Sprint 3 – Validate - AgileBusinessAnalysis/01_TEAM GitHub Wiki

Recommendation concept

The recommendation concept validation – based on the prototype specified in Sprint 3 – Prototype – provided the team with the following insights.

Validation results

A total of 24 participants took part in the validation. The recommendation concept achieved an overall accuracy of 54%.

Tipp A performed not that good. The tipp was recommended 8 times and was right in 50% of the cases. In all negative cases, the right tipp would have been Tipp B (!).

Tipp B performed best – achieving an accuracy of 70% (!). In negative cases, 2 times Tipp A and 1 time Tipp C was suggested.

Tipp C performed worst – achieving an accuracy of only 33% (the equivalency of the given probability of having 3 options). In 75% of negative cases, Tipp B would have been the right one.

The above mentioned findings can bee seen visualised by the pie chart. Tipp B would have been the best alternative in negative recommendation cases. Assumption: Tipp B is too strong for the other tipps (it was already chosen most often by the recommendation engine.

A correlation was identified. In case of a negative recommendation result: If the participants chose answer B in the first question and whatever answer in the third question, the choice of the second question defines what recommendation would be right.

The tipp suggestion provided in a clear result. The recommendation "shopping" achieved almost tripple as much votes as the second most often chosen tipp alternative.

These validation results are taken into account for the following derived improvements.

Derived improvements

Based on the identified correlation above, the following adaptions were made on the recommendation logic.

The marked rows were adapted in the way that: If answer 1 is B, then answer to question B is equivalent to the given Tipp. The initial recommendation logic can be seen here: Sprint 3 – Prototype.