise. It helps a business prepare.
- Define the outcome: decide whether you are forecasting orders, units, revenue, stock or cash.
- Choose the horizon: state the next day, week, month or other period.
- Gather comparable data: check records, opening hours and unusual events.
- Use a simple starting method: calculate an average or a clearly explained assumption.
- Test uncertainty: examine what changes if demand is lower or higher.
- Review forecast error: compare the estimate with what actually happened and revise the method.
Teaching step 1 — state exactly what you want to predict. Imagine a small stationery stall that must order notebook kits for next Saturday. Its most recent four comparable Saturdays recorded 18, 20, 22 and 24 kits sold. The owner wants to plan inventory for the next Saturday. That is a forecast of units sold during one Saturday, not monthly revenue or profit. The starting data should use the same counting rule and reasonably comparable operating conditions. If one Saturday included a large school event and another lasted half as long, those differences matter. Before calculating anything, write the unit, time period and decision.
Teaching step 2 — calculate a simple baseline. An arithmetic mean of the four observed Saturdays is (18 + 20 + 22 + 24) ÷ 4 = 21 kits. You can use 21 kits as a simple baseline estimate, not a guaranteed outcome. The method gives equal weight to the four observations and does not automatically account for trends, holidays, promotions or a change in the shop’s hours. A forecast of 21 kits is easier to interpret when you can explain the input data and calculation. If the stall needs to purchase whole kits, the number is directly usable as a quantity, subject to appropriate inventory planning.
Teaching step 3 — measure the miss and learn from it. Suppose the following Saturday actually produces 26 kit sales. The forecast error, defined here as actual minus forecast, is 26 − 21 = +5 kits. The forecast underestimated demand by five kits. The absolute error is five kits, regardless of the chosen positive or negative sign convention. That does not prove the forecasting method is useless: all forecasts can be wrong. Ask whether an unusual event, better weather, an additional promotion or an incomplete previous record might explain the difference. Then record the result and update the next estimate rather than quietly rewriting the original forecast.
The six-box beginner forecasting card
- Question: what outcome are we trying to estimate?
- Time: for which exact day, week or month?
- Evidence: which observations and facts are available?
- Method: how are those observations turned into a forecast?
- Uncertainty: which events or assumptions could change the result?
- Review: how will we compare the estimate with the eventual outcome?
A practical forecast should fit on one page before it becomes a large spreadsheet. Write the date when it was made as well as the date it predicts. That distinction prevents hindsight from making a forecast look more accurate than it really was. For learning purposes, keep the original estimate visible beside the later actual number.
Forecast, budget and target are different
- Forecast: an estimate of what may happen under stated assumptions.
- Budget: a financial plan the organisation uses to allocate resources.
- Target: a result the organisation wants to achieve.
- Actual: what really happened in the specified period.
- Scenario: a conditional calculation using changed assumptions, not a prediction that must come true.
For example, a stall may target 30 sales, forecast 21 based on earlier Saturdays and budget for a stated purchasing plan. The three figures can be different without one necessarily being dishonest. They serve different purposes. Read Business Budgeting to see how planned revenue and expenses fit together after the sales forecast has been stated.
How to calculate a three-period moving average
A moving average updates as new observations arrive. Suppose a shop records 10, 13, 16 and 19 orders over four consecutive comparable weeks. A three-week moving-average forecast for week four, using weeks one to three, is (10 + 13 + 16) ÷ 3 = 13 orders. After observing 19 orders in week four, the next three-week moving-average forecast uses weeks two to four: (13 + 16 + 19) ÷ 3 = 16 orders. The oldest observation is dropped and the new one enters.
This method is easy to learn and reproducible, but it may lag behind a persistent upward or downward trend. A moving average may also be unsuitable when demand depends heavily on school holidays, event calendars, weather or other changing conditions. Use the formula as a baseline you can criticise, not as an authority that overrides current evidence.
How to read a forecast error
Let the sign convention be error = actual − forecast. If the forecast was 40 orders and actual sales were 35, the error is −5 orders. The estimate was five too high. If actual sales were 46, the error is +6 orders. The estimate was six too low. The absolute errors are 5 and 6 respectively. State the convention before describing positive or negative errors, since some teams use the opposite subtraction order.
For a series of forecasts, average absolute error can summarise the typical size of the misses in the sample. But a smaller error on historical data does not guarantee a better future forecast, particularly if the business changes or the data were selected to flatter one method. Check how the forecasts perform on later observations.
Turn a forecast into a useful purchasing decision
If 21 kits are forecast to sell, should the stall order exactly 21? Not automatically. It may already have six suitable kits in stock. It may also need a small buffer for uncertain demand, have supplier minimums or face the risk of unsold goods. Suppose the store has six usable kits and chooses a provisional buffer of four. A simple inventory need is forecast demand 21 + buffer 4 − stock 6 = 19 additional kits, ignoring open orders and other constraints. The buffer is an assumption requiring judgement, not a universal best practice.
A demand forecast and a purchasing decision are related but separate. The forecast estimates future activity; the purchasing plan adds stock levels, supplier lead times, cash and other constraints. If customers pre-order, those confirmed commitments may provide different evidence from historical foot traffic.
What can make a forecast unreliable?
- Missing or duplicated records: a dataset with bad counts produces a misleading baseline.
- Changing opening hours: comparing long and short operating days without adjustment.
- Special events: treating a one-off surge as ordinary demand.
- Product substitutions: comparing sales of different offers as if they were the same.
- Unobserved shortages: sales may be low because the product was unavailable, not because customers lacked interest.
- False certainty: reporting a single estimate as a guaranteed result.
Your twelve-minute business forecasting routine
- Minutes 1–2: write one measurable outcome and forecast horizon.
- Minutes 3–4: inspect a short sequence of comparable historical results.
- Minutes 5–6: calculate a baseline average and show the working.
- Minutes 7–8: prepare a lower-demand and higher-demand scenario.
- Minutes 9–10: compare one sample forecast with an actual result.
- Minutes 11–12: name one missing factor and one action the business might take.
Quick practice: a small forecast
A stall sells 12, 15 and 18 kits across three comparable days. Use the simple arithmetic average as an illustrative forecast for the next comparable day. If the next day records 14 sales, calculate the error using actual minus forecast. What can you conclude?
Answer: Baseline forecast = (12 + 15 + 18) ÷ 3 = 15 kits. Forecast error = 14 − 15 = −1 kit, so the estimate was one kit too high. This calculation describes that one prediction and its miss; it does not prove future days will stay close to 15.
Common forecasting mistakes
- Calling an estimate a promise: label uncertainty.
- Using inconsistent periods: compare like with like.
- Ignoring stock-outs: distinguish lack of supply from lack of demand.
- Changing history after the outcome: preserve the original dated forecast.
- Assuming an average covers peak demand: inspect variation and capacity.
- Never measuring errors: compare past predictions with actual outcomes.
Check your understanding
1. Are a forecast and a sales target the same? No. 2. If actual is 22 and forecast is 19, what is actual-minus-forecast error? +3 units. 3. Why is the date a forecast was made important? It preserves what was estimated before the outcome became known.
Continue the Business Quick Notes series
Begin with Business Fundamentals, then connect forecasts to Business Budgeting, Business Inventory Management, Business Cash Flow, Business Data Analysis, Business Planning and Business Key Performance Indicators.
For a broader introduction to planning decisions, see the U.S. Small Business Administration business-planning resources. All figures here are hypothetical examples; real forecasts require carefully maintained data and contextual review.
Remember: Define the outcome, show the evidence, make a simple estimate and measure the miss. Forecasting is useful because it reveals uncertainty instead of hiding it.