Improve Analysis Within Your Business By Using Statistics

Most business leaders have encountered statistics at some point in their careers, but I suspect for many this was in their undergraduate coursework and they never really looked back. The implication of this is that many businesses do not actively use statistics in their decision-making framework. I think there are two reasons for this. First, they may be interested in applying statistics but the cost is prohibitive. SAS and SPSS are expensive packages for small and medium sized businesses. I addressed this concern in a recent article where I advocated for bringing in the free R statistics package if you do not already have an advanced statistics solution. I invite you to review that article for a more in depth discussion about R, but to summarize I will just say it is free, it is robust, and it is awesome.

The second reason that may prevent business leaders from adopting statistics into their decision-making framework is that they may be unfamiliar with the different methods that may be applicable to their business. Adding in a new statistics tool does nothing for you if it is not incorporated into the analytics process within your business. So here are some ideas for enhancing your business decision-making with the use of statistics.

  • Segment your customers using k-means clustering. You collect a lot of data on your customers, whether it is demographics, addresses, phone numbers, dates, purchases, behavior on your website, and much more. The k-means method groups your customers based on these data attributes. The resulting output will be x number of groups of similarly situated customers that can be used to create personas of your customers. This can help you identify the patterns of your most valuable customers and can help you tailor your marketing strategy to target each group.
  • Forecast with linear exponential smoothing. Most businesses are subject to seasonal fluctuations within their annual business cycles. Electronics retailers may peak in the holiday buying season while auto parts retailers will peak during the warmer months. A great way to quantify this annual trend is a method called linear exponential smoothing, which deconstructs your trends between natural growth and seasonal fluctuations based on historical data. I have found this to be a simple and useful method for building baseline forecasts. This page from Duke University provides detailed instructions on how to create this model in Excel.
  • Test your theories with controlled experiments and hypothesis testing. A common trend I have observed among many businesses is to implement a new campaign, price, product, etc. without experimenting first. Either the change replaces what was in place before and results are compared or there is no analysis at all. There are two problems with this. First, the change may result in a negative impact on your business, so your business loses money. Second, the change may indicate negative or positive results, but since it occurred over different timeframes you may make decisions based on seasonal or macro factors rather than the change itself. By defining experiments with control and test groups, you can test one to many different changes at the same time with minimal risk while controlling for time and macro variability. A controlled experiment will require you to define your theories as testable hypotheses and define a population from which you can create smaller samples for statistically significant testing. Comparing the observations of the similarly situated control and test groups using the t-test or chi-square test can identify which changes are statistically significant (i.e. not explained by chance), which will help you make informed decisions.
  • Predict and gain insights with regression modeling. Above I reviewed modeling using linear exponential smoothing, which I like to use as a simple method for baseline forecasting. However, it only helps to explain what is going on and provides no insights into the more important question of why. Regression modeling is a method for predicting outcomes based on one or more input variables. With macro, seasonal and internal data you can predict movements in your market, customer actions, experiments and other events using regression. I also use regression to gain insights into the drivers of the outcomes. A well formed model can help you understand and quantify the most impactful factors on your business, which means you will have a better understanding of why.
  • Recommend products or content with association rule analysis (ARA). Your business collects large amounts of transactional data, so use it to uncover patterns in your data. ARA, or shopping basket analysis, is a data mining technique that quantifies the relationship between observations. The application of this technique can provide valuable insights for retailers and content marketers. For retailers, you can identify the relationship between products for the purposes of making recommendations. If you are a convenience store that sells charcoal, the data may indicate plastic cups are frequently purchased with charcoal. With this knowledge you can then test placement of these separate product categories together to encourage purchases. For content marketers, the analysis may indicate relationships between content, which can be used to enhance the experience of visitors by making article recommendations.


These are just a few ideas of how to apply statistics in your business and I feel the application of some or all of these may generate great value for your business. It’s your data, so why not use them?