When Predictive Analytics is Just Forecasting

The term Big Data and Predictive Analytics are quite the rage these days.  Colleges and universities are offering degrees in analytics.  Consultants are offering services and dashboard software to help their clients be on top of their Big Data and Analytic game.  Companies are creating Analytic departments led by either managers, directors, and even VPs.  There is a fair amount of buzz in this area.

There are great examples of where this discipline and capability will help optimize business performance.  Businesses can best use it in two general areas:

1. What is trending right now and what can be done to take advantage of it?

   1. Rental car agencies monitor the prices of rental cars in each market to see what competitors are doing and how demand is tracking.  They monitor this information real time and change their pricing hourly.

   2. Think of Zara the Spanish based clothing retailer. “Zara is renowned for its ability to deliver new clothes to stores quickly and in small batches. Twice a week, at precise times, store managers order clothes, and twice a week, on schedule, new garments arrive. To achieve this, Zara controls more of its manufacturing than do most retailers: About half its clothes are made in Spain or nearby countries. For Zara, its supply chain is its competitive advantage.”  Their supply chain is designed to deliver, very quickly, with what is treading and selling.  This requires on real time analytics to feed the right orders and quantities into their supply chain. Business Week
      

2. What are the buying patterns of customers and classes of customers?  How can this information be used to market other offerings  to individual customers and classes of customers?

   1. The classic Analytics story here is the Target example.  They noticed that when young female customers switched from buying fragranced cosmetics to non-fragrance variants, the same customers started buying maternity and baby products in the following few months. ~  Forbes

   2. Anyone that uses Amazon.com is used to getting suggestions based on past purchases every time they log on.

There is a lot of promise for Big Data and Analytics.  There is also one area in which there some, apologies to Alan Greenspan,  “irrational exuberance” of the dot.com era.  

Irrational exuberance?  

As important and potentially useful as Big Data and Analytics are, there is a one area where the hype is probably overselling the capability:  any application that requires significant forecasting.  There are promises of using company and external data to predict sales, material and commodity prices, and transportation costs to name a few.  It all sounds very good.  It sounds very promising.  But, it is still forecasting.  And, as anyone who has worked in the discipline knows all too well, forecasting is predicting the future. Predicting the future is a pretty hard thing to do.  

Consider a swimming pool supply company we know.  Their business is very seasonal.  The past few years have been very difficult to predict demand because the weather is so hard to predict long term.  Sales are very dependent on pool usage and the start of the season..  A few years ago, we experienced an early, long, and very hot summer.  This was followed by an overly cool summer.  They were out of stock on many items and lost sales in the hot and early summer year.  In contrast, they ended the cool season with a glut of inventory.  They have had a hard time predicting and planning their sales for the past five years.  The bottom line is that climate change has introduced more variation into the system and this is the reality they have to deal with moving forward.  No amount of company data combined with external data sources will be able to predict next summer’s weather with the accuracy the company would like to have..  

They could consider channelling and using data from the following sources::

• Climate Prediction Center of the National Weather Service
  http://www.cpc.ncep.noaa.gov/products/forecasts/

• The Farmer’s Almanac
  http://farmersalmanac.com/long-range-weather-forecast/

• For historical data to make ones own forecast:
  National Climate Data Center of the National Oceanic and Atmospheric Administration
  http://www.ncdc.noaa.gov/

The firm could use their own history and the data they can glean and feed from the above sources.  They could have the best data and forecasts available in December when the need to firm up their orders for the summer of the next year.  They have all kinds of info and they have to decide on what the climate will be next summer in their major markets to determine the quantities of their products to order and have ready for the beginning of the season.  Basically, they still have to make a long term weather forecast to make a sales forecast.

It is not clear that Big Data and Analytics of any kind will help them to any great extent at this point in time.  

This is where the promise and hype is greater than the what is actually possible.  It is reminiscent of the dot.com hype.  Because parts of transactions could be done online did not necessarily lower the cost of the entire transaction.  People thought that grocery stores would be eliminated because people would simply place orders online, the product would be delivered to their home, and the everyones grocery bills would be less.  Grocery stores, as it turned out, exist not because people could not remotely place their order but rather because the it was cheaper overall for consumers to pick their selection and tote them home than to have the grocer do it.  

Bottom line?  The more real time and deterministic the application of Big Data and Analytics the better.  The more that forecasting methods are required to make decisions, the less likely that Big Data and Analytics will be able provide any greater insights.