How to Use Business Statistics to Your Advantage

gorodenkoff/iStock/GettyImages

How often do you make decisions by trusting your gut? Perhaps this works in your personal life, but the business world offers so much information that you can and should leverage before making decisions. After all, with the financial risk associated with running a company, why not take the time to evaluate as many business statistics as possible to make informed decision making?

Sometimes, you'll still rely on your personal experience and that intuitive feeling in your gut, but collecting business statistics will keep you aware of trends among your customers and competitors. In turn, this helps you stay poised to make strategic moves when the window of opportunity presents itself.

While it will likely help you tremendously to invest some time and money in an introductory statistics or business statistics course, this outline of data analysis will give you an idea of what course materials may look like to get your mind flowing. Business decisions will come much more naturally and accurately with some statistics under your belt.

So, at what confidence intervals are you currently making your decisions?

Why Bother With Business Statistics?

Business statistics have a range of applications and help business owners and managers to make decisions at all levels. For example, statistics could be used to determine whether the business's selling proposition is viable, which concerns the entire company, or whether a particular newsletter design converts more leads, which mainly concerns the marketing department.

Stats prove useful in guiding product development, marketing campaigns, market research and even internal processes. For example, if you want to switch to a more productive project-management software, you first need to gather data about your team's current level of productivity and analyze weak areas.

Simply collecting data isn't enough to help business owners make decisions. The information also has to be interpreted, which is where various statistical analysis techniques come into play. Often, the data can be run through a computer program to deliver mathematical calculations in the blink of an eye, but only a human team can understand the significance of those calculations and make a strategic decision as a result.

Data-Collection Methods

Numerous types of data-collection methods exist, and the method you choose depends on your end goal. Don't assume that you should only collect quantifiable data, such as sales numbers, because statistical analyses can still be performed on complex qualitative data, such as survey results.

So, start by making a list of questions that you want answered followed by a list of the data that could provide answers. The final list should include potential ways to procure that information. Some common data-collection methods include customer surveys, focus groups and personal interviews as well as simply keeping detailed records on every purchase, lead, marketing channel and endeavor. Experiments, typically in the form of A/B testing, can also provide insightful data.

Also, you can collect meaningful data from outside your organization. Look up information about your competitors to understand their strengths, weaknesses and foothold in the market. The government, especially the Small Business Administration, also publishes free market reports about many industries. If your business budget allows for it, ready-made market reports can also be purchased from third parties.

Avoiding Common Statistical Biases

Finding an appropriate source for your data is just the first step. Next, you have to consider whether any statistical biases exist in the data you've collected or are about to collect. For example, you need to have as large of a sample size as possible for an accurate analysis. You also need to ensure you look at all appropriate variables ("omitted variable bias") and don't exclude important data points ("survivorship bias").

Biases are especially problematic for data-collection methods like surveys or interviews. If you want to send out a survey, choosing to only send it to your closest friends is an example of selection bias. To combat this bias, try to get random samples of survey respondents as much as possible. However, the people who choose to respond present a self-selection bias, which means you won't get the opinion of people who simply aren't interested enough to participate.

An observer bias tends to occur in focus groups or one-on-one interviews and happens when the interviewer asks questions in such a way that implies an expected answer. If respondents don't remember events very well, they have a recall bias. Although it may be difficult to avoid common statistical biases, being aware of their presence can help you interpret the data with a grain of salt.

Using Statistical-Analysis Software

You don't need a Ph.D. in math to learn how to do statistical analysis, especially since so many tools and programs exist with the sole purpose of making this as easy as possible.

For example, once you know what kind of an analysis you'd like to run on a data set, you can find and select that function in Microsoft Excel or Google Sheets. With the push of a button, you can have the results in front of you, ready for interpretation. If you want a little more functionality, enhanced visuals or perhaps some built-in interpretation guidance, numerous statistical-analysis software programs are available on the market.

Certain departments within your business can use tailor-made analysis software to build reports and easily look at trends. For example, marketing teams typically track website data such as page views, traffic sources and visitor behavior using a tool like Google Analytics.

Your accounting department could tap into the analytics on its financial program, like QuickBooks. Programs like Salesforce allow the sales team to tap into data about leads, and there's even software for the human resource department to track employee data.

