Most SMEs order stock the same way: a glance at last month’s sales, a feel for how things are going, and a number that sounds about right. The result is predictable — money tied up in the wrong products gathering dust, while the items customers actually want sit on backorder. Demand forecasting is how you replace that guesswork with a defensible estimate of what you’ll need and when, and the good news is it doesn’t require a data scientist, expensive software or a statistics degree. It requires the data you already have and a bit of structure.
What demand forecasting actually is (and why gut-feel ordering fails)
Demand forecasting is simply the practice of using past behaviour and known future events to estimate how much of each product you’ll sell over a given period. That estimate then drives what you buy, how much safety stock you hold and when you reorder.
Gut-feel ordering fails for reasons that have nothing to do with how experienced you are:
- Memory is biased. You remember the dramatic stockout and the customer who shouted, not the twelve quiet weeks in between. Your mental average is skewed by the exceptions.
- Last month is a single data point. One month tells you almost nothing about trend or seasonality. A good month before a slow season leads you to over-order right into a downturn.
- It doesn’t scale. You might hold a decent feel for your top ten lines. Nobody holds an accurate feel for four hundred SKUs across three branches.
Forecasting fixes this not by being clever, but by being consistent. It applies the same logic to every product, every cycle, and it writes the reasoning down so you can check whether it worked.
The data you already have
You do not need to collect anything new to start. Almost everything a basic forecast needs is already sitting in your sales and purchasing records:
- Sales history — units sold per product, per period (weekly or monthly). This is the backbone. Aim for at least 12 months so you can see a full seasonal cycle; 24 is better.
- Trend — is the product growing, flat or declining over time? A line that’s clearly climbing needs a different forecast from one that’s fading.
- Seasonality — the repeating pattern across the year. Cold drinks in summer, stationery before the school term, gifts before the holidays.
- Promotions and events — when you ran a discount, a bundle or a campaign, sales spiked. Those spikes are not “normal demand” and must be flagged, or they’ll pollute every future forecast.
- Lead times — how long between placing an order and having sellable stock on the shelf. A forecast is useless if it doesn’t account for how far ahead you must commit.
The single most common data problem isn’t missing data — it’s unclean history. If a huge promo week or a one-off bulk order is left in your baseline, your forecast will quietly assume that spike is the new normal. Before forecasting, mark those events so you can separate ordinary demand from the unusual.
Simple forecasting methods you can actually use
You can get 80% of the benefit from three methods that fit in a spreadsheet.
Moving average
Take the average of the last few periods. If you sold 100, 120 and 110 units over the last three months, your three-month moving average forecast for next month is (100 + 120 + 110) / 3 = 110 units.
It’s stable and easy, but it treats a sale from three months ago as equal to last week’s — which isn’t true if demand is changing.
Weighted moving average
Give recent periods more influence, because they’re more representative of where demand is heading. Assign weights that add up to 1 — say 0.5 to the most recent month, 0.3 to the one before, 0.2 to the oldest:
- Most recent month: 110 units × 0.5 = 55
- Previous month: 120 × 0.3 = 36
- Oldest month: 100 × 0.2 = 20
- Weighted forecast = 55 + 36 + 20 = 111 units
Close to the simple average here, but the moment demand starts trending the weighted version reacts far faster — which is exactly what you want when a product is taking off or dying.
Adjusting for seasonality
If your product has a repeating annual pattern, forecast the baseline first, then apply a seasonal factor. Work out each month’s factor by comparing that month’s typical sales to the yearly average. If December usually runs 60% above your monthly average, its seasonal factor is 1.6.
So if your baseline forecast is 111 units and December’s factor is 1.6:
111 × 1.6 = ~178 units for December.
The rule of thumb: use a weighted average to predict how much on a normal month, then multiply by a seasonal factor to answer which month.
