counter-intuitive results in pricing analysis

Question

Hi there,

I am looking to estimate the price premium that can be charged for a clients new product vs. an existing product they already have on the market.

To do this I have conducted sensitivity simulations using randomised first choice method, varying price of one product at a time while keeping all other products prices constant at the market average, to generate a price demand curve for each brand.

What I wanted to do was compare the curves for the two different products and observe the price point at which the new product obtains the same share % as the existing product does at the actual price it is sold at in the market.

However, the results I am getting do not allow me to do this as the model suggests share % is so high for the new product and so low for the existing product, that there is no overlap between the two brands curves - i.e. even if existing product was sold at a very low price and new product at a very (unrealistically) high price, the simulator still makes out the new product would have higher share. (also, just to note that even before entering the new product into the simulation, the existing product receives very low share which is not reflective of its performance in the market).

I think this problem may be down to the fact that all existing products in the market are very similar, with little to differentiate them other than brand, and the new product has quite a few innovative features that are different to everything in the market at present. While it is sensible that people would pay more for the new product, I think the extent of the premium is being exaggerated by the model.

I have looked into a few solutions for this. Firstly, changing the sim method to shares of preference which doesn't seem to have much of an effect. Secondly, I have looked at changing the exponent from the default value of one to flatten shares. This does help, but I feel its a little arbitrary... Finally, I wonder if because the new product is so different to the others on the market it might actually make more sense to do a purchase likelihood simulation rather than put it in a competitive scenario, and try to visualise the pricing threshold that way.

Any advice on this issue would be much appreciated. Has anyone experienced a similar problem and how did you overcome it? Based on the scenario I have explained, what in your view is better - to change the exponent and keep simulation within a competitive scenario - or to use the purchase likelihood method as an alternative due to the new products lack of similarity to existing competitors in the product category?

Thanks!

Answer 1

Sounds like a challenging problem.  As it is well known, conjoint analysis captures preference from people that in some cases tracks very closely with market shares and sometimes does not.  A product concept could capture a lot of preference (utility) for the marketplace, but unless it has good awareness, distribution, marketing, sales force, communication of the benefits to consumers, etc. it may never achieve the potential that the conjoint market simulator suggests it should relative to other products in the marketplace.

One challenge you are facing is that the existing product has such low share of preference in the market simulator.  You'll need to start thinking about how well the attributes and levels used in the conjoint analysis study capture what is presented to buyers in the real world.  You'll also want to think about your sampling plan (did you interview the right people who are the same people who purchase the product in the real world).  You'll want to think about respondent incentives to complete the interview realistically.

Also, one thing that often happens with conjoint analysis and new product simulations is that the respondents in the conjoint survey are over-educated about the new features (primed) and the new features really jump out in the conjoint profiles--more salient than they eventually will be in the real world in packaging, advertising, marketing communications, etc.  That can cause the market simulator to over-value the new features and overpredict share for the new product rendition.

Comments

Hi Bryan

Just would like to ask a related question to this. I’m doing feature pricing using simulation method. However there are a few instances where the preference share of the new product intersects with the share of the existing product at 2 points.
This is not usually what happens in our simulations. Usually we only have 1.  So we’re wondering if you’ve encountered something similar to this situation. I think it’s not making sense to have two price points (naturally I would choose the more expensive one).

 Thanks