Many other programs that your business already uses have an analytic component. For example, the larger social-media platforms like Twitter and Facebook have built-in analytics to help you track your brand's performance.

Project-management tools have stats about task completion, newsletter programs display stats about the open rate or unsubscribe rate and proposal websites offer stats about the deals you land or lose. Before you try to reinvent the wheel, explore the business statistics currently at your fingertips.

Common Statistical Techniques for Business Data

Now that you have data, it's time to analyze it. Your statistics software or built-in analytics programs might give you some clues as to how to evaluate and interpret your data. However, more often than not, you'll need to tell the programs what to do with the data.

So, what in the world can you do with all this data? It partially depends on the kind of data you gathered and the questions that you want answered. Looking for relationships between variables? Then you'll want to perform a measure of association, the most popular of which are a regression analysis and a chi-square test.

Do you want to distill large data sets down to just a few significant numbers for a presentation or report? Then you'll want to evaluate the measures of central tendency (mean, median, mode) and data distribution. Are you trying to make sense of survey results? A conjoint analysis could prove useful.

Linear-Regression Analysis and Chi Square

One common question that crops up in the business world is, "Does this variable affect sales/profit/revenue?" The variable under scrutiny could be sales price, store location, time of day, product location in the store or any other factor that piques your curiosity. To answer this question, you need to perform some calculations called "measures of association".

One such calculation is a linear-regression or regression analysis, which compares data sets to determine if the outcome of one variable depends on the value of another variable. After completing the regression formula, you'll end up with a number between 0 and 1 called r-squared (also known as the coefficient of determination).

An r-squared number closer to 1 indicates that the outcome of the dependent variable relies heavily on the independent variable. An r-squared number closer to 0 signifies that the two variables act independently of one another, and modifying one won't affect the other.

Another calculation that evaluates the association of two variables is called a chi-square test. This mathematical formula will yield a result between 0 and 1 called a p-value. If the two variables are statistically significant — in other words, they seem to have a dependent relationship — the p-value will be less than 0.05. A p-value greater than 0.05 suggests that the variables don't have a significant bearing on each other.

Understanding Data Distributions

Once you've gathered quantitative data points, you need to understand some trends in the data. For example, you'll want to know which data point occurs most often (the mode) and calculate the average (the mean) of all the data points so that you have one number with which to work instead of hundreds or thousands. Both the mode and the mean can be a little misleading without first understanding how the data is distributed.

Data distribution refers to the full range of the data from the lowest point to the highest point. Extreme outliers, such as one or two very high or very low data points, can end up skewing the average. That's why it's important to also consider a data point known as the median, which represent the exact middle point in the distribution, as well as a statistic known as the standard deviation. The standard deviation indicates how far away the outliers are from the average, so a high standard deviation tells analysts that it might be better to look at the median versus the mean.

Analyzing Survey Results

When conducting surveys in order to gather data, it's helpful to provide a limited set of answers for respondents to choose. This allows you to analyze the distribution of answers. However, this automatically limits what respondents can say, so it's also helpful to provide an option at some point in the survey for an open-ended response.

Regardless, there are several methods you can use to analyze survey results provided that the survey was set up with these analyses in mind. For example, you can use marketing research techniques like a cluster analysis or factor analysis to look for overlapping traits and values among your survey respondents.

Another useful way to analyze survey results is through a conjoint analysis. This technique helps pinpoint the characteristics that are most important to survey respondents, whether that's a low price, a high-quality product, ease of purchase, free shipping, friendly customer support, etc. What if you built your business with the assumption that free shipping mattered most to your customers when in fact it's the least of their concerns? This is certainly important information.

Interpreting and Applying Statistics in Business

Finally, it's time to put all the tools and software aside and use your brain. Better yet, assemble a team of people to interpret the results and develop a strategic plan.

Do the business statistics suggest that you're doing everything right? It's more likely that the stats highlight areas for improvement. Now it's time to shift from the analytical part of your brain to its creative side: What can you do to improve these stats? There's no right answer, but that's part of why owning a business is such an adventure.

References

About the Author

Cathy Habas specializes in marketing, customer experiences, and behind-the-scenes management. Cathy has contributed to sites like Business and Finance, Business 2 Community, and Inside Small Business. She served as the managing editor for a small content marketing agency before continuing with her writing career.