Recently we had a holiday on the NSW Central Coast and needing a few bits and pieces
I ventured in to the local department store.
Being a bloke, I seldom shop and then only when I really need something. So OK, that
means I haven’t been to my share of shops but this place was pretty amazing! It was
absolutely full of stuff! You could barely get down the aisles but, along with the
curios, there were some really handy things. Finding them was hit and miss and it
certainly was for me a case of “better to arrive than journey hopefully”.
Recent styles were “cheek by jowl” with Dickensian artefacts… clothes, crockery, string,
stationary most anything you might need. A bit of a “one stop shop” and the price
was right! It reminded me a bit of when I used to go shopping with my Mum 35 years
ago at the discount food store with everything everywhere. Talk about retail aversion
therapy!
Being an engineer by background and more analytical than is good for me, I left the
shop wondering if the store could survive as the overriding impression was that it
had far too much stock. Simply and nostalgia aside, there seemed to be far too much
money locked up in “them there” shelves. Surely there was a better way!
Retailers are faced with the difficult problem of trying to match the stock they hold
against customer demand in an environment of continual change, often driven by seasonal
demand and fashion. The trick is to carry just enough stock so that each customer
can find what they want when they want it so you don’t lose a selling opportunity
but not so much stock that it sits on the shelf until it is disposed of in next years
sale.
One of the big issues is the “forest for the trees” problem. The buying habits of
customers are diverse as they come in all shapes and sizes with different style and
colour preferences. This means that there are an enormous number of combinations all
being continually influenced by season and fashion. Retailers servicing multiple stores
have this problem only magnified. With this amount of data it is easy to see how inventory
managers have great difficulty seeing the “forest for the trees”.
Faced with an economic and competitive landscape demanding tighter margins for survival
it is imperative that only sufficient stock is held to satisfy customer demand. Retailers
can no longer afford to make their decisions at the class or category level. Just
because there is a run on jeans in one locale it doesn’t mean that the size 12, female
stonewashed is moving in all stores or at all.
In the world of retail, of stores and SKUs, there is a well known maxim “Retail is
Detail” and we all know that the “devil is in the detail”.
Size 12, female stonewashed jeans may not have been sold last week but if buyers are
planning at the total jeans level this will not be visible leading to wrong buying
decisions and sub optimal stocking levels.
Thankfully the ever increasing power of computers and the application of NASA inspired
techniques with fast and easy methods for seeing the data are making the buyer’s job
easier and more accurate.
The power of the latest computers is making it possible to plan at the SKU store intersection
and spot trends at the lowest level but this is still overwhelming from a inventory
buying perspective.
To overcome this automated buying techniques are deployed based on optimal stock models.
Algorithms that review the sales trends determine what the optimal stocking levels
are for each SKU in each store and an inventory order is raised on this basis.
All automation algorithms are not equal and some are clearly better than others in
predicting future sales. At its simplest level the buying decision may be to replace
the inventory from sales from the previous week. The difficulty with this approach
is that it does not take into account the amount of stock on the shelf in the store,
the changes in seasons, population demographics and fashions or unusual purchases.
An unusual purchase may be a local mum buying a dozen pink shorts for her daughter’s
Netball team when only 2 usually sell each week. Ideally the store would only hold
2 or 3 items (enough to allow for restocking lead times).
But how is it possible to forecast the correct stock levels when seasonal, fashion
and unusual events are occurring?
Over the years different techniques have been used such as Moving Average and Replacement
but none of these do a good job in sorting out the “noise” from the “one off” exceptional
sales like the pink shorts for the team or the cyclic run on pencils and pads prior
to the return to school.
This is where a NASA mathematician comes to an unexpected rescue. Rudolf Kalman observed
that he could apply his linear filtering technique that strips unwanted noise out
of a streams of data to the problem of trajectory estimation leading to its incorporation
in the Apollo navigation computer.
The Kalman filter as it is now known is widely used in navigational and guidance systems,
radar tracking and satellite orbit determination as well as in econometrics. It has
now been shown to be very effective at eliminating retail “noise” even a run on pink
shorts or stonewashed jeans enabling retailers to create better forecasts, hone their
stock models and radically drive down inventory levels.
Through the correct application of these filters reductions in inventory of between
10% & 20% and improvements in stock turns of 10% are not uncommon. Shelf space
and dollars freed from over stocking can be refocussed towards more profitable and
faster moving items.
This leads to potential savings from stock reductions and increases in revenue from
related improvements in stock turn amounting to millions of dollars in even medium
sized retailers.
So if next time you visit a store and everything is “just so” and you begin to yearn
for the old cramped, nostalgic quaint experience that has “Gone with Gowings” remember
NASA chose the stonewashed.
Business Intelligence and Performance
Management Home
