Business Intelligence Blog from arcplan

Predictive Analytics: Examples, Advice and Shortcuts


Everyone is throwing around the term "analytics" – about as much as they're throwing around the term "big data." While I might put big data on my list of the Most Overused Phrases, analytics gets a pass. As companies realize the amount of insight and value they can glean from their ever growing volumes of data, there has been a surge in analytics initiatives. The goal of these projects is to use data to analyze trends, the effects of decisions, and the impact of scenarios to make improvements that will positively impact the company's bottom line, improve processes, and help the business plan for the future.

In order for analytics to remain relevant and always provide value, organizations must continually up their game. One way to do this is with predictive analytics, which is becoming more mainstream every day. If you stick around to the end of the article, I'll tell you a simple way to bypass its complexity and still get the predictions you need.

Gettin' Predictive With It
Ventana-predictivePredictive analytics involves making predictions about the future or setting potential courses of action by analyzing past data. A 2012 benchmark study by Ventana Research revealed that predictive analytics is currently used to address a variety of business needs, including forecasting, marketing, customer service, product offers and even fraud detection. While predictive analytics used to be in play in only a small number of companies, two-thirds of companies participating in Ventana's survey are using it, and among those, two-thirds are satisfied or very satisfied. These results indicate the maturity that predictive analytics has undergone over the last few years, as technology has advanced to make it less expensive and more approachable, and therefore easier for more areas of the business to make use of. At this point, it's safe to say that most Fortune 500 companies are churning out predictive insights on a regular basis, but that doesn't mean smaller companies without "big data" can't do the same thing. They can supplement their internal data with external data from social media, government agencies, and other sources of public data to get the insights they need.

In Practice
Let's take a look at finance institutions, which have predictive analytics down to a science. Financial service companies have long relied on customer data, like spending patterns over time and bill payment history, to assess risks in making lending decisions. Predictive analytics plays an integral role in decisions on whether or not to grant a loan, and to determine the amount and terms of a loan…even though these days, it seems the answer is always "no."

The manufacturing and retail industries are also realizing that predictive analytics can open doors for them (literally); they're predicting seasonal demand of products to match the supply and performing "assortment optimization" to ensure that best-selling products are always kept in stock. We have one customer – a grocery chain in the southern US – who uses arcplan to predict the inventory implications of impending storms. Another of our retail clients uses arcplan to properly stock its stores during the holiday season. Predictive analytics are also used to glean customer insights, predict buying patterns, offer promotions based on those buying patterns, and make suggestions for products that would complement a purchase. In effect, predictive analytics empower business teams to anticipate their customers' needs.

Even sports teams are using predictive analytics, beyond what you may have seen in Moneyball. Behind the scenes at the Orlando Magic basketball team, they're optimizing ticket sales, helping coaches choose the starting line-up for each game, and even determining which players will provide the best ROI.

The Dangers
There's a danger here though, as pointed out in the Ventana research. Predictive analytics is still complex and requires special skills. Fifty-eight percent of survey participants said they don't understand the underlying math concepts; this is why hiring specialized data scientists can be critical to the success of any predictive analytics initiative.

Another danger is underestimating the amount of data preparation required before even beginning the project. We frequently mention on this blog that data is the key to BI project success, but that companies often underestimate the amount of data clean-up they need to do before getting the right insights from their BI software. The same goes for predictive analytics – you may build the most accurate model for predicting which customers are likely to purchase from you in the next quarter, but if the data is a mess, it's all for naught. This article, which talks about the importance of data prep, reviews the 9 steps to predictive success and is a great starting point for both business and IT.

Some Advice
One suggestion I can make to anyone embarking on a predictive analytics project is this: keep the business problem at the center of everything you do. Don’t let it become an IT exercise. The success of your initiative will be determined by how well it provides actionable insight, how well the business accepts and acts on the predictions, and ultimately the impact it has on the business. So set the goal first – like upselling customers by x% to increase profitability. It doesn't have to be super specific, as discovery of correlations and new strategies is part of the excitement of predictive analytics, but starting with something quantifiable will help you define success, which is how you'll get further support for similar projects in the future.

A Shortcut?
Now I promised you I'd cover an easy way to bypass some of the complexity of predictive analytics and still get the answers you need. Surprise – it's arcplan! arcplan's BI platform includes built-in formulas, like one for linear regression, that help users analyze historical data to make predictions without needing a math degree. The system does the heavy lifting for you in modeling the relationship between a dependent variable and one or more explanatory variables – for example, sales over time.

In the video below, you'll see how a regression analysis formula is used to predict 12 months of future sales data based on 3 years of historical data. The user can change the assumptions with the click of a mouse and the system instantly delivers a new regression line that projects updated future data.


Click to enlarge

In addition to formulas, arcplan's simulation functions provide an easy entrée into predictive analytics. An example is Monte Carlo simulation, a computer model that essentially makes predictions. It's a risk analysis technique that shows you all the possible outcomes of your decisions by running multiple trial runs (simulations) with random variables. As you can see in the image, it displays the potential outcomes and probabilities they will occur, from the most extreme action to the most conservative. Companies use Monte Carlo simulation to determine how many units of a product to order, for example, to achieve the greatest profit and least overstock.

If you're not ready for a full-scale data scientist-driven, model-building predictive analytics project but want to incorporate predictive elements into your existing analytics applications, BI platforms like arcplan make it possible. It's this type of analysis that will become your company's competitive advantage.

Dwight deVera

About Dwight deVera

I'm Senior VP responsible for Solutions Delivery at arcplan in North America. I also present on a lot of arcplan webinars, so you can sign up to hear me - the events listing on our website is located here: You can also follow me on Twitter: @dwightdevera.