Demand forecasting is estimating how much of each item you will sell in the coming weeks, so you buy and make the right amount. For Indian retailers, brands and distributors, the festive season is where it matters most: a large slice of the year's sales lands in a few weeks, suppliers' lead times stretch, and both stock-outs and leftover stock are expensive.
This guide covers the practical side — data, festival dates, lead times and safety stock — and where AI helps. To see a forecast with festival effects you can switch on and off, try the sales and demand forecasting demo.
Why "same month last year" goes wrong
The most common forecast is last year's sales for the same month, plus a bit. It fails in the festive season for one simple reason: festival dates move. Diwali, Navratri, Holi, Eid and Raksha Bandhan follow lunar calendars, so they shift by days or weeks from year to year. If Diwali was in early November last year and is in late October this year, "October last year" compares the wrong weeks.
The fix is to compare festival to festival: line up last year's sales by weeks before and after each festival, not by calendar month.
The data you need
- Sales by item and week for at least two years — three is better
- Festival and event dates for those years and the coming one
- Prices and promotions you ran, and when
- Stock-outs — weeks when you had no stock sell zero, which is not the same as zero demand
- Supplier lead times for each item, in festive season (often longer than usual)
Most of this already sits in your billing, POS or Tally. The work is cleaning it — merging item codes that changed, marking stock-out weeks, and tagging festival weeks.
From forecast to order
A forecast is only useful when it turns into a purchase or production decision. A simple, sound rule:
Order = expected demand during the lead time + safety stock − stock in hand
- Lead time: if the supplier takes two weeks, you need to cover the next two weeks of demand with what is ordered today.
- Safety stock: a buffer for forecast error — commonly a percentage of lead-time demand, set higher for items where running out is costly.
- Stock in hand: including stock in transit and at other locations you can transfer from.
Run this every week through the season, not once in August.
What-ifs worth asking
- What if the festival lift is smaller this year? Switch festival effects off and see the plain-trend forecast.
- What if we cut the price by 5%? Most items sell more when cheaper; a forecast that knows past price changes can estimate by how much.
- What if the supplier is a week late? Lengthen the lead time and see the order change.
Spreadsheet or AI forecast?
A spreadsheet with festival-aligned weeks and the order rule above is a big step up for a business with a few dozen key items. AI forecasting earns its place when:
- you have hundreds or thousands of items across several locations
- demand depends on many things at once — festivals, weddings, weather, price, promotions, school calendars
- you need forecasts refreshed every week without an analyst rebuilding them
- you want the forecast to learn from its own mistakes each season
A custom model learns from your own sales, prices and calendar, and is judged on how close it came to what actually sold.
Measure forecast accuracy
Track, for each item and week, how far the forecast was from actual sales. A simple measure is the average percentage error (MAPE) — but look at it by item group, because a 20% error on a slow-moving gift box matters less than on your top-selling staple. Also track the results that matter to the business: stock-outs in festival weeks, and stock left over a month after.
A festive-season checklist
- Tag last two to three years' festival weeks in your sales data.
- Mark stock-out weeks so they don't drag the forecast down.
- Confirm festive lead times with suppliers in July or August.
- Set safety stock by item — higher for items you can't afford to run out of.
- Re-forecast weekly from September, using the latest sales.
- After the season, compare forecast with actual and note what surprised you.
The short version
Festive forecasting goes wrong when it compares calendar months instead of festivals, ignores stock-outs, or stops at a forecast instead of an order. Align by festival, plan to lead time plus safety stock, and refresh weekly — with AI when the number of items or factors outgrows a spreadsheet.
Try the sales and demand forecasting demo, see the distributor management system, or talk to us about forecasting for your business.