Demand forecasting for small businesses: how to estimate what you will sell

Demand forecasting for small businesses: how to estimate what you will sell

Buying merchandise is the most repeated and most expensive decision a small business makes, and it is almost always taken by feel: you order what the owner fancies, what the supplier insists on shipping, or what is left over from last time. That approach produces two problems everyone recognizes: stockouts in peak season, when a product runs out in the very month it sells best, and overstock in the slow season, with boxes sitting in the back room, cash tied up and goods slowly losing value on the shelf.

The way out is not to buy more or buy less: it is to buy better, and buying better starts with estimating what you will sell. A simple demand forecast, built from the sales data your own business already produces, gives you a purchasing number to order by instead of a gut feeling. In this article you will see four straightforward methods, a worked example with a table, and the typical mistakes to avoid. No university formulas and no expensive software.

Why buying by feel is expensive

Think of last December: a product ran out halfway through the month and you stopped selling maybe thirty or forty units a week while the supplier took ten days to restock. That lost money never shows up on any report, but it is as real as a bad purchase. At the other extreme, the merchandise left over in January takes up shelf space and ties up capital you could use on items that actually turn.

The root problem is that demand is not guessed: it is estimated. With a little method, your own sales history turns into a number: how many units to order next month, how many boxes of each reference, when to place the order early. You do not need to be a mathematician to do it.

What a demand forecast is (and is not)

A demand forecast is an estimate of how many units of a product you will sell in a future period, based on what you already sold. It is not a crystal ball and it does not promise accuracy: it is an informed bet that reduces surprises. It is not an academic exercise either: with six to twelve months of history per product you can build a useful forecast in an afternoon, on paper, in a spreadsheet or with the tool you already use.

The golden rule is this: the forecast turns your sales history into a purchasing number. If you sold eighty units last month and ninety this month, the number says a reasonable expectation for next month is around ninety or a hundred units, not thirty and not three hundred. It sounds obvious, yet most small businesses never do it systematically.

Method 1: the simple average

The most basic starting point is the simple average: add up the sales of recent months and divide by the number of months. If a product sold twelve hundred units over the year, the monthly average is one hundred; your buying reference is to purchase around one hundred units a month, plus a small margin so you do not hit zero before the next restock.

Its strength is simplicity; its weakness is that it hides seasonality and trend changes. A yearly average of one hundred can hide a December of one hundred and eighty and a February of sixty. So the simple average is the baseline, but almost never the final number: use it as a starting point and combine it with the next methods, especially if your business has marked seasons.

Method 2: the moving average

The moving average fixes the laziness of the simple average: instead of looking at the whole year, it looks only at recent months, which best describe your current reality. It is calculated the same way, but over a short window: if you sold eighty, ninety and one hundred units over the last three months, the moving average is ninety, and that is your forecast for next month.

The three-month window is the most common in small businesses: it reacts to change without jumping at every odd week. If your sales are growing, the moving average notices and raises your purchasing number; if they are falling, it lowers them in time. When your business changes quickly, because of a new product or a shifting market, prefer the moving average over the simple average.

Method 3: the seasonal adjustment

No average helps if your business has seasons and you apply it the same way in December and March. Seasonal adjustment is simple: compare what a typical month sells against the general average. If December sells, say, one and a half times the monthly average, its factor is 1.5; if February sells sixty percent of the average, its factor is 0.6.

To apply the factor, multiply your base forecast by the factor of the month you are estimating: a base forecast of one hundred units times a December factor of 1.5 gives one hundred and fifty. With two years of history the factor gets sharper; with just one you already have a valuable clue. Write the factors down in a table by month and they become part of your buying routine.

Method 4: the judgement of the person at the counter

The number does not replace the person serving customers: it makes them better. Someone who sells every day knows things history does not tell you: that the big client orders two hundred units every time their main supplier fails, that the school around the corner places large orders in March, or that a nearby construction site will bring new people to the neighborhood.

The right way to use that judgement is to combine it: calculate the forecast with the methods above, then adjust it with what the person at the counter knows, always writing down the reason for the adjustment. If you adjusted because a large client arrived, leave a note; when you review next month you will know whether that judgement added value or not, and you will learn to calibrate it.

