What Is a Sales Forecast?
A sales forecast is an estimate of how much your business will sell over a future period — next week, next month, next quarter — based on historical performance and whatever you already know about upcoming conditions. It isn't a guess: it's a calculation built on a method, one you can review, correct, and compare against what actually happened.
For any business that buys inventory, staffs a team, or plans cash flow, the sales forecast is the number almost every other decision depends on. Buying too much because "it felt like it would sell" and buying too little because nobody ran the numbers are two different ways of losing money for the same reason: not having a reliable sales forecast.
Why You Need a Reliable Sales Forecast
Without a sales forecast, every part of the business plans off its own version of reality. Purchasing orders inventory based on last month's sales without adjusting for season. Finance projects cash flow off the best-case scenario. Sales sets targets that sound good in a meeting but have no relationship to the actual trend.
A shared sales forecast fixes that: it gives purchasing, finance, and sales the same starting number. It doesn't eliminate uncertainty — no method does — but it does shrink the margin of error compared to deciding by gut feel, and it leaves a record of how accurate the forecast was against reality, so you can tune the method next period.
The Data You Need Before You Build a Sales Forecast
Before picking a method, you need clean data. Without this, any formula — no matter how advanced — produces a number that only looks precise:
- Sales history by period (day, week, or month, depending on how often you plan), ideally 12 to 24 periods back so you capture at least one full seasonal cycle.
- Sales by product or category, not just the total, because each line can be trending in a different direction.
- Events that distort the normal pattern: promotions, launches, stockouts, marketing campaigns, holidays, or closures.
- Relevant external variables, where they apply: seasonality in your sector, exchange rate if you sell imported goods, or the school calendar if your business depends on that cycle.
- Your current sales pipeline, if you sell B2B or work long sales cycles: open quotes, close probability, and expected close date.
The practical rule is simple: if the data isn't clean and broken out by period and by product, your sales forecast inherits that noise, no matter how advanced the method you layer on top of it.
How to Build a Sales Forecast Step by Step
- Define the period and the level of detail. Decide whether you're forecasting by week or by month, and whether you'll do it at the whole-business level, by category, or by product.
- Pull clean historical data. Export actual sales for the last several periods, without mixing returns or cancellations into net sales.
- Pick a method that matches your data volume. A business with little history doesn't need — and can't support — a machine learning model; a well-applied moving average is usually enough to start.
- Calculate the forecast and document your assumptions. Note which special events you factored in or excluded, so you can explain later why the number came out the way it did.
- Compare the forecast against actual sales once the period closes, and measure the margin of error.
- Adjust the method based on what you learned. If the error keeps repeating in the same direction — you always underestimate December, say — fix the model, not just the number.
Sales Forecasting Methods: From Simple to Advanced
There's no single correct method — there's the correct method for your history depth and how much seasonality you deal with.
Moving Average
The simplest of all sales forecasting methods: it takes the average of the last several periods and uses it as the estimate for the next one. For example, Forecast = (Sales month 1 + Sales month 2 + Sales month 3) / 3. It works well when demand is relatively stable and there's no strong seasonality, and it's the natural starting point when you're just starting to formalize your sales forecast.
Its limit is exactly that simplicity: if your business has a clear peak season — December, back-to-school, Black Friday — an unadjusted moving average will systematically underestimate it.
Seasonal Smoothing
This fixes the problem above. Instead of averaging with no context, the method calculates a seasonality index per period — how much more or less each month sells compared to the yearly average — and applies it on top of the general trend. So if December historically sells 40% above average, the December forecast starts from the trend and adds that adjustment, instead of treating it like any other month.
This method needs more history than a moving average — ideally two or three years — because the seasonality index is only reliable when it's calculated from several complete cycles, not just one.
Machine Learning
When a business has enough data volume and several variables that influence sales — price, promotions, weather, web traffic, available inventory — a machine learning model can find patterns a simple average or a seasonal index misses, because it combines several signals at once instead of a single historical series.
