Non-linear performance pricing NLPP

How the automatic price-performance evaluation works

| by

Mathematics is incorruptible, so do we: We use multiple regression analyzes to determine thanks to a Special software, which products promise a particularly good price-performance ratio based on their product data and their price. We explain how it works.

No question: it is a silver bullet for evaluating the price-performance ratio of a product, and we are already going in many product categories: We test the relevant devices in a product category, determine an overall rating, relate this to the current device prices and thus receive a ranking that it is CHIP Enables users to find products that promise particularly good performance for their price. However, this procedure has a very important limitation: It is very complex.

We are currently testing a good 1 in the CHIP test center. 000 products per year. That is enough for many leaderboards, which hopefully provide our users with valuable information. For example, to fill our notebook leaderboards, go through well each year 100 Devices the test center. In the same period, however, around 50 times as many notebook configurations come onto the market. This makes it impossible to really test all products. We would also like to be able to present our readers with a selection of exciting products in categories for which we have not developed a test procedure. And this is where NLPP comes in.

NLPP calculates price-performance formulas

NLPP stands for “Non-linear Performance Pricing”: This is what the Swiss software manufacturer calls “ Saphirion AG “Its software for the analysis of product and price data. It is used primarily in the industrial environment in the area of ​​purchasing and is intended to allow companies to buy components or products as inexpensively as possible. After we became aware of NLPP, we tested the software and now use it for the price-performance evaluation of consumer products.

We use NLPP as follows: We get the relevant product data including the prices of products in an interesting category. In our case, this data comes from online shops such as Cyberport or from the manufacturers themselves. We clean and analyze them and, if necessary, we add additional information, such as a performance index for the notebook CPUs. We import this refined data into the NLPP software and then inform it of the so-called “price drivers”. This is data that has an impact on the price. Two simple examples from the field of notebooks: Whether an inexpensive Atom processor or a powerful Core i7 CPU was installed usually plays an important role in the sales price. Likewise, whether, and if so, which operating system is included.

Based on this data, the software determines a formula for the so-called target price, the results of which are as close as possible to the real prices. In order to minimize this difference across all products fed into the system, different mathematical methods are used, so-called regression analyzes. The NLPP software uses six different methods and ultimately chooses the one that shows the best fit between the actual and target price.

Then the NLPP software enters each device in a coordinate system, the vertical axis of which shows the real price, the horizontal axis represents the target price. As an illustration, you can look at the diagram that can be found below in the article: It shows the result that the software delivers when we compare it with the product data of just under 500 Notebooks up to 2. 000 Euro feed. Each point stands for a specific notebook model. Devices that lie directly on the straight line of origin cost exactly as much as the software considers appropriate based on the performance parameters we have taken into account. Notebooks below are cheaper and therefore offer an exciting price-performance ratio. Devices that lie above it cost more than calculated using the price formula and are therefore potentially too expensive for the performance offered.

Results are only as good as the data

Basically, the software provides mathematically “correct” results that we believe have proven themselves very well in practice. However, you should keep the following in mind when evaluating the results: The software can only take into account information that is explicitly contained in the product data. For example, it is possible to differentiate whether a smartphone camera has photos with a maximum of 8, 12 or 20 Megapixel shoots. However, the software cannot know how good this photo really is without qualitative information (such as the results of camera measurements).

NLPP cannot and should not replace tests and expert assessments for us, but only supplement them. We use the tool to filter out and present to you a large number of products that appear particularly interesting from a price-performance perspective. Not more but also not less.

Related Articles