Utility-based recommendation systems

Decisions are hard, and with access to more and more options, we as customers encounter our limitations when it comes to making purchase choices. Luckily, companies provide us with recommendation systems that help the customer choose. They suggest similar products, link it to previous behaviour or show us what other customers have chosen. But recommendation systems do not only help the customers.  With the collection of all this, mostly personalized, data, it could also provide the company with more detailed information about their customers. This information can be used to come up with personal offers or adjustment of the prices based on the popularity of products.

Scholz and colleagues researched this possibility by testing if utility-based recommendation systems are able to measure a customers’ willingness to pay (WTP) for a product (Scholz et al. 2015). This post will discuss their paper and describe their research briefly.

Currently, there are decision models available that measure customer preferences and WTP, however they are costly and ask a lot of effort from customers on specifying their preferences (Scholz et al. 2015). But utility-based recommendation systems are argued to solve both these problems, because customers are willing and motivated to use recommendation systems while these also provide information to the company. This makes them more reliable, low in effort for customers and less costly for the companies.

For this study, they used utility-based recommendation systems. This type of system requires active input from the customer about utility preferences for a single attribute (SAU) and is able to compute the customers’ utility for all the attributes into a total utility factor, which is called the multiple-attribute utility (MAU). The factor can either be linear or non-linear, both are tested in this study. The use of this system is ambiguous, because the researchers stated that the effort of a customer needs to be low. However, the paper argues that a lot of this active input that was asked from the respondents, can be derived from online sources in real life.

The WTP variable is measured on product level. The respondents were asked to fill out the indifference WTP for a specific product. This WTP was split up in different levels:

  1. Floor WTP: this is the highest price at which the customer would definitely buy the product (100% purchase probability).
  2. Indifference WTP: This point is the indifferent point between buying or not buying the product (50% purchase probability).
  3. Ceiling WTP: The highest price of all, but the lowest price where customer would definitely not buy the product (0% purchase probability).

The WTP level of the different products was then calculated based on the utility threshold of the products. This was calculated with the different SAU functions and the indifference WTP that was answered by respondents.

In the study, the respondents were told about digital cameras. They had to perform different tasks. First, they were explained what the different attributes added to the product and after that they were asked to add weights to the different attributes, based on their preferences. This resolved in recommendations that were shown to the respondents in the next task. The respondents were asked what they were willing to pay for the camera that was recommended (WTP for best recommendation). The last task that they performed was to also state their WTP for the top ten recommended products.

The results show that the relationship between attributes and utilities is perceived as an exponential function. This indicates that customers think in trade-offs when they evaluate attributes. For example the attribute size, it has a trade-off of the utilities fragility and the ease of transportation (Scholz et al. 2015). The big difference with WTP is that this was better measured in the linear function. But both treatments underestimated the real price. The interesting part is that based on the WTP that was stated, there was still a rise in revenues.

This is an important implication, based on this study. It shows that with the help of recommendation systems, it is possible to generate personal offers to customers based on their willingness to pay for a product that suits their total utility the best. Recommendation systems are able to gather information on the preferences of customers on a large scale, while it does not request much effort from the customers. Companies are able to, and should, use this information to generate more revenue. They are able to do so with special offers for customers, based on their preferences or to by selling their information to the product manufacturers so that they can design products based on the preferences of customers.


Jack (342660)


Scholz, M. et al., 2015. Measuring consumers’ willingness to pay with utility-based recommendation systems. Decision Support Systems, 72, pp.60–71. Available at: http://www.sciencedirect.com/science/article/pii/S0167923615000263 [Accessed February 6, 2016].

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s