What will people pay for?

People have different preferences about what to buy, and one of the main jobs of marketers is to appeal to those preferences so they can sell their products. It wouldn’t be a useful exercise to just ask the open ended question “what do you want to buy?”. As Steve Jobs said, “People don't know what they want until you show it to them”. Conjoint Analysis is a market research method that shows people different combinations of attributes for a new potential product (or an existing product on the market) and asks the consumer to choose between usually 2-3 options. Attributes could be different prices, product features, taglines, logos, packaging, colors, really anything you deem important to the success of the product.

These cards containing unique combinations of attributes get shown at random to the target group, and therefore you can get relatively unbiased data on which features are actually important to the market. For example one user might see a blue product that costs $40 with the tagline “Win Parenting”, next to a red product at $50 with the tagline “Be Healthy. Another user might see a red product at $40 with the tagline “Be Healthy” next to a blue product at $50 with the tagline “Win Parenting”. Using Regression Analysis, it’s possible to disentangle these choices and determine what the relative importance was of each feature in terms of predicting user preferences.

Typically these analyses are done with a specific survey tool like Qualtrics, Conjointly or Sawtooth. These tools handle the asking of the questions and the analysis afterwards, as well as providing flexible rules for how those questions are asked. It is possible for you to create your own survey however, you just need to account for every combination of the product attributes, which can be tiresome to set up. For example if you had 4 taglines, 2 colors, and 3 prices to test, that would be 4 x 2 x 3 = 24 variations. One thing to remember in terms of the final analysis: when doing regression analysis you need to drop one category from each attribute. So for example the data would have columns for 3 out of 4 taglines, 1 of the 2 colors, and 2 of the 3 prices. This is because the first row would have all zeros, and therefore would be the baseline in the regression. It would represent the option of Tagline A, Color A, and Price A. Then as you move down the rows it would be a zero or one depending on if that attribute was present in that card or not.

Once you’ve got your data and you do the regression analysis, you’ll have coefficients you can use to predict preferences. You can simply select the coefficients that are highest, one from each attribute. It’s also valid to choose for example a higher price if the coefficient is not too much more negative, as that might improve the profitability of your product. In advanced Conjoint Analysis tools it’s also possible to use these coefficients to predict market share, though that’s complicated to do, relatively theoretical, and rarely correlates to the real world, so is outside the scope of this tutorial.