Why a Group Recommendation Never Fits EveryoneWhy a Group Recommendation Never Fits Everyone

In the Malang souvenir recommendation study, the group ranking tracked three of four users closely: 0.97, 0.99, and 0.88. The fourth sat at 0.69. Averaged out, the group result correlated 0.88 with individual preferences. That number sounds strong until it is read as what it is, an average that hides a member who was not well served.

Each traveller sets their own criteria weights before the ranking runs. In the worked example one user treated distance, price, and shop age as equally important, while the other leaned heavily on distance and price. Those two weightings pull the same list of options in different directions, and the Borda count is what settles the argument.

• Let every member set weights separately. A single shared weighting hides whose preference is being ignored.
• Show the reasoning behind a top result. Users accept a compromise more readily when they can see what drove it.
• Expect one member to be least aligned, and consider a second option that suits them.

The demonstration used fifteen alternatives, three criteria, and a small set of decision makers in one Indonesian city. Correlation testing shows that two rankings move together, not that users were happy with the outcome. Satisfaction was not measured here.

Even with those limits, the study supports a useful design habit: treat a group ranking as a starting point for discussion rather than a verdict, and give every member a way to see and adjust their own weights.