SlideShare una empresa de Scribd logo
1 de 10
Descargar para leer sin conexión
WHITE P APER
                                                               Streaming Video Quality and User Engagement
                                                               Sponsored by: Akamai

                                                               Karsten Weide
                                                               July 2011


                                                               IDC OPINION
www.idc.com




                                                               In an industry first, IDC conducted a statistical analysis of the server log files of six
                                                               major 2010 sports events that were streamed live to consumers in both North
                                                               America and Western Europe with a total of more than 2 million users.

                                                               The analysis found that both user engagement (measured as session length) and,
F.508.935.4015




                                                               consequently, unique user numbers were influenced by video quality. Several factors
                                                               were shown to have an impact:

                                                                Higher bit rates do increase user engagement. For each event, after a certain bit
                                                                 rate threshold, a further increase of bit rates had no additional positive effect on
P.508.872.8200




                                                                 user engagement anymore.

                                                                An important factor negatively impacting user engagement was the number of
                                                                 rebuffering events per hour.

                                                                Other, less influential negative factors were the share of time the video player
Global Headquarters: 5 Speen Street Framingham, MA 01701 USA




                                                                 spent rebuffering during users' sessions and the number of dropped frames per
                                                                 hour.

                                                               Our research suggests that measuring and monitoring key performance indicators (KPIs)
                                                               for video quality is of critical importance for publishers because they affect user
                                                               engagement and audience reach and therefore publishers' revenue and competitiveness.



                                                               METHODOLOGY
                                                               Akamai tasked IDC with a research project to explore the impact of different aspects
                                                               of online streaming video quality on user engagement. To that end, Akamai provided
                                                               IDC with the server log files of six major 2010 sports events that were streamed live
                                                               to consumers via Akamai's HD Network employing HTTP streaming, using adaptive
                                                               bit rate technology as outlined in Table 1.

                                                               For all of the following analyses, keep in mind that the bit rates offered by publishers
                                                               were vastly different between events. The FIFA World Cup soccer events in particular
                                                               were offered at comparatively low bit rates because of the massive crowds expected
                                                               to watch.
TABLE 1

    Event Overview

    Event                                     Region           Number of Users   Bit Rates Served

    Soccer A: FIFA World Cup 2010             North America       235,052        400, 750, 1000, 1300, 1800 kbps

    Soccer B: FIFA World Cup 2010             Western Europe       76,843        700, 1300, 2200, 3000 kbps

    Soccer C: FIFA World Cup 2010             Western Europe      125,244        400, 800, 1200, 1600 kbps

    Sports Event A: Major 2010 sports event   North America     1,376,727        564, 1064, 1564, 2200 kbps

    Sports Event B: Major 2010 sports event   North America        25,644        564, 1064, 1564, 2200 kbps

    Sports Event C: Major 2010 sports event   North America       469,876        Four bit rates under 2000 kbps

    Source: IDC, 2011




The log files were cleaned up before statistical analysis as follows:

 Sessions were consolidated by user ID. Each log file entry originally represented
  one viewing session. Where there were two or more separate sessions for the
  same user ID, these sessions were consolidated so that log entries represented the
  complete viewing experiences for each user ID for each event.

 Logs were cleaned up. We removed any entry where it was clear from the data
  that it was either impossible for the respective user to have seen the video
  (average playback bit rate was zero, number of average frames per second
  [FPS] was zero) or where more than 20% of total session time was spent
  rebuffering (with the picture frozen), making it unlikely for the user to have
  endured that bad of a viewing experience. We also ignored cases where the total
  aggregated session time was less than one minute, assuming that shorter
  sessions could not be counted as "viewing" a live video.

We also assumed that each user ID related to one person, even though several
persons or different persons at different times may have watched the video.
Furthermore, we assumed that users had spent the entire total aggregated session
time watching the video. In practice, users might have walked away from their PC or
could have had the video run in the background. For both assumptions, there was no
way for us to determine from the log files whether they held true.

IDC then conducted a statistical analysis of the remaining cases using the statistical
software package SPSS. The approach was to correlate user engagement (measured
as the total aggregated session time per user ID per event [short: session length or
session duration]) with certain measurements of streaming video quality for that
user's session during that event (see the Correlation section for details on the
statistical method):

 Average playback bit rate: The average bit rate at which the video was rendered
  on the user's screen as reported to the server by the user's video player


2                                                  #229083                                                ©2011 IDC
 Rebuffering events per hour: The number of times the buffer ran out of data
  and had to be replenished, possibly with the picture frozen if the rebuffering
  event was long enough to be noticeable by the user

 Percent of time spent rebuffering: The share of the total aggregated session
  time that was spent rebuffering

 Dropped frames per hour: The number of frames that the user's video player
  did not show

(The project did not analyze the impact of video start-up times on user engagement
because the log files did not include that information.)

