Have an idea?

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

Warning-Simulation contains an extrapolated value


I was trying to do a willingness to pay using the simulator by raising the price of a product until its share of preference drops to the original value. However, after a specific price (the max price listed on my study), I get a warning saying that the simulation contains an extrapolated value, and the simulator will use straight-line projection to estimate level utility. And probably because of this, the share does not drop to the original value regarding of whatever the price increase.

Can I only use the prices specified for that particular attribute level? The pricing attribute had 3 levels (Low, Medium, and High), and I used conditional pricing with respect to another attribute with 2 levels. So in total, there were 6 prices.
asked Dec 3, 2017 by menon (240 points)

1 Answer

0 votes
Change from using Randomized First Choice to "Share of Preference" and you will be able to extrapolate.  The software will still give you a warning, but it will perform the simulation.

But, extrapolating takes you into a dangerous area, since you are making assumptions about the sensitivity to price in an area that was never asked of respondents.  The software will assume a straightline projection on price using the two most extreme price levels included in your study.
answered Dec 4, 2017 by Bryan Orme Platinum Sawtooth Software, Inc. (136,265 points)
Thank you Bryan. I am using Share of Preference, but I do not want to extrapolate. While trying to figure out the willingness to pay, the software extrapolates automatically. I would like to know if its possible go get the willingness to pay without extrapolating. As I mentioned, the shares of preference does not fall back to the previous one before the start of price increase.
Yes, this is one of the challenges of WTP, which is a tricky and messy area: the WTP sometimes takes you outside the range of prices that were even asked of respondents.

The way you are trying to do things is to compare the original product (without the enhancement) to the same product with the enhancement, but at a higher price until the shares of preference from the simulator become 50/50.  I'd recommend setting the original product without the enhancement about at the price you'd expect it to sell for in the real world.  That sets the context in the approximate price range that is relevant to the business question.

However, if doing the above causes the enhanced product to have to be raised outside the range included in your experiment, then this is a problem.  One possibility is just to annotate it in your report to indicate that the WTP measured was extrapolated, and should be interpreted with caution.  Another approach, which also is shady, is to lower the price of the original product without the enhancement so that the simulation vs. the enhanced product achieves 50/50, but stays within the price range you measured.

WTP is a tricky area, and the way you are doing it represents the theoretical maximum that a person would be willing to pay if there were no substitutes or other competition available.  As soon as a competitor came in offering a lower price for the enhanced feature, people would no longer be willing to pay that amount.