When the focus is on metrics or statistics, there are really three things that marketers must know to use them effectively.
#1: Observed and Expected
Statistics typically are based on a simple equation: “Observed minus Expected.”
Ex 1: Comparing the target audience (the Observed) against the total population (the control or Expected). High or low indices can inform marketing strategies and tactics.
Ex 2: Assessing company sales performance, where the actual sales (Observed) are compared against prior sales (Expected).
Even as statistics get more complex, we see this. An “average” (aka “mean”) is another way of saying “Expected.” Conjoint (trade-off) analysis, a complex multivariate technique, works under the assumption that at the start, all attributes have an equal chance of being selected.
Most marketers don’t know it, but they “talk” in “Observed minus Expected” all the time: “If we do A, I hope to see B change by X.” Measuring that goal becomes the foundation for designing research, for the statistics used in analysis, and for seeking unexpected insights.
(For those of us into social media/discourse analysis and data mining, a special note: there often isn’t an “Expected” with these methods, which is one reason why they’re harder to analyze. The goal, moving forward, is to start establishing benchmarks against which findings can be assessed – or it will always be difficult to separate new, critical insights from the undefined norm.)
#2: Correlation vs. Causation
Because two things are related doesn’t mean that one thing caused another. We know that over-eating (unfortunately) causes most people to gain weight. However, most smokers – some say 90% of them – don’t get lung cancer. Smoking doesn’t cause cancer, but it’s clearly a (correlated) high risk behavior because many more of those who smoke will get ill when compared with those who don’t.
How does this impact marketers? You can’t really assume that doing A will cause B to happen. However, business models built on connections between operations, consumer attitudes, behaviors, etc. are very powerful in finding ways to reach business goals.
#3: Compound Findings
Independent facts can’t be compounded or strung together to simplify the “story” being shared with others.
Does this example tell us that our target should be women, age 25-34, who participate in outdoor sports? Not necessarily.
Assuming a perfectly distributed sample, we’d estimate that the target is 25% of our population (65% x 65% x 60%). But what if all the men – this is a hypothetical! – said they were into outdoor sports? That would mean that only 25% of the women said the same – reducing our estimated target size to 11% (65% x 65% x 25%).
While compounding or grouping results makes for a clearer story, marketers need to know that the story may not be accurate.