The hypotheses were that:

 Where positive KPIs such as playback bit rate or average FPS were higher (i.e.,
  video quality was better), user engagement would also be higher (i.e., session
  times would be longer). The expected correlation coefficient would be > 0.

 Where positive KPIs were lower (i.e., video quality was worse), we expected user
  engagement to also be lower (i.e., session times would be shorter). The
  expected correlation coefficient would be < 0.

 Conversely, where negative KPIs were higher (i.e., video quality was worse), we
  expected user engagement to also be lower (i.e., session times would be
  shorter). The expected correlation coefficient would be < 0.

 Where negative KPIs were lower (i.e., video quality was better), user
  engagement would also be higher (i.e., session times would be shorter). The
  expected correlation coefficient would be > 0.

All correlation coefficients reported in this document were significant at the 0.01 level.
This means that mathematically, there is only a 1% likelihood that the reported
correlation occurred by chance.


Correlation

Correlation is a statistical method that analyzes the relationship between two sets of
data and expresses the closeness of their relation in a "correlation coefficient," a
single number between 1 and -1.

For instance, we compared the average bit rates at which thousands of users
watched a video and the total time they spent watching the video.

 If higher bit rates in each case translate into longer viewing times in a certain
  proportion, the correlation coefficient would be 1.

 If there was no relation at all between bit rates and viewing times, the coefficient
  would be 0.

 If higher bit rates in each case translate into shorter viewing times in a certain
  proportion, the correlation coefficient would be -1.

 Values between 0 and 1 and 0 and -1 would express varying degrees of relationship.


©2011 IDC                                      #229083                                       3
Correlation does not necessarily indicate causation (i.e., two sets of data might be
shown to relate to each other statistically even though there is no relation between
the two data sets in the real world).



IN THIS WHITE P APER
This IDC white paper explores the impact of different aspects of online streaming
video quality on user engagement based on the statistical analysis of server log files
of six major 2010 sports events that were streamed live to consumers.



SITUATION OVERVIEW

Statistical Analysis of the Impact of
Streaming Video Quality on User Engagement

IDC's statistical analysis of the server log files of six major sports events that were
streamed live to users found that user engagement was influenced by video quality.

We found the following two factors had a positive impact on session durations (i.e.,
they tended to improve user engagement):

 Higher playback bit rates (which of course are based on higher transferred bit
  rates) had the greatest impact in terms of extending session lengths, but only up
  to a certain optimal bit rate threshold. If bit rates were further increased beyond
  that threshold, session durations were not further increased, or they were not
  increased as much.

 Higher frame rates (frames per second) also had a positive impact on session
  lengths, but to a lesser extent than higher bit rates.

The following factors had a negative impact on session lengths — that is, they tended
to worsen user engagement (in sequence of their level of impact):

 The number of rebuffering events

 The share of the session time spent rebuffering

 The number of dropped frames

Of the preceding factors negatively impacting user engagement, one of the most
important was the number of rebuffering events. The share of viewing time spent
rebuffering and the number of dropped frames/s had less of an impact.

Our research suggests that measuring and monitoring KPIs for video quality is of
critical importance for publishers because they affect user engagement and audience
reach and therefore publisher revenue and competitiveness.




4                                             #229083                                     ©2011 IDC
Playback Bit Rate
Online video publishers all adopt high-quality or high-definition video for competitive
and branding purposes, also based on the experience in cable TV, where higher
resolutions translated into greater user engagement.

There has been a lot of discussion in the industry about whether increasing the bit
rate available to the user and thereby improving video resolution has a positive
impact on user engagement. Therefore, we began our analysis by correlating users'
average playback bit rates and session lengths. We expected a positive correlation
(i.e., that higher bit rates come with longer sessions).

Our statistical analysis showed that users did watch the video streams for a longer
time if they watched the event at higher playback bit rates (i.e., at higher video
resolutions) — but only up to a certain bit rate. That is, for each event, if we analyzed
only the cases up to that event's optimal bit rate threshold, correlation between
playback bit rates and session length was positive, which means that higher bit rates
tended to go with longer sessions (see Figure 1). The impact was slight, but
statistically significant.




