Our matching algorithms evaluate more than just location to determine if two phones are a possibly match (and highly accurate location is fairly rare). So often we are able to match the correct phones even if there are others in close proximity that are also bumping. In the event that we match >2 phones, we prompt those users to bump again and compare both datasets.

Although you can sometimes trick the algorithm into matching phones that did not physically bump by shaking them from across the room, bumping will not work reliably if you do not bump into each other.

Read the article.

I did read the article. Was looking for real world experimentation or maybe a comment from people at Bump. Bumps don't need to happen phone-to-phone, and geolocation data isn't all that accurate.

I saw multiple Bump interactions in highly dense technophile populations at SXSW where two people got a random contact when they bumped.