Tagged!! A new approach to demand forecasting and assortment planning in Fashion Apparel
The fashion apparel industry has a unique problem - designing and forecasting the demand for their latest collections and lines. The designers naturally base their designs on their inspirations and new innovative concepts. Which means that the seasonal lines and their associated characteristics could vary dramatically each season. While creativity drives designers, at the same time it is important to create lines, styles and collections that connect with the consumers and sell fast. One of the problem is that many of these collections /lines are increasingly required to connect with an increasingly broad audience with growing importance of the international markets, and expansion of target segments. How do designers stay in touch with the changing consumer?
Merchandisers and planners also need to figure out exactly how much they might end up selling of each line. The typical demand forecasting tools used by many consumer goods companies, including those apparel retailers with relatively standard products, don’t work so well for truly creative design houses. For instance, last season’s yacht themed summer collection might have sold out fast but the demand pattern might be very different for this year’s Hamptons-themed summer collection. How do merchandisers forecast demand accurately?
I think part of the solution is by using Web 2.0 concepts like “tagging” and then using predictive modeling using tags.
So, how exactly do you tag the past? For past collections, designers, planners, marketers, merchandisers and consumers are asked an open ended question, “What did you think of that line/ collection? What comes to your mind?” – kind of a word association test but with actual garments. The audience’s description is captured as tags, and a tag frequency map created. For instance, the yacht themed collection might end up with tags such as cool, relaxing, adventurous, sporty, white, ocean, tropics, classy, yacht, etc. This is repeated for each line from the recent past. At the end of a short exercise, there would be a tag map for each line. Then the sales data for each past line is mapped to these line tag map. Over multiple seasons, a “tag pattern” would emerge that would show the tags that are growing sales and those that are not.
And how do you use the tag maps? When designers are in the process of designing new collections and lines, a similar tagging/ association exercise is done based on their paper designs or physical samples. Those lines that are out of sync with the hot tag trends could be discarded or reworked. Subsequently when the designs are baselined, predictive demand forecasting using advanced statistical techniques can be done to help the merchandisers and planners come out with overall demand. This can even be taken to the next level of granularity to predict demand at store level based on past tag maps that have sold well at the store.
Of course, this approach is not a silver bullet and some truly path-breaking themes need to be excluded. For instance, concepts that aim to change the customer thinking and start a completely new fashion trend.
Do you think this approach would work? Is there applicability in segments beyond seasonal apparel?
Comments
Very interesting insight.However in practical scenario how easy would it be to capture such data ? True data on tagging can some only from the store ;sales staff need to be trained and equipped with suitable mechanism to make data available for creating a tag cloud.
If there are some pilot results, please share them .
Thanks.
Posted by: Balaji OS | October 7, 2009 12:53 PM