Both stockouts and overstocking are expensive — one loses sales and damages customer trust, the other ties up cash and risks obsolescence. Good forecasting is how you minimise both at once.
Why forecasting matters more as you grow
When you're selling a handful of orders a week, gut feel and manual reordering work fine. As volume grows, that approach breaks down: a stockout on a bestseller during a promotional period can cost more lost revenue in a single week than months of careful forecasting effort would have cost, while overordering on a product that turns out to have limited repeat demand ties up cash that could have funded a better-performing line.
Good forecasting isn't about predicting the future perfectly — it's about systematically reducing the size of your errors and building in the right buffers where uncertainty is highest.
Core forecasting methods
Moving average forecasting (using average sales over the past 4–12 weeks to predict near-term demand) is simple and works reasonably well for stable, non-seasonal products with a decent sales history. Exponential smoothing weights recent periods more heavily than older ones, responding faster to genuine trend changes than a simple average while still smoothing out day-to-day noise. For products with strong seasonality (garden furniture, Christmas decorations, swimwear), seasonal decomposition — separating the underlying trend from the seasonal pattern using at least one full prior year of data — is essential, since a simple moving average will badly underforecast in-season and overforecast out-of-season.
Most inventory management software automates these calculations, but understanding the underlying logic helps you sanity-check the system's suggestions rather than trusting them blindly, particularly for new products with limited history where any automated model is working with thin data.
Accounting for lead time and safety stock
Forecasting demand is only half the equation — you also need to forecast against your supplier's lead time (the gap between placing a reorder and receiving stock), which for UK-based suppliers might be 1–2 weeks but for overseas manufacturing can run 30–90 days including production and shipping. Your reorder point should trigger early enough to cover expected sales during the full lead time, plus a safety stock buffer sized to your lead time variability and how costly a stockout would be for that specific product.
A simple reorder point formula: (average daily sales × lead time in days) + safety stock. Safety stock itself should scale with both demand variability (more volatile products need bigger buffers) and lead time variability (an unreliable overseas supplier needs a bigger buffer than a reliable UK one, even at the same average lead time).
New product forecasting without history
For genuinely new products, use analogous forecasting — basing initial estimates on the launch performance of similar products you've sold before, adjusted for known differences in price, marketing push or market conditions. Start with a conservative first order sized to validate demand rather than to hit your best-case sales projection, and build in a fast-reorder capability (a UK or nearby supplier, or air freight for the first few reorders even if it's costlier per unit) so you're not stuck out of stock for months if the product sells faster than the initial conservative estimate.
Common forecasting mistakes
- Ignoring marketing and promotional calendar impact — a planned sale or influencer campaign will spike demand well above the organic baseline a pure historical model predicts.
- Treating all SKUs the same — an ABC analysis (classifying products by revenue/margin contribution) lets you apply tighter forecasting rigour and safety stock to your top 20% of SKUs that typically drive 70-80% of revenue, without over-investing analytical effort in your long tail.
- Failing to adjust for discontinued or soon-to-be-discontinued products, where historical sales data actively misleads future forecasts.
- Not revisiting forecasts regularly — a forecast made at the start of a quarter should be checked against actuals at least monthly and adjusted, not treated as fixed.
Building a forecasting habit
Set a recurring weekly or biweekly review of your top SKUs' actual sales versus forecast, current stock cover (weeks of inventory remaining at current sales rate), and any upcoming lead-time-relevant reorder decisions. This regular cadence, more than any specific statistical technique, is what actually prevents the expensive extremes of stockouts and overstocking in a growing UK e-commerce business.
Where This Fits Into Your Wider Strategy
As with most things covered in Supply Chain & Logistics, inventory forecasting techniques to prevent stockouts and overstocking works best when it's revisited periodically rather than decided once and forgotten. For the practical mechanics, see Inventory KPIs Every UK Online Retailer Should Be Tracking.
Visibility is the other recurring theme. Whether you're managing your own stock or working with a third party, the retailers who scale smoothly are usually the ones who can see accurate, near-real-time data on where every unit sits — rather than reconciling spreadsheets after the fact. Investing in that visibility earlier than feels necessary tends to pay for itself the first time a supplier or carrier issue would otherwise have gone unnoticed for days.
For related reading: It's worth reading this alongside Dead Stock and Markdown Strategy: Recovering Value from Slow-Moving Inventory before making a final decision. If you haven't already, Managing Seasonal and Perishable Inventory Without Getting Burned is a useful companion read on this exact point.