Our first experiments were simple 2-way calls using the three most popular video conferencing applications: Zoom, Google Meet, and Microsoft Teams. Like most online apps, these programs both pull data from the Internet (download) and send data to the Internet (upload). Maintaining an unrestricted flow of these data downloads and uploads produces a high-quality experience for the user.

On average, these applications used about 1-2 megabits per second (Mbps) of download bandwidth and 0.75-1.4 Mbps of upload bandwidth for a single video call, with Microsoft Teams on the higher end of both measures. Most residential Internet plans provide plenty of capacity to handle this data; the FCC defines broadband as a 25 Mbps download and 3 Mbps upload, both much higher than the average used in a video call.

Competition

But imagine a house where parents are working remotely and participating in video meetings while their children attend remote school through their own video conferencing apps, or use a video streaming service for entertainment. How would the data needs of these applications stack up and affect each other?

While most available Internet plans provide download speeds that can handle more than one video call at a time, the upload limits can come into play with only 2 simultaneous Teams meetings or 3-4 concurrent Zoom or Meet sessions. If the applications are receiving only a portion of their normal upload speed, it could mean sending choppy video to other participants on the call.

What’s more, these applications sometimes do not play fair with each other, competing for that limited bandwidth. In our experiments with Zoom and Teams, the first call receives precedence over a second call, capping the upload speed available to the latter application and lowering the quality of the call for the second user. For homes without Internet plans that exceed the minimum FCC broadband standard of 3 Mbps upload capacity, this effect can easily produce headaches during simultaneous meetings.

Meeting Size

In remote work or school environments, video meetings are rarely one-on-one. Our experiments found that, in certain situations, the number of users in a meeting and how they are watching your video can affect the amount of bandwidth used. For example, a teacher lecturing from home to a virtual classroom is likely watched by her students in “speaker mode,” where the video is enlarged on each screen. We found that Microsoft Teams meetings used more upload data on the teacher’s network when more participants were watching and using speaker mode, testing the 3Mbps limits of a minimum broadband connection.

Interruptions & Recovery

Any of the above situations, as well as interruptions to service caused by a drop in WiFi signal or a brief network outage, can cause one of the most annoying video call glitches: frozen video. Our researchers simulated these disruptions by reducing available bandwidth for 30 seconds and seeing how long it takes video conferencing apps to recover. Depending on the magnitude of the bandwidth drop, the apps could take 25 to 50 seconds to return to their pre-interruption performance — a significant and inconvenient gap in anyone’s meeting or online class.

Because these were laboratory experiments, we could induce artificial and controlled interruptions to service. But in your home, these disruptions could be longer or more frequent, and the maximum bandwidth available may be lower in areas with poor WiFi coverage. Therefore, these experiments may actually underestimate the amount of Internet needed for smooth performance and the effect of service interruptions on the users’ quality of experience.

Policy Implications

These results help establish a more realistic, household-level benchmark for the amount of Internet bandwidth required for remote work and school. Rather than classifying households by whether they meet the FCC definition of broadband access, a city government might create a new standard around these modern technologies. For instance, does a residence have enough upload and download bandwidth for three kids to do remote school simultaneously with sufficient quality?

Policymakers could use this benchmark in future times when remote schooling is required, guiding the distribution of higher-capacity broadband access and devices to ensure that it is meeting realistic bandwidth requirements. In addition, consumers could use this information to purchase an Internet plan to meet their needs, whether that means multiple people video-conferencing at the same time or just a single person streaming videos. The information can also help us assess our resident-level data on the true Internet speeds that people receive in their homes and whether they are sufficient for frequently-used web applications.