Data Analysis

What's a good survey response rate, 1%, 15%, 50%? It's the wrong question. What actually matters is whether you've got enough responses to draw a conclusion, and that's a different number for every situation. Here's how to calculate the response rate you actually need, with the maths shown.

Statistical packages are powerful, expensive and overkill for most customer feedback work. Excel can do almost everything they can do, and it's already on your desktop. Here's a step-by-step walk-through of analysing survey data in Excel, simple stats, error bars, histograms, with a video to follow along to.

Customer feedback isn't a thermometer reading, it comes back tangled with halo effects, three-year-old memories and irrelevant tangents. That doesn't make it useless. Here's why messy data is still rich data, and how to stop "the data isn't clean enough" being used as a reason not to act.

A great score on one feedback attribute often means nothing more than a great score on brand image colouring everything else. Psychologists have known about the halo effect for a century, most CX teams still report scores as if it didn't exist. Here's how to spot it in your data, with a worked example from a Canstar mobile-phone survey.

Comparing two survey averages and calling them different (or the same) is the most common mistake in customer feedback reporting — and it leads to killing things that are working and persisting with things that aren't. Here's how to run Student's t-Test in Excel to know whether the change is real, with a step-by-step worked example.

"I don't believe that NPS, you don't have a big enough sample size." It's becoming the go-to objection from anyone who doesn't like a feedback result. Most of the time it's wrong, and a five-point dataset can be statistically meaningful. Here's what sample size actually means, and how to get past slide one without an argument.