Close this search box.
Close this search box.

Demand Forecasting for Fashion Apparel and Footwear: 5 Ways to Improve Accuracy

graph lines

Demand forecasting is one of the bigger challenges for retailers, wholesalers and manufacturers who sell fashion apparel and footwear. Overly optimistic forecasts cause overstocks. And overstocks force markdowns, which in turn reduce revenue and profit.

But pessimistic forecasts carry an opportunity cost for merchandise you could have sold at good margins but didn’t. Presales forecast error for new items typically range from 50 to 100 percent. This is according to Marshall Fisher and Ananth Raman, professors at the Harvard Business School, who reports on ways to mitigate forecast error in their book The New Science of Retailing: How Analytics are Transforming the Supply Chain and Improving Performance. Although you can improve forecast accuracy, your forecasts will never be perfect. Considering the inevitability of forecast error, the authors say, hedging strategies can help you make better buys. With better forecasts and better buying, you’ll accumulate less overstock, take fewer unwanted markdowns, and achieve higher revenue with better gross margins.

Why fashion is hard to forecast

Fashion items are hard to forecast for six main reasons:

• They’re new. New items, by definition, have no sales history. You have to base your forecast on the sales history of similar items. But no two fashion items are ever exactly alike, so you have to interpolate and approximate.

• Tastes are fickle. A color that sold well last year may bomb this year. You have to judge trends. This involves guesswork.

• Fashion merchandise has short life cycles. You can rarely accumulate enough sales history to generate a statistically accurate forecast before the item’s season has ended.  

• Because life cycles are short, you have little opportunity to correct for error. If lead times are longer than the life of an item, you have no opportunity to re-order from your supplier. And if the season of an item is shorter than a few weeks, you may not be able to re-allocate.

• Forecasting is hard to automate. People must make judgments that are prone to interpretation and error. When a new item comes out, someone must decide whether it will sell more like a prior item that is similar in some respects but different in others. Someone must also judge how much fashion trends have changed.  

• In the absence of good data, forecasts are set by whoever is most vocal, persuasive or authoritative.

How to improve forecast accuracy for fashion items

Fortunately, several practical measures can help offset these challenges:

1. Collect more detailed sales history. Ideally, collect sales history by item by week. Also collect more sales history by item and by location. Data storage is cheap compared to the value of having more history to guide your forecasts. If you can’t collect detailed sales history, collect as much as you can.

2. Tag the attributes or characteristics of your fashion items. Identify any characteristics of an item that may affect its appeal to your customers: color, fabric, cut, sleeve style, cuff style, collar style, neckline, pleating, pockets, retail price, placement and size of logos, and so on.

3. Make sales history more easily accessible to the people who need it. Put the data online. Forecasters don’t have time to consult paper reports. Show sales history graphically so patterns are visible at a glance. Make it easy to see sales history for items tagged with similar attributes or characteristics.

4. Seek multiple forecasts for the same item. For important items, don’t rely solely on the judgment of a single forecaster. Take the average forecast of several.

5. Track the accuracy of each of your forecasters over time. You’ll see that some people are more consistently accurate than others. As you develop a sense of whose forecasts are better, weight their forecasts more heavily.

6. Adjust your forecasts as soon as you have actual data from the first week’s sales.

Fisher and Raman’s book shares specific ideas for ways to implement suggestions four, five and six, above. Hedging strategies mitigate forecast error but even if you successfully implement all six, your forecast error is still likely to be higher for fashion items than for basics. With that likelihood in mind, the authors suggest that you tailor your buying strategy to the amount of forecast error you expect.

They present data to compare the profitability of each of four different buying strategies. Their analysis considers lost margin on sales as well as opportunity costs. It shows that the cost of under buying may often be higher than the cost of overbuying. So sometimes it pays to buy more than the forecast, sometimes less.

Fisher and Raman recommend a hedging strategy that’s too detailed to present here. But it’s well worth the price of the book to understand it. If twenty bucks for the book and an hour’s reading can help you reduce unnecessary markdowns by just 10 percent, you’re unlikely to find a better return on your investment of time and money.

[jesta_exit_intent text=”Looking for future-focused technology that drives business results?” strong_text=”Talk to an expert!” image=”” yes_link=”/contact-us?form_source=demand-forecasting-for-fashion-apparel-and-footwear-5-ways-to-improve-accuracy-2″]


Most Recent


Want to Learn More?

Skip to content