Retail inventory forecasting software is for the buying decisions a reorder point cannot make: how much of a seasonal range to commit to before the season starts, how much of a new product to order when there is no history, how much extra to hold for a promotion, and how to phase deliveries across weeks so the stock arrives when it sells. With Pentoggle, a retailer can describe how they plan and buy and generate the starting application around it.
This page is about planning ahead. Reordering items that sell continuously, where the decision is a threshold rather than a plan, is Retail Replenishment Software. Placing the resulting orders is Retail Purchase Management Software. What to do with the stock a forecast got wrong is Retail Dead Stock Management Software.
Most retailers already run an accounting system such as QuickBooks, Tally or Xero, and a POS that records what sold. Those stay where they are. What is often still managed outside them is the plan: what the store expects to sell over the coming season, what it has committed to buy against that, and how the two compared last time.
Key takeaways
- A forecast is an assumption written down. Its main value is not accuracy but that it can be compared to what happened, which is how buying improves.
- Sales history has to be cleaned before it can be used. A week with a stockout understates demand; a week with a promotion overstates it; a week with a closure says nothing at all.
- Seasonality is usually the largest single pattern in retail sales, and it is measurable from a store's own history rather than assumed from the category.
- New products have no history and need a comparable, an assumption and a small first commitment, not a confident number.
- A useful number is forecast error by category: how far the plan was from actual sales, and in which direction.
The spreadsheet is often not the problem
A buy plan in a spreadsheet, built from last year's numbers and the buyer's judgement, is a working method and a great deal of good retail buying has been done that way.
The trouble starts at identifiable points.
When last year's numbers include last year's problems
The range sold 400 units last season. It was also out of stock for three weeks and discounted for two. The 400 is not the demand and using it as the plan repeats the same shortfall.
When the plan and the orders drift apart
The plan said 1,200 units across the season. Orders were placed over eight weeks by two people. Nobody has added them up against the plan, and the total is 1,600.
When the phasing is ignored
The right total arrives in the wrong weeks. Half the season's stock lands after the peak, and the same units that would have sold at full price in week three are marked down in week nine.
When nobody compares afterwards
The season ends. The leftover stock is marked down and the buying starts again for the next one, with no record of which parts of the plan were right and which were not.
What retail inventory forecasting software holds
Sales history, cleaned
Sales by item, category and location over time, with periods marked as stockout, promotion, closure or otherwise not normal, so they can be excluded or adjusted.
Seasonality
The shape of demand across the year, by week or month, per category or item, calculated from the store's own history.
Forecasts
Expected sales for a period, item or range, with the method and the assumptions recorded alongside the number.
Buy plans
The quantity to commit, phased across delivery dates, with the supplier, the cost and the resulting stock position week by week.
Plan against orders
What has actually been ordered against the plan, so the commitment is visible before it exceeds the intention.
New product forecasts
A comparable item or range, the assumption made, the initial commitment and the review point at which the assumption is tested.
Promotion forecasts
Expected uplift for a promotion, the stock built for it, and the actual result held for the next time.
Forecast against actual
The comparison after the fact, by item and category, with the error and its direction retained.
Clean the history before using it
Sales history is a record of what was sold, which is not the same as what customers wanted to buy. The difference matters most in exactly the periods a forecast pays attention to.
An item that was out of stock for three weeks sold nothing in those weeks. Its history shows zero. Its demand was not zero, and a forecast built on the raw history will under-buy and produce the same stockout again. Marking the stockout period allows the forecast to exclude it or to estimate what would have sold.
An item that was promoted sold at a rate it does not sustain. Its history shows a spike. A forecast that treats the spike as the normal run rate will over-buy for the period after. Marking the promotion allows the uplift to be separated from the baseline, which also makes the uplift itself a useful number for planning the next promotion.
A period when the store was closed, or a location was refitted, or a nearby road was dug up, says nothing about demand at all and should be excluded rather than averaged in.
The application's job here is to make marking these periods easy and to keep the marks with the history, so that every forecast built afterwards uses the cleaned version. Most of the value in forecasting software for a mid-sized retailer comes from this step rather than from any sophistication in the method that follows.
Seasonality is measurable from your own history
Many categories sell in a shape rather than at a rate. Warm clothing, garden goods, school supplies, festive food, gifts, sports equipment and many others have a strong annual pattern, and the pattern for a particular store is often more specific than the category's general shape.
