Criterion - ProRank
ProRank - Software for multi-criteria evaluation and decision support

One consequence of our information society is an enormous increase in available data, with many people attempting to glean as much information as possible from these data, especially for comparative evaluations and related decision processes. Various tools are available to support these activities. However, a common difficulty with the evaluation step is that many of the methods mask and aggregate the data, and therefore both valuable information and transparency are lost. This translates to potential impacts on acceptance of the final decision.

ProRank presents a rather new approach based on partial ordered sets that can be used to avoid the merging of data and thus preserve important elements of the evaluation and decision-making processes.

Information and partial order ranking

What are the comparatively best

  • alternatives
  • strategies
  • products
  • materials
  • objects
  • etc.

What alternative meets the criteria?

What criteria are most important/sensitive for evaluation/decision?

This is only a selection of information that ProRank can generate from your data.

Graphical illustrations facilitate the visual analysis of results as well as the handling for presentation purposes.

Beyond the partial order ranking data matrices can be analysed and filtered by means of Boolean and logical operators. Therefore the program has data mining qualities too. A component manager enables an easy navigation between the different program windows, which is in particular important when several data matrices are to be considered at the same time.

For decision purposes in particular, partial orders can be transformed into a linear and total order respectively, by applying the concept of average rank probability.

Features of ProRank
  • Data evaluation and analysis with partial order ranking
  • Visualisation of results by Hasse diagrams
  • Analysis of ranking results:
    • Discovering products, objects, etc. with similar qualities concerning all criteria/attributes
    • Discovering alternatives, materials, etc. that meet certain criteria values (e.g. substitute materials, keeping threshold values)
    • Classifying environmental pollution, products, chemicals, etc.
    • Importance of criteria/atriibutes

    Objects with similar qualities in all criteria→

  • For decision purposes transforming a partial order into a linaer/total order by the average rank concept
  • Context-specific selection of objects and/or criteria/attributes from the data matrix by Boolean arithmetic and logical operators (data mining)
  • top
    • Several algorithm for criteria/attribute aggregation
    • Basic statistics for yout data matrix
    • Easy navigation between program windows
    • Import and export of MS-Excel sheets and text-files
    • Export of diagrams supporting several graphic formats
    • Fully platform-independent, i.e. ProRank runs on Windows®, MacOs® and Linux
    What is partial order ranking?
    Multi-criteria evaluation can be described as ranking of objects (representative for all items of evaluation) by means of more than one criterion.
    Object Criterion #1 - consumption, l/km
    Boran 7
    GW 8.8
    Relant 7.6
    Flint 9.2
    Rami 8.0
    Assuming that lower values are better than higher values a ranking of objects is very easy:

    Ranking by one criterion = total order

    Multi-criteria evaluation by partial order ranking
    When the cars are to be evaluated by two or more criteria then an aggregation to one criterion might be crucial because of different dimensions of both criteria, i.e. fuel consumption in liter/km and car boot in volume (liter).

    In partial order ranking both criteria are taken into account at the same time. That means: a car is "better" than another car, if it has in both criteria lower values. Since a larger volume of a car boot is better than a smaller one, the sign of all criterion values has to be reversed.

    In the visualisation of the partial order a line indicates the comparability of cars: for instance, it is seen that Relant is "comparable better" than GW, Rami and Flint, because it has lower values in both criteria. The absence of a line indicates incomparability between the cars because their criteria values are antagonistic: Boran is incomparable with Relant because Boran has a lower fuel consumption but a lower car boot volume too (here a higher value). The result of partial ordering can be interpreted as follows:

    There are two "best" cars compared to all others, namely Boran and Relant. However, the reasons for that are different: Boran has a lower fuel consumption as Relant whereas Relant is characterised by a higher car boot volume in comparison to Boran. Therefore, by partial ordering it is seen 'why an alternative is incomparable with another or better than another'. Since aggregation of criteria is avoided and moreover both criteria are processed as their are, the main advantages of partial order ranking can be characterised as:

    • no information loss
    • transparency

    Decision: Transforming a Partial Order Into a Total Order

    After obtaining two "best" cars with respect to fuel consumption and car boot volume, one may decide which one to prefer. For these purposes ProRank features the average rank concept. By decomposing the partial order in all kinds of linear orders an average rank (Rkav) of each car can be calculated, where no criterion/attribute is prefered.

    Whereas in the partial order both Boran and Relant were evaluated as the best alternatives compared to all other cars, the average rank yields a distinct order with Relant as the best alternative.


    System Requirements

    We recommend running the operating system on a machine with 64 MB or more memory, 20 MB Free of Hard Disk Space and SVGA Adapter and Monitor.

    Important Installation Notes

    ProRank requires Java Runtime Environment (JRE, at least version 1.4.2) that is usually installed on Mac and Linux machines. For Windows user we recommend to make sure that JRE is installed. If required you can download the JRE from

    For a full version of ProRank please contact us.