How to prevent or fix continually increasing revenue curves in simulations?

Question

Suppose I do a CBC study with a handful of attributes, one of them being price.  After obtaining utilities I run some simulations to get a price curve recording share for each price point.  As expected, as product price increases, product share decreases.  However, when I calculate revenue on these same shares, it is not uncommon to see revenue continually increase as price increases.  This result seems troubling as I can’t tell a client to just continually increase price as high as you want and you’ll keep gaining in revenue.  So in this instance I’m wondering what are some standard approaches or methods employed to “fix” or deal with this situation?  Any thoughts or advice is appreciated, thanks!

Answer 1

Make sure that your simulations represent your client's product, plus relevant competition at their relevant price points.  And, include the None alternative.  If you fail to put in the relevant competition or don't have a very good measure of the None (traditional "None" implementation is often too weak in marketing research applications; dual-response None can be better), it increases the likelihood of seeing problems as you mention.

Another thing to watch for is a small segment of your respondents who might have "reversals" on price.  This could be just due to random noise coupled with the fact that price isn't very important to the respondent, or due to bad respondents.  If some respondents have price utilities tipping the wrong way, then increases in price get more and more revenue from these folks.  If the product you are simulating has relatively low share of preference, this can especially be a problem.

Doing some data cleaning and/or constraining the utilities to follow monotonically decreasing values can help and bring things into more realistic focus.

Comments

To elaborate further, since some people get this wrong, make sure that your price sensitivity simulations involve assigning your client's product to just one price point, then simulating vs. the relevant competition & the None.  Then, change your client's price point (holding all else constant), record the new share of preference, etc.