Making Predictions Through Correlations
A correlation is a form of association between two variables: for example, height and weight are correlated because taller people tend to weigh more. Correlations can be statistically tested to ascertain their validity. According to Dr. Jordan Sudberg, one misconception about correlation is that it implies the causation of a particular phenomenon. Two phenomena can be highly correlated, but only one of them is causal.
5 Ways A Company Can Make Predictions Through Data Correlations
1. Learn About How the Company Operates
Learning about a company’s business operations and its customers is essential. For example, McDonald’s has discovered that it can predict its sales by correlating its breakfast sales with the temperatures in different parts of the country. McDonald’s also connects its sales with how many people attend sports events in a particular area.
2. Look for Patterns
It is often possible to find patterns among the data relationships of different companies and industries. For example, companies can look at their data and similar companies to find ways to help predict what will happen in the future. For instance, if a company discovers that working too many hours at a time leads to an abnormally high number of workplace injuries and accidents, it can use that information to predict that it may be coming up on a busy period and adjust accordingly.
3. Look for Contradictory Data
It is possible to look for data that contradicts what seems logical, leading to new insights about the business. For example, the data could include information about customer preferences or the performance of a particular product over time.
4. Use Consumer Data
Learning from talking with customers and observing them in stores is possible. For example, a store that has been attracting customers with a particular product in its cafeterias may discover that the main reason for its success is that it uses a specific type of broiler and serves large portions of food.
5. Use Industry Data
Industry data can help companies predict what will happen to their business in the future. For example, instead of predicting the weather, it may be possible to expect if there will be an increase in sales due to an upcoming sporting event even though no one knows when or where such an event will be held.
As Dr. Jordan Sudberg pointed out, one of the things to consider about using data to predict the future is that correlation does not prove causation. The data may suggest a pattern, but there can be any number of other factors causing the way in the first place. One example is that in certain parts of China, wages are high, and housing and food prices are high, while other parts of China have low salaries but low prices on those same items. In this case, people’s incomes are correlated with their spending habits, but it doesn’t necessarily mean that higher income levels come from higher spending levels. Dr. Sudberg also suggests that companies should use correlations for decision-making.