When you work in market research and deal with analysis long enough, you'll run into a lot of techniques and statistical models. Some last while others phase out quicker than the XFL. Although the term multivariate analysis may seem complex, it's a relatively common and easy to adopt approach in your daily market research projects.
What is multivariate analysis?
It is any type of statistical analysis that reviews more than one variable. As a marketing strategist there is a lot of value in comparing the importance or impact of specific variables. If you are launching a new product or service and the likelihood to purchase is extremely low based on your survey results, you'll want to know if this was caused by price, the features, the color, the design, etc. Using multivariate techniques in market research help add clarity to this process and support a data-driven strategic approach. As I mentioned to a client the other day, "it helps quantify your gut feeling."
Here are 3 examples of multivariate analysis:
1. Multiple Regression
This uses your long list of grid satisfaction ratings and works them into a model to make a prediction as to which factor has the most impact on overall satisfaction or likelihood to purchase. It's used often in forecasting. It starts by asking your respondents their overall likelihood to purchase on a 1-10 scale followed by asking them the appeal of each of the following features (price, feature 1, feature 2, color, etc.) using the same 1-10 scale. Using regression you will discover the feature has the largest impact on likelihood to purchase. Therefore it helps you understand priorities as to what features or benefits might cause a product failure and what needs improvement.
2. Conjoint Analysis
This is most closely associated with "trade-off." It is asking: would you rather have feature 1 or feature 2? Would you rather have color options or a lower price? Would you rather have feature 1 or color options? And so on. Conjoint is a common form of trade-off analysis in online survey research. Just keep in mind, if you have a long list of factors it may seem very repetitive to the survey taker, so look for ways to randomize lists across series of respondents or limit your factors. Long statistical models like this in surveys often increase the rate of drop-outs.
3. Discrete Choice Modeling (DCM)
This is similar to traditional trade-off conjoint but typically includes a series of variables. In this option of multivariate analysis, the survey designer creates a packaged concept and forces a choice. Think of it as almost a Package A or Package B question.
Drive Research is a market research company in Central New York. Questions or need advice on your next market research project? Visit our contact us page. That will make it easy for you.