page contents

I was recently informed that you cannot forecast fashion.  If that were a true statement, I would not be writing this blog, or selling and supporting a superior Demand Planning software solution.  Lighthouse.

This perception was created because each selling season requires merchandise that is fresh and new.  We have no way of knowing how the new seasons fashion will sell—or do we?  How can we possibly plan with so many unknowns?  That said, the new fashion line might be a success, or could bomb.  We know consumers are fickle, and there is no guarantee what merchandisers introduce to the market will be appealing and sell in huge quantities.  We understand too that manufacturers show new lines at trade shows where they take retail customer orders.  The buyer’s reaction serves as an indicator of how popular the line might be.

There are two distinct aspects of fashion retail we need to consider.  The first is manufacturers supplying retail customers, the second are retailers developing their own lines and sourcing production.  This, for US retailers is generally from overseas, very often from China, Thailand, India, Bangladesh, Jordan, Indonesia, or other low wage, low cost countries, and occasionally within the USA where responsiveness and quality is paramount.  We discuss the manufacturer option initially.

That still begs the question, can demand be forecasted?  Yes.  Well, how?

Most fashion creators have lines that are a variation on a theme of what they produced last season.  No company makes shirts or blouses for the upcoming season when they have never made this in the past and where they enjoy a reputation of only producing the highest quality rain or over coats.  The manufacturer will have established a reputation for producing merchandise that appeals to mass market, or luxury lines, but never the same label to address both consumer demographics.  As an aside, if the organization is adding a brand new product line, never having manufactured it before, then yes, this too can be forecast and tracked, but that is a discussion we touch on briefly near the end of this writing.  For this discussion let’s stay with an ongoing fashion line of seasonal shirts.

What type of shirt to we make?  Dress shirt, T-shirt, golf shirt, Rugby shirt, sweatshirt, pajama shirt, long sleeve, short sleeve—and the list goes on.  To keep the discussion simple, I’ll focus on dress shirts.  In the past decades, our manufacturer made and sold thousands of this type of shirt.  Society experienced a rise and fall of this category of clothing.  Suits requiring smart button up dress shirts and ties are standard wear for professionals, especially attorneys working in large legal practices, or corporate executives meeting with customers where image is all important.  The trend, however, has been to more casual wear where open neck dress shirts and slacks are the order of the day.  Then too we have a new dress statement where the shirt is not worn tucked into the pant, but displayed on the outside.

So as not to lose you in this discussion, let me assign last years’ fashion line shirts to a product family code of ABC, and the new season line to XYZ.  At the outset we do not have a clue what our rate of sales of XYZ will be.  Our marketing, sales, and merchandizing team can provide their best guess based on experience, but that does not help if we are way out of line with estimates.  Produce too few and we give business to our competitors, too many, and we have to rid the excess inventory through a heavily discounted fire sale.

When examining historical orders and sales for the ABC family, we determine style, color, and size information aggregated from the lowest level of SKU (Stock Keeping Unit) and recorded by retailer store level sales and date.  From analysis we can quickly determine size popularity, and this may vary by geographic region.  We may experience a shift in the size trend over the past several years as we better understand population weight gain, or success in exercise and diet programs.  This could vary according to the region and weather conditions where the consumer resides, and different for inner city and suburb depending on store location.  Demographic information is important.  I’ll explain later why I wrote about “past several years” of history while understanding the ABC was only last season’s fashion statement.  Style and color choices make up the balance of required analysis information for the ABC line.

We examine and understand both orders received, and shipments, where timing is different.  Orders received for fashion goods may not require shipments until the start of fashion season.  Tracking orders against expectations, or order forecast, provides early indicators as to the potential success of the campaign.  Once goods are shipped it is critical to track the rate of sales meeting consumer demand.  With a successful campaign there may be a need to increase orders to meet demand on the production facility if material, labor, cut, make, and trim capacity can be scheduled in a timely manner.  This early demand signal is more important if production is in Asia for the US market with longer lead times.  Merchandise arriving after the season end is of little value.  The converse applies.  If merchandise is not selling, then cutting back on planned production is vitally important with communication flowing throughout the supply chain to prevent unnecessary cloth from being woven and printed, and with additional value added to finishing and packaging processes adding to the potential waste.

