When a naming project reaches the point where dozens or even hundreds of candidates sit in a spreadsheet, the immediate problem is not choosing the best name but reducing the pile to a size a group can actually talk through. A practical starting point is to apply hard elimination criteria that require no judgment calls: names that conflict with existing trademarks in your category, names that are impossible to pronounce in your key markets, or names that are already taken as domain registrations or social handles you would need. This pass is mechanical and can be done by one person working from a checklist. Depending on how large and varied the original list is, it may clear out a meaningful number of candidates before any discussion begins.
After the hard filters, the remaining candidates usually still number too many for a productive meeting. A useful next step is to cluster them by strategy or theme. You might find that several candidates are all variations on the same metaphor, or that a group of them are slight spelling tweaks of one root word. Grouping this way lets you keep one or two representatives from each cluster and set the rest aside, not because they are bad but because they do not add a distinct option to the discussion. The goal here is variety of direction, not volume.
With a clustered set of roughly eight to twelve names, you can run a structured evaluation with a small group. Give each reviewer two or three specific dimensions to score on, such as memorability, fit with the project's intended tone, and ease of use in a sentence. Avoid asking people to simply rank their favorites, because that tends to produce a spread of personal taste rather than a comparison you can act on. Collect the scores, look for names that consistently land in the middle or bottom across reviewers, and remove those. What remains is a shortlist small enough for a focused conversation about trade-offs, context, and final preference.