FIGURE 1

Correlation Between Session Time and Playback Bit Rates for
Cases up to the Optimal Bit Rate Threshold for Each Event


           Soccer A

           Soccer B

           Soccer C

   Sports Event A

   Sports Event B

   Sports Event C

                       -0.4     -0.3     -0.2     -0.1     0.0     0.1      0.2         0.3     0.4
                                                (Correlation coef f icient)

Note: For the optimal bit rate threshold (i.e., the playback bit rates up to which cases were
analyzed for each of the above events), see Table 2.
Source: IDC, 2011




©2011 IDC                                            #229083                                          5
After that threshold, there was no additional positive effect on session times, or
the effect decreased. For the six events analyzed, the threshold was at different
levels (see Table 2). For publishers, this means that it is necessary to carefully
measure and monitor the impact of bit rate on session lengths to establish the optimal
bit rate range.



    TABLE 2

    Maximum Playback Bit Rate Level Showing Positive Impact on Session Length

                                     Maximum Bit Rate (kbps) up to Which
                                      Higher Bit Rates Further Improved
                                         Impact on Session Lengths                  Correlation Coefficient

    Soccer A                                         1000                                   0.120

    Soccer B                                         2500                                   0.148

    Soccer C                                         1000                                   0.031

    Sports Event A                                   2000                                   0.021

    Sports Event B                                   3500                                   0.048

    Sports Event C                                   1500                                   0.023

    Source: IDC, 2011




It is difficult to arrive at a formula that would express how much user engagement
(i.e., session lengths) increases as bit rate increases given the many factors that have
an impact on video performance (see next paragraph). But based on the kind of
performance increases we have seen in the data, we would expect to see, as a rule
of thumb, an increase of 10% in session lengths per 500 kbps increase in average
playback bit rate.

Keep in mind that the bit rates offered by publishers were vastly different between
events. This may be one reason why the cutoff is at different levels for different
events. We also theorize that there may be other effects at work as well. For instance,
those users who watch video at the highest bit rates also must have the infrastructure
(e.g., a high broadband access speed) in place to be able to watch at these rates.
Those users are also more likely to have incomes and busier lives, which could
explain why they are more likely to watch for shorter periods of time. More research is
needed, taking into account cultural, social, and situational factors.


Rebuffering
Rebuffering events are among the most frustrating experiences when watching a live
video stream. We analyzed the impact of rebuffering events per hour. These are
incidents where the buffer of the user's video player runs out of data and must be



6                                             #229083                                                 ©2011 IDC
replenished by the server while the picture freezes. Rebuffering is caused either by a
connection slowdown or by bad heuristics (i.e., when the player waits too long to
switch to a lower bit rate).

We expected a negative correlation between the number of rebuffering events and
session durations (i.e., for more rebuffering events to go with shorter sessions)
because with more interruptions of the video stream, users would become more
frustrated and more likely to stop watching it.

That is precisely what we found. The number of rebuffering events turned out to be
one of the worst factors impacting user engagement (see Figure 2).




FIGURE 2

Correlation Between Sessi on Time and Rebuffering Events per
Hour


           Soccer A

           Soccer B

           Soccer C

   Sports Event A

   Sports Event B

   Sports Event C

                      -0.4   -0.3   -0.2     -0.1     0.0     0.1      0.2   0.3   0.4
                                           (Correlation coef f icient)

Source: IDC, 2011




Then we looked at the impact of the total share of the session time that users had to
spend waiting for the rebuffering to complete and the video to resume (aggregating
the waiting time incurred by all rebuffering events) on session lengths.

Again, we expected a negative correlation (i.e., for higher rebuffering time shares to
go with shorter session durations). And again, we found that to be the case (see
Figure 3). Rebuffering duration was the second most important factor negatively
impacting user engagement. This also means the negative impact of rebuffering time
was smaller than that of the number of rebuffering events. Apparently, viewers find
that disruption as such is worse than waiting for it to end.




©2011 IDC                                       #229083                                  7
FIGURE 3

Correlation Between Session Time and Percent of Time Spent
Rebuffering


           Soccer A

           Soccer B

           Soccer C

    Sports Event A

    Sports Event B

    Sports Event C

                      -0.4   -0.3   -0.2     -0.1     0.0     0.1      0.2   0.3   0.4
                                           (Correlation coef f icient)

Source: IDC, 2011




Video Frames
The number of frames per second (FPS) or frame rate expresses the number of
consecutive images shown in a video transmission per second that create the illusion
of motion. Higher FPS numbers translate into better video quality because the video
plays more smoothly; lower FPS numbers conversely result in worse video streams.
Dropped frames are images that are not displayed by the user's video player, either
because local resources (CPU, graphics adapter, memory, etc.) are not sufficient or
because there is a disruption in the video transmission. From the user's perspective,
dropped frames translate into choppy video.

