All the analytics (computation) is performed at the edge in the camera. The metadata (results) are stored in the cloud. This hybrid approach means we can offer cool functionality like motion search, heat maps, and motion based retention with low bandwidth usage. This allows for deployments on small WAN links like cellular.
When you create a motion search, dashboard is querying the cloud's store of metadata, the table that is created at the bottom of the screen is derived from this. When you click an entry it then pulls the associated video from the camera and streams it to you.
What this means is the amount of up-link WAN bandwidth needed by a camera is variable dependent on the amount of metadata. The maximum it will be (design limit) is 50Kbps. We expect a typical camera to average about 11Kbps, and a camera that sees very little activity to average <3Kbps.
@Jsnyder81 if this helps, and to add to the reply from @GeorgeB see this screenshot of the network usage from the MV71 camera over my driveway that also covers the residential street in front of my home. So there is a fairly regular amount of motion and analytics info, like vehicles moving by, people walking, and I can even pick out the windy versus still days by how much the trees/leaves sway back and forth, so there's a reasonable amount of metadata being sent to the cloud. As you can see, my MV71 has generated a total of about 16MB, and averaged under 2Kbps over the last 24 hours. See screenshot.MV71 Metadata Usage
@Jsnyder81 Just to add the opposite of @MerakiDave's high traffic camera, here is the usage chart from an MV21 that's in a server room where there is typically no motion all day. You'll see the network usage is almost identical, so I think it's safe to plan for ~2Kbps of throughput and ~15-20MB of traffic each day per-camera.