Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

WTP from ACBC (again)

Dear Sawtooth-team,

I've read all the available information on WTP (forum and white paper) and do not really see a solution with regards to my data. A sensitivity analysis does not work, as I cannot replicate the whole market with data, i.e. the shares of preferences are unrealistic and fall short of a lot of alternatives. Even if I tried to replicate the market, the theoretical combinations of attributes (8) and levels (24) would create a matrix way too big. My idea now is, to compare purchase-likelihood of two sample products that differ in one level only and vary the price of the "better" product until purchase likelihood is equal. Wouldnt that represent a WTP for a particular level given all others are equal?  It is not under your recommended solutions, so I was wondering if its a realistic approach.

The only problem I encountered with this approach but also with a share of preference approach is that price changes have minor impacts on either of the two approaches, i.e. is completely unrealistic. E.g. a change of a level to a more preferred level of minor impact (partworth +5, while none alternative 200), gives a 10 dollar WTP at a particular point (bases on share of preferene or purchase likelihood as discussed before). This is for a product that ranges between 5 and 70 euros in my conjoint highly unrealistic. Any ideas?
asked Nov 26, 2017 by Luke (280 points)

1 Answer

0 votes

The approach you describe (looking for a share-equalizing price between  versions of a product that differ in their levels for only a single attribute) is one way of measuring WTP, in fact it's a measure we discuss in our book Becoming an Expert in Conjoint Analysis (http://www.sawtoothsoftware.com/81-products/conjoint-analysis-software/1852-new-advanced-conjoint-analysis-book).

The problem you describe, unrealistically large WTP, is a common one, also discussed in the book.  It's been suggested that these kinds of WTP calculations may measure what economists call the "maximum price," in other words the maximum price a customer would be willing to pay if the feature whose WTP you're measuring was not available in any other way, not from any other brand for any lower price.  

Note that the academic literature is not replete with stories of measurement of WTP so successful as to satisfy all critics:  WTP is a controversial metric and research about how to best design and analyze choice experiments to capture it is ongoing.
answered Nov 26, 2017 by Keith Chrzan Platinum Sawtooth Software, Inc. (53,875 points)

many thanks for your answer! Two comments on this:

a) is the approach of purchase-likelihood equalizing price also a valid approach? As I said, the share-preference appraoch looks a bit flawed in my case as I would rather compare two versions of a product and not replicate the whole market. I therefore thought purchase likelihood - not taking other products into account - would therefore be the better approach

b) if we are talking about maximum WTP, as you descibe, then this approach (as well as share of preference approach) are equally flawed I guess. an Interpretation is therefore rather, that there is an additional WTP, if purchase likelihood and share of preference are higher for the more expensive product, yet questionnable whether in the actual magnitude (e.g. price diff of 10 euros, still preferred)

Yes, the purchase likelihood equalizing method seems reasonable.  

There are some other things you can try, and which I find sometimes produce more reasonable WTP.  For example, I'll compute WTP at the respondent level and report the median rather than mean, as the latter can be seriously affected by outliers.