Video Checkup from YouTube gives you an easy way to compare mobile providers based on their ability to stream YouTube videos to your device.
Scores are based on real measurements from billions of YouTube views worldwide, sorted by geography and mobile provider.
The measurements show whether a provider serves mostly standard definition (SD) or high definition (HD) videos in a given area. They also indicate how much buffering a typical viewer can expect while watching YouTube videos using that provider.
Individual results may vary based on device, location, data plan, and other factors. Data throttling and MVNO relationships may affect scoring.
YouTube wants you to be able to watch videos whenever you like at the highest quality possible.
Mobile providers play a crucial role in determining quality, including whether your videos experience annoying lags known as buffering.
By analyzing billions of anonymized watch results each month, YouTube is able to score video quality by provider as well as map areas of higher and lower quality.
The scores represent the video streaming quality you can expect when you watch YouTube using a specific mobile provider in a specific area. These scores are based on the streaming capability of the provider at standard definition (SD) and high definition (HD) resolutions, as measured by YouTube mobile applications in the past 35 days. The scores and associated descriptions should give you a good indication of expected video quality and buffering.
Rather than being based on a small sample of users, this analysis is based on billions of YouTube videos across hundreds of mobile providers.
Data is anonymized and represents aggregated YouTube views using a given mobile provider.
Many factors can independently affect user experience on the Internet, including but not limited to a user’s Internet connection speed, reliability of the Radio Access Network (RAN), policy enforcement and decisions on the core network, availability and load characteristics of the application servers, choice of Internet data plans, and capability of the devices and applications in the network. Individual and isolated measures like access speed test, transport metrics, or server capacity do not capture the real user experience. A better approach is an end-to-end, application-level performance measurement that includes all factors in the equation.
Below is a methodology to rate mobile networks in terms of YouTube video streaming capability based on sustained application-level performance measurements. The objective is to present a score that is meaningful, easy to understand, and closely reflects the real-world experience of YouTube viewers.
A typical YouTube video playback consists of the YouTube client (player) fetching video bytes in a streaming fashion from a YouTube server (CDN) in one or more video requests (e.g. HTTPS GET). The first step in determining provider scores is to measure the sustained speed at which these video bytes are transferred from server to the client. To measure the achieved application-level throughput (goodput), the following are recorded for each network request:
Based on these measurements, the goodput for a given request ‘R’ is computed using the formula below. Each measured request is considered a goodput sample.
GoodputR = Response SizeR / Response TimeR
Goodput is defined as application-level throughput, or provider throughput measured at the application layer of the protocol stack. This is the number of useful information bits delivered by the provider to the client application per unit of time. The amount of data considered excludes protocol overhead bits (e.g. TCP headers) as well as retransmitted data packets in lower layers of the network stack (e.g. RLC, TCP). Since goodput is application specific measurement and excludes overhead bytes, this makes an excellent network quality metric to infer actual user experience in real world mobile applications like YouTube.
YouTube mobile applications request for video and audio in fragments over the course of the video playback, and will try to maintain a healthy readahead to protect from buffer underruns. YouTube mobile applications have the capability to measure goodput on these media streams on both HTTP and HTTPS connections, over both TCP and UDP (QUIC). The measurement system also accounts overlapping request patterns (e.g. audio and video streaming in parallel) and can correctly compute the effective goodput in such cases. Multiple observations are reported for every video playback.
Scores are derived by aggregating relevant goodput observations recorded in the measurements phase. The methodology supports scores to be computed at various levels of granularity for the selected dimensions. For example, the score for a network could be calculated for various time slices (e.g. hour, day, week, month) and/or at various geographical levels (e.g. country, province, region, city).
For a given time period ‘T’ (e.g. trailing 35 days) and geographical location ‘L’ (e.g. San Francisco, CA, USA), the score for a service provider ‘P’ (e.g. Verizon) is computed as follows:
Compute GAT (Goodput Above Threshold) volume: Place each goodput observation in HD or SD buckets based on thresholds from the table below.
An observation greater than 2.5 Mbps is placed in both HD and SD buckets; an observation between 0.7 and 2.5 Mbps is placed only in the SD bucket; and an observation of less than 0.7 Mbps qualifies for neither. This provides the raw aggregated GAT Volume for the selected dimensions.
|GAT Bucket||Goodput Threshold||Reasoning|
|HD||>= 2.5 Mbps||Recommended goodput required to smoothly stream YouTube HD video at 720p resolution|
|SD||>= 0.7Mbps||Recommended goodput required to smoothly stream YouTube SD video playback at 360p resolution|
Compute GAT percentage: Based on the computed GAT volumes (SD and HD) from step 2, compute GAT percentages (SD% and HD%) based on total number of observations from step 1.
|HD capability (HD%)||HD observations / Total observations||% of video streams that achieved network goodput sufficient to smoothly stream YouTube HD playback at 720p resolution|
|SD capability (SD%)||SD observations / Total observations||% of video streams that achieved network goodput sufficient to smoothly stream YouTube SD playback at 360p resolution|
Assign scores: Place the HD% and SD% (from previous step) in the 2D chart below to determine the final scores. The 10 segments in the chart reflect various levels of YouTube experience based on the HD and SD capability delivered by the network goodput.
A provider can achieve higher scores by moving up along either axis, by offering a smooth streaming experience at SD resolution to greater population (higher SD%), or supporting smooth streaming at HD resolution for a greater subset of the population (higher HD%), or both (ideal). The final score for a provider in a given location is determined by placing the computed GAT percentages (SD% and HD%) from step 3 in the chart below. The highest score an operator can currently achieve is 4 stars, 4.5 and 5 are reserved for future use.
Maps provide an easy way to browse and explore points of interest. These maps were constructed using video performance data from YouTube mobile applications based on real-world usage and do not indicate operator’s service coverage. The data points are based on quality measurements collected from YouTube usage by LTE-enabled users in a given geographical location over the past 35 days (see FAQ for qualification criteria). The individual data points are aggregated to a coarse location (like city or neighborhood) and drawn on the map. Scores are shown only for locations that meet the necessary privacy requirements. The map won’t provide scores for locations where there is insufficient data.
Scores are centered around mobile providers, not users. User identity like email addresses or usernames are never collected or reported as part of performance measurements. All observations are fully anonymized, and no personally identifiable information is used in the scoring algorithm. If the observation volume or the device volume is below certain thresholds, no score will be assigned for that location (for a given provider).