If you work in IoT, you’re no stranger to leveraging data for business applications. But are you using that analytics know-how to market your own organization?
Regression analysis is a statistical method for estimating how independent variables affect dependent variables. To answer questions like “which factors are associated with my employees scoring higher on performance reviews?” you can use regression analysis. There’s a reason it’s the “go-to” method in analytics. Use it effectively and get powerful business intelligence.
Regression analysis may sound intimidating, but it’s just a way to see the relationship between variables. To explain how it works, let’s walk through a hypothetical scenario.
Say you own an ice cream truck. You have a hunch that you sell more ice cream on hot days, but you want to see if the data backs that up. You decide to use regression analysis to answer your research question. The independent variable (x) is temperature, the dependent variable (y) is ice cream sales. If temperature increases, do ice cream sales increase?
Fortunately, you’ve been recording both sales figures and the local weather report over the years, so with a few strokes of your keyboard, you can answer this question. Using the statistical software of your choice, you collect all the data points in a scatter plot, then fit a regression line through it.
The line is going up, indicating a positive relationship between our independent and dependent variable. In other words, on hotter days, you do indeed tend to have higher ice cream sales.
If you wanted to predict sales figures more accurately by including other variables (foot traffic, school holidays, local unemployment, etc.) you could build a multivariate model using related methods.
Of course, you can use these techniques for a lot more than ice cream. Depending on your business, there’s all sorts of questions regression analysis can help you answer.
Identifying the factors that influence sales is one potential use. Does national disposable income affect sales? Is higher marketing spend associated with increased revenue?
If you have access to location-based data, that opens up a whole new realm of insights. In the telematics sphere, a recent study used regression analysis to help identify the most risky driver behavior (it’s speeding).
Regression analysis is powerful, but it isn’t a magic bullet. There are a few major caveats you need to consider before you start making decisions for your IoT business based on regression analysis results.
Analysis performed on shoddy data is worse than nothing at all. Don’t lose sight of the fact that those numbers should represent reality. If they don’t, don’t use them.
Don’t be fooled by a straight line through a mess of dots. Just because there’s a relationship between two variables doesn’t mean it’s a strong one.
Returning to the ice cream example from earlier, here’s a famous cautionary tale shared with Stats 101 students everywhere. Ice cream sales and drowning deaths have a strong, positive relationship. That doesn’t mean ice cream causes drowning! Instead, both ice cream sales and drowning deaths tend to increase in the summer.
Think through results before implementing them. Regression insights are a jumping off point, not the end of the road.
Here are five steps your IoT firm can take to start using regression analysis to make business decisions.
You can’t analyze data without data to analyze. This can be a frustrating obstacle for companies who have only recently taken an interest in analytics, but you can’t make sound sales predictions if you only have three months of sales data.
Once you do have data, make it clean and consistent. No punctuation, mixed currencies, or inconsistent reporting methods (gross or net, pick one).
While the underlying concept of regression analysis is simple, the nitty-gritty can be time-consuming for the amateur. For an IoT business, an analytics expert is a good investment.
Regression analysis is most powerful when it’s well-thought out and specific. At the end of the day, all those numbers and calculations are connected to real-world logic. You’ll get better results from asking “did new client acquisition increase with digital ad spend?” than “what increased sales?”
A correlation coefficient shows that there is a relationship, not how it works. Once your analysis has established a relationship between variables, think outside the box about why they may be related.
Once you have your insights, think carefully about how to use them. Establishing a relationship between variables doesn’t tell you how best to capitalize on that relationship: that’s for you to decide.
Regression analysis is a powerful tool, especially for IoT professionals who have access to a wide range of data. Still, it’s a tool, not a panacea. Without clean data, smart analysis, and thoughtful utilization, it’s not worth incorporating in your strategy.