Three Statistical Concepts All Marketing Professionals Should Know and Use

Monday, 11th March
Abbe Lefkowitz

I genuinely love statistics, and I’m getting my Master’s degree in Applied Statistics to prove it! Most people are scared of statisticians because we use lots of weird Greek letters and never can say anything with 100% confidence. However, there are plenty of simple statistical concepts that marketing professionals can and should understand. Here are my top 3 simple statistical concepts that everyone should know:

1. Sample size

Sample size is one of the most basic statistics used to describe a dataset. It is just the size of the sample! As simple as it sounds, sample size is fundamentally important to statisticians because the size of a sample can make or break an analysis. Too small of a sample, and the results may not be significant. Too large of a sample, and the data can become unwieldy and difficult to manage.

In the marketing world, it is always important to be mindful of sample size, especially with all of the recent buzz about “big data” floating around. It seems that data sets are only growing; I will have the opportunity to work with a data set that has upwards of 400 million records. In a situation like this, sampling will come in handy because it will reduce processing time and make for a quicker analysis.

Generally speaking, sampling 10% of the population is a good rule of thumb in statistics.

2. Mean/median

The mean and median are both statistics used to measure the center of a sample. The mean is what most people refer to as the average, and the median is the middle data point (or spot between two middle points if the sample size is even). These two statistics can have similar or different values depending on the distribution or shape of the data. Conversely, knowing the mean and median can allow the distribution of data to be determined if it is unknown.

In the image below, three different distributions are presented: symmetric, right-skewed, and left-skewed. A symmetric distribution has the same mean and median. A right skewed distribution has a greater volume of data toward the left (skewed right refers to the direction of the tail), so the mean is greater than the median. The opposite is true of a left skewed distribution.

In marketing, I have found the mean to be the appropriate measure more times than the median, but that does not mean the median cannot be used. The mean is more sensitive to outliers and influential points, while the median is more robust. That is why the mean gets pulled in the direction of the tail for a skewed distribution. Therefore, when there are many extreme data points, the median should be used instead.

For example, in a traffic by source report (direct, banner ad, search, referral, social networks, etc) one can determine where people are coming to the site from. If there are normally 30 visits from social networks but one day there is a jump to 2,500 visits due to a widely popular post of Facebook, the visits from social network for that day would be considered an outlier. They would falsely inflate the year-to-date mean number of visits from social networks, so using the median would be useful in this case.

3. Correlation/causation

“Correlation, not causation” is a catch phrase commonly used by statisticians. Knowing the difference between correlation and causation is one of the most important concepts in statistics (and life). Many people do not know their proper definitions and often confuse the two.

Correlation occurs when two or more variables have a relationship with each other. For example, height and weight are positively correlated because as height increases, weight generally increases as well. These variables display correlation, but not necessarily causation. Height and weight may seem directly causal, but there are other confounding variables that have an effect on a person’s weight such as health, gender, race, diet, exercise, and so on.

Causation is similar to correlation in that it occurs when two or more variables are related to each other. However, it differs from correlation in that one variable is directly responsible for causing a change in other variables, while with correlation that is not necessarily the case. For example, the greater amount of time you exercise, the more calories you will burn and this is a direct causal relationship. These variables display both correlation and causation.

It is so important not to confuse these two concepts. Unless you know for sure that two variables have a causal relationship, always assume correlation and not causation. For example, a company may launch a new marketing campaign with a lot of banner advertisements and then receive a lot of traffic to their site. However that same week there was also a Facebook post by that company about winning a free prize. While the number of ads and number of visitors to the site are positively correlated, it is not correct to assume that the ads themselves were responsible for increased traffic because the Facebook post is a confounding variable most likely causing the traffic.

With a new understanding of these three statistical concepts, you are now ready to impress your co-workers, peers, and colleagues. At your next networking event, why not strike up a conversation about correlation? Some will chime in, and those who do not will just think you are really smart, so it’s definitely a win-win. Knowledge is power!

MaassMedia works with a variety of clients in a range of industries. If you’d like to know more about how your organization can use marketing analytics to find Transformative Insights™ about your audience, contact us today.