By Luke Allison, TransPro Senior Project Analyst
“Statistically valid results.” When it comes to customer surveys, the term is critical. But what does it really mean and how is it achieved?
What Are “Statistically Valid Survey Results”?
By definition, statistically valid results refer to data deemed accurate and reliable. This matters because when we survey customers, we work with a sample of customers. Our goal is for that sample subset of customers to accurately and reliably reflect the opinions of the entire customer base. Surveying all customers is time and cost-prohibitive. Thank you, Statistics, for making accurate and reliable sampling possible!
How Do We Achieve Statistically Valid Results?
Statistical validity is born from data collection methodology. Statistically valid results are not fundamentally based on how many customers you survey, but rather in HOW you select them. The key ingredient? Randomness.
Randomness, in surveying, is the great equalizer. A 400-customer sample collected randomly is far more likely to yield accurate and reliable results than a 1,500-customer sample without randomness. That may seem counterintuitive, but it has been proven time and again: without randomness, survey results risk reflecting a subset of your customer base rather than the entire population.
This is why, at TransPro, we employ a randomized intercept method for surveying customers. Our surveyors approach every fourth customer, a technique known as systematic sampling. Because no one can predict who the next fourth customer will be, bias is minimized.
- Maybe that customer would have asked to fill out a survey on their own – maybe not.
- Maybe they are shy or hesitant due to language barriers – maybe not.
- Maybe they are from a demographic minority who assumes their voice won’t be heard – maybe not.
- Maybe this is their first time riding transit – maybe not.
Randomness ensures we hear from all customers – their “maybe not” counterparts.
Conclusion
Our sample goal is simple: a sample that truly reflects the customer base. That’s another way of saying we want statistically valid results. The best way to accomplish this? Random sampling.
This is another feather in the cap for randomized intercept surveying compared to online, mail, and phone.
Here’s to randomness – for giving us statistically valid results!
