What will customers trade off?
Run a conjoint analysis on your survey data
Customers say every feature matters. A conjoint analysis shows what they give up to get one. Put the choices or ratings from your survey in one file and see what each feature level is worth next to price.
What a conjoint analysis tells you
Instead of asking people to rate features one at a time, a conjoint study shows them whole product profiles and records which one they pick, or how they rate each. From those answers the tool estimates a part-worth for every level of every attribute, how much of the total each attribute accounts for, and, when price is one of the attributes, what a feature is worth in dollars.
Three questions this answers
- Which attribute drives the choice most: price, or something else?
- What is each level worth next to price?
- Does the average hide customers who want opposite things?
What your file needs
- Choice data: one row per alternative shown, a column marking the one chosen as 1 and the rest as 0, and a column that groups the alternatives shown together. Each task needs its own identifier.
- Rating data: one row per respondent and profile, with the rating or rank.
- One column per attribute, each holding a few tested levels. Include price as one of them if you want dollar values.
- Optional: a respondent column.
- Enough rows to estimate every level: at least 20, and at least three for each level. The tool says which levels it cannot separate if the design never varied them independently.
What you get back
- The part-worth of every level, with a confidence interval.
- The importance of each attribute, shown next to the range of levels that was tested.
- Willingness to pay when a price attribute is found and respondents gave up value as price rose. If they did not, the tool reports no dollar value instead of a nonsense one.
- A comparison of product configurations built from the levels you tested.
- With rating data and a respondent column, a check on whether the average blends groups with opposite preferences.
- The method named and the R code.
Questions people ask
What is conjoint analysis?
A way to measure trade-offs. People choose between, or rate, whole product profiles, and the analysis works out how much each feature level and each price level contributed to those choices.
How many respondents do I need?
The tool needs at least 20 rows and three rows for each level it estimates, and it shows an interval for every level so you can see how firm each one is. More respondents narrow the intervals.
Choice-based or rating-based?
Either works. The tool reads the file, decides which of the two it has, and says how it read it. Choice data is fitted with a multinomial logit, rating data with a least squares fit.
How do I get willingness to pay?
Include price as an attribute whose levels are money amounts. The figure is a stated-preference estimate, not a price customers will actually pay, and the report says so.
Does it forecast market share?
No. It compares configurations built from the levels you tested. It does not forecast what the market will do.
Have your survey results ready?
Try the example choice file first, then drop in your own.