A pairwise ranking survey is a machine optimized process for ranking a large number of items smartly without overloading the respondents cognitively. Typical ranking models where users are asked to rank items (in order of preference) does not folllow human cognitive process. We are not programmed internally to ranking order our preference of wine! We can however discriminate clearly. Given a pair of wine glasses, we can easily say that we like one over the other.
The pairwise ranking survey process uses a machine optimized process to display items two at a time. The respondents have to pick one of the two items. Using a dynamic lookup model, the pairwise ranking process then optimized for orthogonality first. This means that - all the items are randomly divided into groups of 2 each and presented to the respondent.
After that, for the items that are selected - are again recursively grouped two at a time - that is again randomized until the final item is reached.
This then deterministically defines the best option and a tree is created. Once the tree is created, the system can then rank order all the items based on the respondents input.
This model allows for a simple and effective way - that is mobile friendly, where users are used to swiping left and right - to determine the efficacy of an item - and can rank order respondents preference without resorting to a complex cognitive load.
Our research has also shown that the gamification and anticipation of the options allows for extremely high response and participation rates.