So how do we go about the forecasting process?  Recognize that with the new line, XYZ, it is a variation on a theme of the ABC line.  If we look at ABC in aggregate, we quickly learn the rate of orders and sales in units by day or week over the season.  This should likely be the starting point to determine what merchandizers understand should sell, based on history.  Since intelligent input can be used to adjust these numbers, there may be valid reasons to increase or decrease this aggregated number over the duration of the season.  When planning in aggregate it is possible that the ABC line consisted of 7 colors, whereas XYZ only has 5 color choices.

In forecasting it is important to have a minimum of three-year history to accurately determine trend and seasonality.  With a single year of history it is impossible to tell seasonality.  With two years history you (or the forecasting algorithms) cannot be sure if we are witnessing a coincidence in timing of seasonality.  With a three year historical picture we will have a lock on seasonality.  (With Easter being celebrated on different dates each year requires special treatment).  So how do we build this 3-year history?  Your forecasting application must allow for a supersession function.  Here we string together data from season to season of a similar line of merchandise.  Last year we had the line ABC, the year before DEF, and three years ago GHI.  As stated earlier, some of the data will transfer such as timing, quantity, and size.  Style and color may not be appropriate for this level of planning, nor required.  The historical timing and quantity will be the main drivers into the forecast to predict demand for the new season XYZ line.  Recall we are talking about two different demand streams: orders and shipments.  Each will be analyzed separately.  First plot the predicted orders to be received and track actual orders received to this forecast.  Establish alerts if the order rate is outside of a predetermined tolerance.  This information will tell management if the new line is more successful than we predicted, or may potentially bomb, in which case we adjust production or sourcing plans accordingly.

The second step is to plot sales.  This is not quite as simple as it sounds.  Where do we get this data?  It could come from our internal records, or it could be shared with our organization from our customers POS (point-of-sales) data by retailers with information to a specific door within their network.  In some situations we may be able to work with a draw of inventory from the retailer’s distribution center—or both DC and store.  If that is not challenging enough, then it may benefit us to recognize if within the historical data we had periods where promotional activity took place.  A BOGO (buy one, get one) may have resulted in an increased demand in an earlier period and cannibalized sales of other lines or sales taking place in a later time period.  To generate an accurate forecast requires that we normalize this historic data.  The forecasting application can automatically trim the peaks and fill troughs to provide an improved historical pattern from which to generate a forecast.  Absent this function, a manual intervention may be required.

Are all forecast always accurate?  No.  There could be extrinsic factors that come into play.  Milwaukee, Wisconsin in 2015 experienced the warmest Decembers in recorded history.  This impacts sales of gloves, ear muffs, coats, and other merchandise for this time of year, and normally ideal as Christmas gifts.  As the economy cycles through its ups and downs impacting consumer confidence, and their purchase decisions.  A systematically calculated forecast is still significantly more accurate than a thumb suck or a SWAG (smart wise guess).  That said, it is the reason why we plan and test sales daily to determine if the rate of sales remains valid.

This base line forecast helps lay out a plan of expected demand over time.  Most of what I have written about so far is looking at planning data at an aggregate level.  This information must now be prorated or dis-aggregated down to the detail level.  As appropriate we plan down to customer, customer DC, customer store, style, color, and size.  The size component is likely the easiest to plan as we track what has taken place by region over time.  The style and color options may require the creation of a planning BOM (Bill of Material).  Here we are looking at the likely popularity option of styles, and then colors within style.  Please understand that these plans are not cast in stone.  If you experience a shift in the mix popularity of one style or color over another, then modify the ratios.  To expand on the planning BOM concept, understand that we establish a relationship from the parent (line) to the children (styles, colors, or sizes) and a percentage relationship of child to single parent.  The aggregate forecast per period (day or week) is now applied to the lowest level of detail.

