Festival Demand prediction

The crackers, lamps and Diwali decoration sales start from zero, increases slowly and converts into almost a mad rush on the eve of Diwali. Bulk of the crackers is bought in the last 2-3 days of Diwali and none on the next day. How do you do the demand prediction of a product that is sold for only 3 days in a year? Not to mention that all unsold inventory has to be disposed of at a fraction of the price as there is no longer any market till the next year. Also the lead time for any item would be weeks/months, hence if you are stock out of did not guess the fashion correctly, you have no recourse. Most festival collection is sold to retailers on cash and carry (rather than credit) and there is no mitigation for unsold inventory.
Furthermore it rained on the Diwali night in Bangalore which means people were less eager to light crackers and the poor drainage in the streets means fewer footfalls of customers. A conventional shop would lower its prices so that it does not end up with unsold inventory. However seasoned sellers know that the buyers are here out of compulsion/family pressure. They will not have enough time to browse through the shop and would buy only the essentials. Hence it is best to raise the prices (earn the same amount per customer and compensate for the reduces sales)
Every merchant dreams of a hot-selling item, but it is a nightmare for festival retailers. Last thing they want is a shop that sells 100s of different kinds of crackers but their sparkles/fooljhadi is sold-out. Smart retailers need to work with their customers trying to influence them to buy a larger variety and manage a more evenly sales distribution. Since the entire sales happen in a few days, the bundling, pricing, placement decisions have to be taken in real-time.
Sometimes we get daunted by the complexity of simple business but we fail to realize that the shopkeeper who aces it often has not even completed his high school.