Pairwise Analysis

The pair-wise analysis is a simple tool to rank competing requirements. Each requirement is tabulated for importance using a binary comparison. This will help your understanding of which requirements are more important than others.

Example applications

The pairwise technique is a useful technique for ranking all sorts of things, such as applicants for a job or entires in a design competition. Pairwise approaches are used to in probabilistic models to aid decision-making, and can be used to aggregate a range of opinions.

Customer requirements v Design requirements

It’s important to note in the requirements phase the difference between customer and design requirements. Customer requirements tend to be ‘wants’, ‘needs’, or ‘desires’ from the client. Design requirements tend to be ‘whats’, or the measurable aspects of the design. At this stage, we are concerned with prioritising the customer’s desires, and so will use the customer’s perspective here.

Steps

1. Construct a list of potential customer requirements (for this exercise, less than 5 becomes trivial, and more than 10 becomes unmanageable) to bring into the activity.
2. Create a table with the customer requirements listed in the rows and columns in the same order Create a table with the customer requirements listed in the rows and columns in the same order
3. Black out the diagonal.
4. For each row, compare the requirement in the row to the requirement in the column.
   • award a 1 if the row requirement is more important than the column requirement
   • award a 0 if the row requirement is less important than the column requirement
   • in this process, rows and columns cannot be equal
5. Tally up the rows and calculate the sum. The most important customer requirement will have the highest sum.
6. If you have equal rankings, split them by comparing them to each other

Hints

• If you have equal ratings, make sure you split them - the purpose of the tool is to rank requirements
• Complete your pairwise individually, and combine them using the Borda count technique in Dym reading on rank ordering

Core resources

• Dym, C.L., P. Little, E.J. Orwin and R.E. Spjut, 2009. Engineering Design - A Project-Based Introduction, Third ed., John Wiley & Sons. pages 60-62 provide an example of a pairwise using a ladder as an example [PDF, 3 pages]
• Dym, C.L., et al. Rank ordering engineering designs in Research in Engineering Design November 2002, Volume 13, Issue 4, pp 236-242 [PDF, 7 pages] Note that this also introduces a more complicated idea of group aggregation