While at the SKU level, plan the replenishment based on on-hand inventory.  If you are provided daily POS data, this should include the sales and inventory balances.  Inventory holding may optionally be calculated based on initial shipment, depletion from sales, and adjusted for returns or shrinkage (a euphemism for theft).  Much of this depends if you operate a Vendor Managed Inventory (VMI) strategy where the supplier manages the retailer’s inventory at store level, and in return is provided with accurate daily data.  Looking ahead, plan on replenishment depending on the life cycle of this fashion item and the remaining time of the campaign.  If this is a basic goods item (a plain white shirt), then planning is significantly simpler.

Your early warning signal allows management to take corrective action.  With a lower rate of sales, should you offer your customer an incentive program to help move merchandise by agreeing to a discounted sale program?  Then too how quickly can you alert your suppliers that you have a winner and need to increase the shipments over a very short lead time to capitalize on this opportunity that will benefit every company in the supply chain.

Returning to the concept of a planning BOM for a moment.  If you have a totally new line that you are introducing, and you believe it may have a sales cycle similar to another line, then you can use the selling pattern of the like product and apply to the new.  The strategy is to plot the sales pattern of the current line and reduce it to percentages over time.  That provides you visually with a sales curve.  Next apply the best intelligence you have and determine a total demand for the duration of the sales campaign.  Apply that total estimate to the curve, and let each time period calculate a unit demand.  However we are still at the planning stage.  As the new product is rolled out, and you record actual sales, this total demand figure is recalculated based on actual to-date sales.  Demand will either increase, decrease, or stay much the same.  The replenishment calculations will bring orders into an earlier period if actuals exceed demand, or defer replenishment orders if actuals do not meet the demand criteria.

An important strategy is to contain costs.  Shipping can eat away at profits if you find a need to rebalance merchandise by moving inventory from one location to another.  Following a demand pull strategy works ideally.  Start with the consumer at store level, pull replenishment product from the supplying DC, and replenish from the supplier or the retailers own central distribution hub.  Then too if the retailer has embraced omnichannel strategies, which location to ship internet orders from requires planning.  Do you wish to rob your store of on-shelf merchandise for a rapid delivery, or will your customer be willing to wait an additional day or two if you ship from an out of state DC?  In our forecast planning it is important that we identify all the channels of distribution and carefully plan to achieve very high levels of on time delivery to customers with the lowest investment in inventory, and the highest inventory turns possible.  This can only be achieved through a highly responsive planning and replenishment tool, almost always a niche solution that adds value to your enterprise system.

Remember, for any retailer to be out of stock of needed merchandise is the very worst form of customer service.

Keeping the discussion simple, I deliberately left out important information such as product cost, average selling price and margin.  These factors are critical to help executive management to know and understand how successful the new fashion line will be, and its contribution to both the overall sales revenue and profit performance.

There is not much I can add if we address retailers who source their own products and manage their own supply chain.  Most of what I stated above applies equally to this sector.  Often the assumption is that for best pricing an order must be issued for the entire season delivered up front and then to see what transpires.  A preferred strategy is to agree to a total order quantity and ship regularly over time against required inventory releases.  Tight collaboration with suppliers, where data is constantly shared, works well.  Companies generally work within the constraint of typical 6 week replenishment lead times from Asia.  We need to plan around the Chinese New Year where the country shuts down and shipments may not be realized.  What planning methods can we employ to maintain or reduce this lead time criteria?  Or what can we do to plan around the reality of the situation?  If we plan special merchandise for key holidays like Easter, Independence Day, or Christmas, there is little the retailer can do but to look at last year’s figures, adjust with a macro look at current economic conditions, and hope for the best in determining what is the greatest business risk—having too much, or too little inventory.  We live in a Goldilocks world.  Lighthouse is a “just right” solution.

Comments are closed.