Course · 1. Background and motivation
What conjoint analysis is
Ask people what matters to them and they will tell you everything matters: quality, price, brand, size. Ask them to choose and you learn what they actually trade off.
Conjoint analysis is a family of survey methods built on that idea. Respondents see a few products at a time, each described by a handful of attributes (brand, flavor, can size, price), each set to one of a few levels (Volt or Surge; Berry or Citrus; $1.99 or $2.49). They pick the one they would buy, or "none". After ten or so of these choices per person, and a few hundred people, a statistical model can work out how much each level is worth to them, including in money.
Why choices beat ratings
- Trade-offs are forced. You can rate everything 5 out of 5; you can only pick one drink off a shelf.
- It looks like the real decision. A shelf of options with prices is closer to shopping than a rating scale.
- Value comes out on one scale. Because price is one of the attributes, every other level can be expressed as willingness to pay: "people would pay $0.32 more for Surge than for Volt".
What it is used for
| question | what conjoint gives you |
|---|---|
| Which features matter most? | importances and part-worths |
| What can we charge? | willingness to pay, price curves |
| Which configuration wins? | a market simulator to test products against competitors |
| If we launch X, where do its sales come from? | source of volume: how much it takes from each existing product |
| Who wants what? | how tastes differ across people |
Most of the value comes out of the simulator: you describe a market, the model predicts shares, and you change one thing at a time. The rest of this course is about getting that simulator right, and the part standard methods most often get wrong is in the next pages: how people switch between similar products.
Next: From choices to value