Have an idea?

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

Measuring willigness to pay - whats the best way


i try to measure the willingness to pay for a product. I already collect the data, so i can not change the attributes.

My problem is, that i collected two price attributes one which have to paid monthly and another one which is one-time. Also there are 3 other attributes which are features of the product.

There are many calculations examples where the standardization of the attributes levels in price units is explained. I have also read the suggestions from Bryan which are on the website about this topic. But i dont know if its even possible to do a standardization with two price attributes.

So in short: it is possible to measure the willingness to pay with two price attributes? Or can i just use the market simulator to give an advise which product will be prefered by the most?

The next problem is that i want to do a segmentation with hb (i collected some segmentation data) to test if the willigness to pay is different (higher) for the segments. Im not really sure how to test this without having a average price for the segments. Is there any advice?
asked Aug 24, 2018 by user384 (190 points)

1 Answer

0 votes
You have two price variables, so I think the best way to calculate WTP would be to fix one and estimate WTP using the other.  So for example if you have levels A, B and C for your one-time fee and 1, 2 and 3 for monthly fee, you could figure out the monthly WTP given one-time fee A, then given B then given C.  Or you could fix the monthly and figure out the one-time WTP given monthly fee 1, and again with monthly fee 2 and monthly fee 3.
answered Aug 24, 2018 by Keith Chrzan Platinum Sawtooth Software, Inc. (85,300 points)
Before asking why my WTP calculations are so implausible...

I coded my attributes for HB linear:
1) the monthly costs are (1,2,3,4,5) which I coded in (2,1,0,-1,-2)
2) the one-time is (0,100,200,300,400) which i coded also (2,1,0,-1,-2)
3) other attributes was include/not include which i coded in (1,-1)
4) there is one attribute which is feature a, feature b, which I coded (-1,1)

I was not able to figure out if i should always use the lowest effect (-1) for the worst feature (not include or highest price). In my opinion (and this is also pointed out in the manual) the coding affect the results.

I run the market simulator(which is included in lighthouse) with continuous/price on monthly/one-time with the same values as i coded in the HB estimation. When I put 2 products in the simulator which only differs in monthly price, the shares for the higher price are not zero. This makes absolutely no sense to me from a rational thinking. Is that a result from wrong settings or does  some respondents prefer a higher price over a lower?

Another problem arises from the coding of the attributes. When I use the dropdown menu to select the attributes for the market simulator there are the labels and the values. When I select a level, the software puts the value in the field. For one attribute it enters a -1 and shows an error "no viable alternative at input '--'". I am only able to select the level with the value of 1. What does that mean?

Other questions about this:
1) does the internal labels does have any effect on this? I labeled the attribute levels for include with "1" and for not included with "0".  

2) does the ordering of the levels in the CBC-Task settings have any effect on this?
one attribute is something like ... 0, 50%, 100% which i coded(values) in (-1,0,1) BUT the order in the task menu was 50%, 100%, 0. This ordering is also in the HB-Settings. I give 50% the value of 0, 100% the value of 1 and 0 the value of -1. The settings are accepted, no errors. But when I put these values in the market simulator on attribute info it shows me "level values must be in ascending or descending order".

I read the manual when I design the study and i thought that the order or the labels have no effect on anything. Is it possible to change the labels or the order after I collect the data ? I am a bit aware of doing that.

This all confusing me and I am not sure if that is the reason why my calculations are really implausible.
Ok, I think i figured out what is wrong...

When I enter zero centered values in HB the market simulator is not able to function right. When I linear code the attributes with positiv values (1,2,3) and let lighthouse zero center them internally to (-1,0,1) the market simulator seems to get the same results for part-worth / linear coding...

When I only use positive values and let lighthouse doing his zero center thing... i only get nearly the same share percentages as I get in part-worth coding(but they differ from 1-2%).

The dropdown menu in the market simulator enters values , not labels. This is really confusing for me, because I only get the same shares when I enter labels. I can not get the order working. In HB I can enter values like I want to do this, but when I try to enter them in market simulator (attribute info), it gives me the error that they have to be in ascending or descending order. Also confusing the market simulator takes the right values from HB settings, but gives an error by just click on it.

is this a bug? the manual says that I should value them zero centered when linear coded and there is a example of this (page 31). What are the right numbers to put in the market simulator? values? labels? does that come from entering numbers as labels?

And the most important question: what are the right share preferences? I get really different results and coefficients in HB too.
Given your questions, I think you'd benefit most if you contacted support@sawtoothsoftware.com
Value of none option in WTP calculations