A worked example with numbers

Let us take a concrete product, an artisanal jam, and its actual sales over the last six months, exactly as they come out of the invoicing system. With the three-month window, the moving average can be calculated from March:

The month's moving average is the estimate you would have made at the end of the previous month: at the end of June, the forecast for July is the average of April, May and June: one hundred and ten plus one hundred and twenty plus one hundred and fifteen, divided by three, equals one hundred and fifteen units. That is your base number.

Now apply the season. If July is a mid-season month, the factor is close to one and you stay at one hundred and fifteen. If instead you were forecasting December, with a factor of 1.5 the number would be around one hundred and seventy-three units. The table below summarizes the seasonal adjustment:

Notice something important: the forecast is never a single final number; it is a base number plus a seasonal judgement, revisable every month.

Where to get the data

The data comes from your invoicing system's history, not from memory. Memory inflates the good month and forgets the bad one; records do not lie. Ask the system for a report of units sold by product or by category, month by month, and put it into a simple sheet: one row per month, one column per product or category.

If you invoice on paper or in a notebook, the work is the same: gather the sales of the last few months even if you have to count them by hand; the effort pays for itself on the first better purchase. Group the data sensibly: do not mix products with different rhythms on the same line. And if your inventory system already keeps that history in order for you, as Kardex Tauro does with an up-to-date stock card, half the work is done: you just export and look.

How to turn the forecast into a purchase order

The forecast tells you how much you will sell, but the purchase order must cover something more: what you will sell while the order is in transit, plus a cushion for the unexpected. In plain words: quantity to order equals forecast sales over the delivery time, plus safety stock, minus what you already have in the back room.

An example with the same product: the monthly forecast is one hundred and fifteen units, about twenty-seven per week. Your supplier takes two weeks to deliver, so you will sell about fifty-four units in that time. You add a safety stock of twenty units and subtract the ten you still have: you order sixty-four, rounded to sixty-five according to the supplier's packaging. Without a forecast, that order would be a lottery.

Two concepts deserve their place: lead time, which is the supplier's delivery window, and safety stock, the cushion for selling more than planned or receiving late. With forecast, lead time and safety stock you have a complete buying formula, with no sophisticated models required.

What to do when the forecast fails

The forecast will fail, and that is fine as long as you learn from the error. You forecast one hundred and sold one hundred and thirty: before adjusting wildly, ask yourself why. If it was a one-off event, a customer who bought fifty at once, a supplier promotion or a finished construction project, do not change your number: the event is over. If it was a trend, the product is gaining traction or stable new customers arrived, then yes: raise your forecast.

Measure the error every month by comparing forecast against actual sales, even by hand: an error of ten or fifteen percent is normal; an error of forty percent calls for a review of the method or the data. That monthly five-minute comparison is what separates businesses that improve their buying from those that repeat the same mistake with confidence.

Typical mistakes to avoid

• Forecasting everything with the same number. Each category has its own rhythm: milk turns weekly, gifts explode in December. Forecast by product or by category with a similar rhythm, not with one global average.
• Ignoring one-off events. A customer who bought five hundred units once is not a trend; if you feed it into the average, you inflate purchases for months. Separate the occasional from the recurring.
• Never comparing the forecast against reality. A forecast that is never checked against actual sales is just another opinion; the monthly review is what turns the method into learning.
• Confusing sales with demand. If you ran out of stock, the recorded sales are lower than real demand: a month with a stockout underestimates next month's forecast. Note the stockouts down.
• Forgetting delivery time. A perfect forecast is useless if the order arrives after the season; buy with the supplier's lead time in hand.

To close: forecasting is a habit, not a project

You do not need a degree or an enterprise forecasting system: you need fifteen minutes a month and an organized history. Calculate the moving average, adjust for season, ask the person at the counter, and turn it into a purchase order. The first time takes longer; by the third month it is routine.

And when the forecast tells you to buy, do it with the peace of mind of someone who no longer guesses: an orderly sales history, by product and by month, is the raw material of this whole method. Tools like Kardex Tauro keep that history up to date and leave you the time for what really matters: deciding better. Start with a single product this week and watch what happens to your stockouts next season.