Operations
Know what to reorder, how much, and exactly when.
Smart Inventory Forecasting is inventory forecasting software for teams that want the reasoning shown, not hidden. Plans start at $49 a month with a 14-day free trial.
Most small distributors and retailers reorder on gut feel, a spreadsheet of last month's sales, or a min/max level somebody set three years ago and nobody has revisited. That works until a supplier slips a week, a line starts trending, or a seasonal peak arrives earlier than last year. Smart Inventory Forecasting replaces the guess with a per-SKU model, and prints the arithmetic next to every number so your buyer can argue with it.
Upload sales history as CSV and the engine fits a demand model per SKU, not one blended model across the catalogue. Fast movers and long-tail items behave differently, and averaging them together is how a forecast ends up confidently wrong about both.
For each item it produces a daily forecast across your chosen horizon, a safety stock figure, a reorder point, an economic order quantity, the projected stockout date, and days of cover at current stock. Every one of those comes with the inputs that produced it.
The usual spreadsheet approach to seasonality is to group historical sales by day of week and compare each day to the average. It looks reasonable and it is quietly broken: if demand is trending upward, later days in the data are higher simply because they are later. Group by weekday and the trend smears across the buckets, and a steadily growing product looks like it has a Thursday spike.
This engine removes the trend first. Seasonal indices come from the ratio of each observation to a centred moving average, which is the classical decomposition method, and the result is only applied when the pattern holds consistently across both halves of the history. A one-off promotion in March does not become a permanent March uplift.
Most safety stock formulas only consider variability in demand. That is half the problem. If your supplier quotes fourteen days and delivers anywhere between ten and twenty-four, the lead time itself is a source of risk, and a buffer sized only on demand variance will not cover it.
The formula used here is the standard one that includes both terms:
SS = z x sqrt( L x σd² + d² x σL² )
where z is the service-level factor, L is the average lead time in days, σd is the standard deviation of daily demand, d is average daily demand, and σL is the standard deviation of lead time. Set a 95% service level and z is 1.645; set 99% and it is 2.326. The page shows the figure it used and what each input was, so if the buffer looks wrong you can see which input is driving it.
There is no single forecasting method that wins on every product, so the engine fits two and lets the data pick. The important part is how it picks: on data the model has not seen.
This matters more than it sounds. Scoring a forecast on the same data you fitted it to rewards models that memorise noise, and the usual symptom is a smoothing parameter that drifts to its maximum and produces a trend pointing the wrong way. A holdout is the cheapest protection against shipping a model that looks excellent in the backtest and loses money in the warehouse.
A few hundred SKUs, one stockroom, and no planner. The Starter plan covers 300 SKUs and emails a reorder list; the useful output is the ranked stockout radar, not the forecast chart.
Stock split across locations with different demand profiles, and a supplier lead time that moves. Growth adds per-location forecasting, seasonality modelling and supplier lead-time tracking.
Thousands of SKUs where the real job is deciding what to ignore. ABC classification by revenue plus supplier-grouped purchase plans turn a 25,000-line catalogue into a short list of decisions.
If your current process is a pivot table and a reorder column, the migration is a CSV export. You keep the spreadsheet until the two agree, which is usually the fastest way to build trust in a forecast.
Upload sales history and the engine fits trend and seasonality per SKU, then computes safety stock, reorder points, economic order quantity and projected stockout dates — with the working shown so your ops lead can sanity-check every number.
Smart Inventory Forecasting starts at $49 a month with a 14-day free trial. Netstock is from $900 per month per company.
Seven days is the minimum, but the model is noticeably better from about sixty. Below seven days it refuses to forecast rather than guessing, and tells you why.
It accounts for variability in both demand and supplier lead time: z-score for your chosen service level, times the square root of lead time by demand variance plus demand squared by lead-time variance. The formula and every input are printed next to the result.
No. Seasonality is detected from ratios to a centred moving average, which removes the trend first. Grouping raw sales by weekday — what spreadsheets usually do — makes a steady trend look seasonal, and that is the mistake this avoids.
Also searched for: demand forecasting for small business · reorder point calculator software · safety stock software · netstock alternative · inventory planning software · stockout prevention software · eoq calculation software