Think of a business owner “Person A” observes a rise in sales of 15 per cent from the previous month and predicts that sales will be the same 15 per cent higher next month. Another owner "Person B" examines two years of sales data, looks at seasonal sales trends, compares marketing expenditures, and creates a forecast based on the actual data.

Both owners are using historical data. But only one wants to know what is causing the numbers.

History doesn't predict future events. It helps you to decide on what you think may happen. And that difference is important for companies that have to make decisions about inventory, cash flow, sales goals and hiring plans.

Suppose a company with an online store wants to project their sales for the upcoming month. Its income in the last six months is recorded here in Table 1:

_{Table 1: Six months sales data}

These sales have been generally increasing, although there has been no constant increase. June was less than May and September less than August.

If a business owner only considers the sales of the first and last month, they may think that sales are increasing steadily. On closer inspection, there's a more complex story.

The average month's sales for these six months was about $14,217. That provides a helpful starting point, but doesn't guarantee that $14,217 is the appropriate forecast for October.

Why not? An average reveals nothing about whether the business is increasing its sales, whether the season is coming to an end or whether the business has recently adapted its marketing or pricing policy.

The first rule of forecasting is: Don't try to predict the future until you know the past.

Three Practical Steps to Build a Better Forecast

1. Identify the pattern

Look for trends, seasonal changes, unusual spikes, and sudden declines.

In our case, the sales figure went up from $12,000 in April to $15,800 in September, which is an increase of approximately 31.7% over the five month period. This doesn't imply that the business is expanding by 31.7% each month, however. This is an overall increase from one point to another, not a monthly rate of increase.

It is a helpful step to compare the same months in the previous years. If an online store tends to sell more before the big holidays, this could be a more useful trend to focus on than just month versus month.

2. Interpret the causes of the changes

Numbers tell us what has transpired. Complementary business data may provide the answer as to why.

For instance, if the retailer ran a paid advertising campaign and had a jump in sales for that month, August, then the sales would be higher than usual. The owner should first observe if the campaign will be continued, if the cost of acquiring customers is affordable and if the customers' sales were profitable, before projecting that growth into future months.

There are several other factors to consider:

• Price changes and discounts

• Availability and shortages of products

• The marketing campaigns and customer demand.

• Seasonal events and celebrations.

• One-off orders, returns and cancelations

A six months forecast may not be reliable if the promotion was unusually large or unusually small during one of those months or if an unusually large order fell on one of the months.

3. Choose a method that fits the decision

It is possible to make a valuable prediction without relying on an elaborate statistical model. Using a method that is understandable and can be advanced.

_{Table 2: Forccasting Methods for corresponding relavant task}

From Table 1, if we take the three-month moving average (see Table 2) of July, August, and September to estimate October sales:

(15,000 + 16,200 + 15,800) / 3 = $15,667

This provides the retailer with a very basic prediction of around $15,667 for October.

It is not a promise, and implies that the past will be a good indicator of the future. October is normally a quiet period, the business should plan accordingly. The forecast might need to be adjusted if a major campaign is being planned.

Don't Just Produce One Number: Plan for Different Outcomes

One forecast can create a false sense of certainty. A more useful approach is to consider several plausible outcomes.

Using the previously calculated moving-average $15,667 as baseline, the retailer could build these illustrative scenarios:

These are planning scenarios, not statistically calculated confidence intervals. The retailer should adjust them using its own historical variability, seasonal patterns, and knowledge of upcoming events.

Even a simple range can improve a business decision. If the retailer needs to order stock, for example, it can examine whether a purchase remains affordable under the lower-sales scenario instead of assuming the expected case will definitely occur.

A good forecast does not eliminate uncertainty. It helps a business prepare for it.

How Excel Can Help

Excel is often enough to get started. A business can organise historical sales by date, check for missing records, calculate moving averages, plot trends, and compare actual results against forecasts.

For a more reliable process:

1. Maintain historical data in a consistent format containing one record for each transaction/reporting period.

2. Distinguish between actual and forecast values.

3. Document key assumptions (e.g., price changes, promotions).

4. When new data is available, compare each forecast to the actual data.

5. Discuss any mistakes in predictions that were made and change the strategy if needed.

Here, the fourth step is particularly important. A forecast should not be created once and forgotten. If the retailer predicted $15,667 but actual sales were $12,900, the difference deserves investigation. Perhaps demand weakened, a product went out of stock, or the forecasting method did not capture an important pattern.

Over time, this review process helps the business understand which assumptions are useful and which need improvement.

A Forecast Is Only Useful If It Helps You Decide

George Box did not say forecasting or it's models are useless. The actual objective is to employ a technique that is useful in the current decision, to be aware of the limitations and to revise it as and when things change.

Some small businesses need a monthly sales forecast to plan their stock. As a business expands, it may require a cash flow forecast to determine it can afford a further member of staff. A marketing team can predict the leads it expects for the purpose of budgeting marketing expenses.

Every business requires a forecast that addresses its particular question and isn't just a fancy chart.

Conclusion

Historical data becomes valuable when helping a business to get away from guesswork. Understand the performance of the past; investigate factors to explain the performance; choose an appropriate forecasting technique and make comparisons between the forecasts and actual performance.

Don't try to anticipate the future exactly. Today you need the right level of insight to make informed decisions.

The real value of forecasting is not knowing exactly what comes next. It is being better prepared for what might.