Factoring in the real world
Your history tells you what usually happens. Your judgement handles what’s about to be different. A forecast should always be adjusted for known future events the data can’t see:
- Planned promotions — if you’re running a campaign, add the expected uplift on top of the baseline. Use the actual results from your last comparable promo, not a hopeful guess.
- Holidays and paydays — demand clusters around paydays and public holidays in ways that differ by market. If most of your customers are paid at month-end, your late-month forecast should reflect it.
- Supplier lead times — this is where forecasts turn into decisions. If a supplier takes six weeks, your forecast for six weeks out is what you buy today. Long lead times mean you’re always ordering to a further-out, less certain forecast — which is why safety stock exists.
- External shocks — price changes, a competitor closing, a new contract, currency swings, import delays. These won’t be in your history. Note them as manual overrides so you know later why the forecast was adjusted.
The discipline that matters: keep the statistical baseline and the human adjustment separate. When you review later, you want to know whether the maths was off or your judgement was.
Turning a forecast into action
A forecast is a number on a page until it changes what you buy. Three moves connect them:
- Set safety stock deliberately. Safety stock covers the gap between your forecast and reality over the lead time. Faster-selling and less predictable items need more; steady, slow items need less. The goal is a buffer sized to real variability, not a flat “hold two weeks of everything”.
- Time the reorder to the lead time. Reorder when stock on hand is enough to cover expected demand during the replenishment lead time, plus your safety stock. Forecast the lead-time window, not just “next month”.
- Buy to the forecast, not the fear. The hardest part is psychological. After a stockout you’ll want to over-order; after being stuck with dead stock you’ll under-order. Trust the forecast, adjust it for known events, and let it — not last week’s emotion — set the quantity.
A quick forecasting checklist
- Pull at least 12 months of unit sales per product
- Clean out promos and one-off orders from the baseline
- Pick a method: moving average, or weighted for trending lines
- Apply a seasonal factor where a pattern exists
- Adjust for known upcoming events
- Convert to a buy quantity using lead time + safety stock
- Record the forecast so you can compare it to actual
Reviewing and improving the forecast
A forecast you never check never gets better. Each cycle, put forecast next to actual and measure the gap.
The simplest useful metric is forecast error: the difference between what you predicted and what sold, as a percentage. If you forecast 100 and sold 130, that’s a 30% error, and you were short. Track it per product and watch the direction:
- Consistently forecasting too low? Your trend or seasonal factor is understated — nudge it up.
- Consistently too high? You may be carrying old promo spikes in the baseline, or a product is quietly declining.
- Wildly erratic? That product is genuinely unpredictable — the right response is more safety stock, not a more elaborate formula.
You’re not chasing a perfect forecast; perfect doesn’t exist. You’re chasing a forecast that’s less wrong than last quarter’s, and one where the errors are small enough that your safety stock absorbs them.
Why this is painful with disconnected data — and easy without it
Every step above assumes you can see three things together: what sold, what’s in stock, and what’s on order. When those live in separate places — sales in one system, stock in a spreadsheet, purchase orders in an inbox — forecasting becomes a monthly archaeology project. You export, you reconcile, you copy-paste, and by the time the numbers agree they’re already out of date. Most SMEs abandon forecasting not because it’s hard, but because assembling the data is exhausting.
When sales, inventory and purchasing share one source of truth, the friction disappears. Sales history is already clean and attributed. Current stock and open purchase orders are live. Lead times are recorded against each supplier. The forecast stops being a report you rebuild from scratch and becomes a view that’s always current — so reordering to a forecast is a decision you make in minutes, not a project you dread.
The takeaway
Demand forecasting for small business isn’t a leap into advanced analytics. It’s a habit: use the sales history you already have, weight recent months, adjust for seasonality and known events, convert the result into a buy quantity, then check it against reality and improve. Do that consistently and you stop swinging between overstock and stockout, and start holding roughly the right stock for roughly the right reasons.
That habit is far easier to keep when your sales, stock and purchasing already talk to each other — which is exactly the kind of connected foundation a business operating system like IXL CORE is built to give you.
