Have an idea?

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

Calculating Willingness to Pay ACBC

Hi all,

I have run two ACBC models with pricing in each. I am aware of the risks with Willingness to Pay (WTP) but the client is desperate to understand to what extent every options is valued, ideally in a monetary format.

I have calculated WTP by running simulations comparing two products against each other, only changing the attribute I want to get an estimate for (in relation to the other attribute).

I have run into two problems:
a) I have used the Calibration module to set the none threshold at 4 (probably/definitely will buy). I find when I run WTP simulations on the uncalibrated data, it doesn't matter what I set the other attributes to (so long as they are consistent between the two products), but when I use the calibrated data the choice of the other attributes has an impact. Which approach should I be using, and if I should be using the calibrated data what should I set the attributes to?
b) Some of the values seem very high - too high. Is there a way of showing this data that shows the comparative value of the attributes without actually allocating prices to them? One approach I tried was to zero-centre the pricing for each attribute, but then I worry the client will think the less valued attributes have to be associated with a discount, which they don't necessarily.

Any advice would be most welcome.
asked Oct 22, 2019 by Sharon Morris (180 points)

1 Answer

0 votes
To clarify for others that might be reading this post, the calibration step in ACBC only affects the scaling of the None utility; it does not modify the HB estimates of the attribute levels included in the experiment.  

I gather you are using the simulation method for estimating WTP, based on finding the difference in price that causes the share for two products (one with enhanced feature and an otherwise identically defined product without it) to be 50/50.  Because a constant utility added to alternatives in a logit simulation factors out, it wouldn't matter what other attribute levels the two products were given, as long that they are equally given to both products.

However, when a third (or additional) alternatives are specified in the competitive scenario (such as the None alternative), then the WTP will change.

In general, we think that WTP estimates (from the two-product simulation scenario approach) are exaggerated and tend to represent the theoretical maximum the market would pay if there weren't competitive offerings also providing those enhanced features or other features that would otherwise compensate.

So, to obtain more realistic estimates of WTP, a suggested approach is to try to represent as completely as possible the full set of competitive alternatives in the marketplace, which includes the ability to opt out (the None choice).  So, let's assume it is believed that the market is adequately represented by 12 products specified in a competitive market simulation scenario (the highest market share options represented at their respective brands, features, and prices) along with a 13th option being the None.   One of those products becomes the test product for which WTP for enhanced features will be estimated.  We first simulate that test product versus the other 12 alternatives at the base price and base feature for the test product.  Next, we modify that product by enhancing a given feature and simultaneously adjust the price upward until its share is driven back to the base case value.  Note we are not adding a new enhanced product alternative to the simulation scenario; we are modifying the test product to improve it while holding the number of product alternatives constant.
answered Oct 22, 2019 by Bryan Orme Platinum Sawtooth Software, Inc. (170,215 points)
Thanks Bryan, that's helpful.

Unfortunately it wouldn't be possible to try to replicate a competitive set - there are an estimated 3,500 competitor products available in the market.  And realistically, everything we tested is probably already offered in the market. The client just wants to get a feel for the perceived consumer value of the different levels.
OK, but understand that if you only use two products in the market simulation (product without the enhanced feature and one with it), without giving respondents any other option in the simulated marketplace to get enhanced features via any other competitive offering, it will overstate the WTP.  In reality, respondents have other alternatives in the world other than picking between two products.   This approach should bring WTP to a reduced and more realistic level compared to the two-product simulation approach that ignores the None and available competitive offerings.

I would recommend creating a sample of the available and largest competitors in the market and representing them in the market simulator.    This sampling should include products offering the variety of product capabilities and prices in the marketplace.  For example, if you represented the marketplace by creating a sample of 18 or 24 competitive products in the marketplace.  Plus, include the None alternative, for a total of 19 to 25 competitors.
...