That pattern can be measured. Two or three years of cleaned weekly sales for a category, expressed as a share of the year's total by week, gives the store's own seasonal curve. Applied to a total expectation for the coming year, it produces a week-by-week expectation, which is what a buy plan actually needs.
The curve is also what makes phasing possible. Knowing that a category does a large share of its year in six weeks tells the buyer when stock has to be on the shelf, which sets the delivery dates on the order, which sets the order date given the supplier's lead time. A plan without phasing is a total, and a total can be right while the season still goes wrong.
Two cautions are worth stating. A store's own curve reflects when it had stock, so a category that was short during its peak will show a flatter curve than its real demand. And a curve built from a single year is a description of that year, which may have had a hot spring or a late festival. More years is better, and where the store does not have them, the buyer's knowledge of the calendar is a legitimate input.
New products need a comparable, not a number
The hardest forecast is for an item with no history, and it is also the one most often done with false confidence.
The workable approach is to be explicit about the reasoning rather than the number. Choose a comparable: an existing item or range that this new one most resembles in price, category and customer. State the assumption: this will sell at roughly the rate of that comparable, or some fraction or multiple of it, for a stated reason. Commit small: order enough to test rather than enough to cover the assumption being right. And set a review point: a date or a sales threshold at which the assumption is checked against actual sales and the second order is sized from real data rather than from the guess.
Holding the comparable and the assumption in the system, rather than in the buyer's head, is what makes the review possible. When the review comes, the question is not "how is it selling" but "is it selling at the rate we assumed, and if not, by how much." That comparison, repeated across many new products, is how a buyer learns which comparables work.
The same structure applies to promotions. An expected uplift, a stock build against it, and an actual result recorded afterwards. A store that has run twenty promotions with the results kept has a far better basis for planning the twenty-first than a store that has run the same twenty and kept nothing.
Where forecasting looks different by business type
- Fashion Retail Software, where the buy is committed a season ahead, cannot be repeated, and the whole margin depends on how much sells before markdown.
- Apparel Retail Software, where the forecast has to be a size ratio as well as a total, and the wrong ratio leaves stock nobody wants.
- Sporting Goods Retail Software, where categories peak in different parts of the year and the store is really several seasonal businesses.
- Home Furnishing Retail Software, where ranges are introduced and cleared on a cycle and the clearance date is planned at the time of buying.
- Grocery Store Software, where the forecast is short-horizon and driven by day of week, weather and local events more than by season.
- D2C Brand Software, where production lead times are long and the forecast decides a manufacturing commitment rather than a purchase order.
Why retailers choose Pentoggle for forecasting
History you can trust
Stockouts, promotions and closures marked, so the forecast is built on demand rather than on what happened to be on the shelf.
Your own seasonal curve
The shape of your year measured from your sales, by category and location, rather than assumed.
Plans that hold their assumptions
The comparable, the uplift and the reasoning kept with the number, so the review after the season is a real comparison.
Plan against committed orders
What has been ordered against what was planned, visible before the commitment overshoots.
Sits around your accounting and your purchasing
QuickBooks, Tally, Xero and comparable systems continue handling invoices and payments. Pentoggle adds the plan and the comparison, and passes quantities to purchasing.
A useful number for forecasting
Forecast error by category: how far the plan was from actual sales, and in which direction.
The direction matters as much as the size. Consistently forecasting high produces markdowns and dead stock; consistently forecasting low produces stockouts and lost sales. They are different failures with different costs, and an average error that ignores direction hides which one the store has.
Read it by category and by buyer rather than in total. Errors in opposite directions cancel out in a total and tell nobody anything.
Read it against what was done about it. A forecast that was high but caught early, with orders cut before delivery, cost far less than the same error discovered at the end of the season. That is a reason to compare plan against actual during the season, not only after it.
Ready to build retail inventory forecasting software?
You know what you sold last season.
You may not be able to say how much of that was demand, how much was the discount, and how much you never sold because the shelf was empty.
Describe your categories, your seasons and how you plan a buy to Pentoggle in plain English and generate a working first version in hours, then refine it around your process.
Related resources
- Retail Replenishment Software →
- Retail Purchase Management Software →
- Retail Dead Stock Management Software →
- Retail Sales Analytics Software
- Retail Promotion and Discount Management Software
- Fashion Retail Software
- AI Software for Retail Businesses →