This is the ninth post in my Customer Research series.
I’ve been dragging my feet on this post because I have imposter complex on primary quantitative research. I learned how to do qualitative research by working shoulder to shoulder with established masters. But I never was formally trained in quantitative research. So I know what I don’t know. I’m not terribly deep in my expertise.
For what it’s worth, I’ll share my not too profound thoughts on online surveys.
There are certain pieces of data that cannot be efficiently or effectively collected with qualitative research, either due to the sample size required to get a trustworthy answer, or because qualitative techniques introduce an unacceptable risk of observation bias in emotionally charged questions.
Here are some example questions that I believe are better asked with an online survey:
- How many people in my customer base own which smart phones, and how does this compare with the population at large?
- How satisfied are my customers with their experience as a whole? (This is the top box question)
- How likely are my customers to recommend my products to their friends and family? (This is the Net Promoter question)
- What attributes of the product or service affect purchase interest most? (Probably best done with a conjoint analysis)
- What is the purchase intent for a particular product or service, presented with our best creative messaging efforts, and set at a particular price?
- What is the pricing elasticity for a particular product or service? (Efficiently asked with the Van Westendorp methodology)
In all of the above examples, statistically significant data is required to answer the questions posed. For instance, the proportion of people with iPhones in your customer base can hardly be credibly answered by interviewing 60 people. You would want to survey 5000 people and get results from 500 respondants (assuming a 10% participation rate). Same with top box/net promoter metrics – they become meaningless when the sample size is too small.
As for pricing, that’s an emotionally charged topic and the presence of researchers can very well affect what subjects say and do – some may be driven to present themselves in a different light than their natural tendencies. Anonymity is a great way to minimize observation bias here.
The only one that I can go either way on is product features or attributes. Yes, you can do a conjoint analysis, and yet, I think you learn much more by qualitative product discovery research. I prefer qualitative in early stages of product ideation, and then one can always follow up with a conjoint analysis in survey form to verify the directional input from the qualitative phase still holds when you cast a wider net.
Running the survey is only half of the equation. Analyzing the result is something else entirely. There is the naive data crunching from raw responses, and then there’s data mining – swimming around in the raw data looking for patterns. I love the fishing exercise, but there are many structured approaches that make this work faster and better. One nifty example is “verbatim tagging”. This basically involves looking for keywords in essay responses to open ended questions. You read all the responses, figure out which words keep coming up, then tag each response for the presence and absence of those words for each response, then you tally them up. This is hugely time consuming but can also offer incredible insights.
I would love to get deeper in quantitative research – anyone has suggestions for books to read or websites to peruse?