The tradeoff is that these models need clean data in sufficient volume to train on: applying one to a short or gappy history doesn't produce a better forecast, it produces a model that memorizes noise and fails exactly when precision matters most. That's why it makes sense to start with a moving average or seasonal smoothing, and move to machine learning once your history and operation actually justify it.
Sales Forecast Example With Moving Average
Take a business that sold $180,000, $210,000, and $195,000 over the last three months. With a simple moving average, next month's forecast is:
Forecast = ($180,000 + $210,000 + $195,000) / 3 = $195,000
If next month lines up with a known peak season — say it historically sells 15% above a regular month — you apply the seasonal adjustment on top of that result:
Adjusted forecast = $195,000 × 1.15 = $224,250
Once the month closes, you compare actual sales against the $224,250 you forecast. If actual sales came in at $230,000, your margin of error was about 2.5% — a number worth tracking month over month to see whether your sales forecast is improving or the same bias keeps repeating.
Common Mistakes When Building a Sales Forecast
- Forecasting only the total, without breaking it down by product or category. A stable total can hide one product declining while another compensates for it; that signal disappears completely at the aggregate level.
- Ignoring one-off events when building the history. If a month had an aggressive promotion or a stockout, including it unadjusted skews the average and drags the error into the following months.
- Not measuring the previous forecast's error. Without comparing what you forecast against what actually happened, there's no way to know if the method is working or if it's time to switch it.
- Jumping straight to advanced models without the history they need. A machine learning model trained on six months of data isn't more accurate than a moving average — it just looks more sophisticated.
- Updating the forecast only once a year. A forecast that never gets checked against actual sales for each period stops reflecting what's actually happening in the business.
Automating Your Sales Forecast With Dashboards
Calculating a sales forecast by hand in a spreadsheet works fine while the business is small and the data fits on one tab. Once you're managing several categories, several locations, or several sales channels, that same exercise gets slow, error-prone on formulas, and hard to share with the whole team at once.
A web dashboard connected to your point of sale or invoicing system solves that bottleneck: it recalculates the moving average or seasonal adjustment every time new sales come in, and shows the forecast next to the real history, without anyone updating a spreadsheet every Monday. It's the same logic behind a sales dashboard tracking retail KPIs: the more automated the capture, the more you can trust the decision.
If your business also manages inventory, a reliable sales forecast is also the foundation for calculating each product's reorder point, and if you use the forecast to set sales targets, it's worth connecting it to your team's OKRs instead of setting targets separately.
Frequently Asked Questions
How often should I update my sales forecast?
It depends on your business's sales cycle, but the minimum recommended cadence is once at the close of each period you forecast — weekly if you forecast by week, monthly if you forecast by month — so you can compare it against actual sales and tune the method.
How much history do I need for a reliable sales forecast?
A simple moving average only needs a few months of stable history. A seasonal adjustment needs at least one full cycle, ideally two or three years, because the seasonality index is only reliable across several comparable periods.
Is a sales forecast in a spreadsheet enough to get started?
Yes, for a small business with few categories, a spreadsheet with a moving average is a reasonable starting point. The problem shows up as categories, locations, or channels grow: at that point, a dashboard connected to your actual sales avoids the manual work and formula errors.
Does machine learning always produce a more accurate sales forecast?
Not necessarily. A machine learning model only outperforms a moving average or seasonal adjustment when there's enough clean data volume to train it on. With little history, an advanced model can end up less reliable than a well-applied simple method.
What's the difference between a sales forecast and a sales target?
A forecast is a data-based estimate of what's likely to happen; a target is a goal the business chooses to pursue, which can be equal to, above, or below the forecast. Confusing the two leads to targets that are disconnected from the business's real trend.
If your sales forecast still lives in a spreadsheet someone updates by hand every Monday, at AISDC we build custom web dashboards that connect your point of sale or invoicing system to automatic forecasts, updated in real time and visible to the whole team.