For the number of dropped frames per hour, we expected a negative correlation with
session durations (i.e., for more dropped frames to coincide with lower user
engagement) because dropped frames disrupt the viewing experience. This is what
we found in the numbers, too. Dropped frames per hour were the third most important
factor negatively impacting session lengths. However, the impact was fairly minimal
(see Figure 4).




8                                               #229083                                  ©2011 IDC
FIGURE 4

Correlation Between Session Time and Average Number of
Dropped Frames per Hour


           Soccer A

           Soccer B

           Soccer C

   Sports Event A

   Sports Event B

   Sports Event C

                      -0.4   -0.3     -0.2     -0.1     0.0     0.1      0.2     0.3     0.4
                                             (Correlation coef f icient)

Source: IDC, 2011




FUTURE OUTLOOK
This research has established that there is an impact of streaming video quality on
user engagement (i.e., session durations) and, therefore, on audience reach. As
consumers embrace online video distribution as a viable alternative to cable and
broadcast TV, their expectations of video quality will continue to increase. This will be
even more so as online video makes its way into consumers' living rooms, where it
ends up on high-definition television sets and will have to compete with the quality
that cable routinely provides. Here, a reliable online transmission at a resolution of
720p is only the beginning.

Publishers will need to embrace measuring and monitoring video quality such as
buffering, drop-off, and bit rate consumption on an ongoing, routine basis to tune the
experience and avoid dips in video quality and the resulting drop in user engagement
in order to protect their financial performance and competitiveness.

Given the wide range of bit rates offered in the six events analyzed, and the different
infrastructures given for them, it is difficult to arrive at a universal bit rate benchmark.
Two of the three soccer events in particular offered comparatively low bit rates
because of the expected huge numbers of viewers. If one looked only at the United
States, recommended bit rates would have to be set quite a bit higher.

Based on the given events, for the bit rate provided by publishers, IDC suggests
maintaining a level of at least 1200 kbps to attract the kind of audience numbers we
saw in the events analyzed. If one wanted to increase audience reach and user
engagement beyond that level, it would be prudent to increase the average bit rate to
1500 kbps.



©2011 IDC                                         #229083                                      9
The single most negative impact on engagement was the number of rebuffering
events. We believe the best practice is not about keeping rebuffering events to a
certain exact number; rather, it is about ensuring that the largest share of your
audience experiences no buffering at all.

Of course, only part of the occurrence of rebuffering events can be controlled by
publishers. Again, an optimized distribution technology and measuring and monitoring
rebuffering events are key to an optimized user engagement.

The second most important negative impact on viewer engagement is the amount of
time spent rebuffering. IDC recommends, as a rule of thumb, maintaining a level of
video quality at which users experience rebuffering for a maximum of 1% of the time.
This lines up with the experiences that publishers have had in practice. In an
interview with IDC, Glenn Goldstein, MTV's VP, Video Technology Strategy, said,
"Once rebuffering time hits 1% of the playback time, we know we're in trouble."

More research is needed regarding the impact of start-up times on user engagement
(which was not explored in this research) and the influence of demographic,
psychographic, and situational factors on video consumption.




Copyright Notice

External Publication of IDC Information and Data — Any IDC information that is to be
used in advertising, press releases, or promotional materials requires prior written
approval from the appropriate IDC Vice President or Country Manager. A draft of the
proposed document should accompany any such request. IDC reserves the right to
deny approval of external usage for any reason.

Copyright 2011 IDC. Reproduction without written permission is completely forbidden.




10                                           #229083                                   ©2011 IDC

Más contenido relacionado

Más de Akamai Technologies

Replacing recovery with resilience
Replacing recovery with resilienceReplacing recovery with resilience
Replacing recovery with resilienceAkamai Technologies
 
Competitive EDGE - Data Driven Differentiation
Competitive EDGE - Data Driven DifferentiationCompetitive EDGE - Data Driven Differentiation
Competitive EDGE - Data Driven DifferentiationAkamai Technologies
 
3 Reasons You Need Proactive Protection Against Malware
3 Reasons You Need Proactive Protection Against Malware3 Reasons You Need Proactive Protection Against Malware
3 Reasons You Need Proactive Protection Against MalwareAkamai Technologies
 
3 Reasons It's Time for a New Remote Access Model
3 Reasons It's Time for a New Remote Access Model3 Reasons It's Time for a New Remote Access Model
3 Reasons It's Time for a New Remote Access ModelAkamai Technologies
 
새로운 원격 접속 모델이 필요한 3가지 이유
새로운 원격 접속 모델이 필요한 3가지 이유새로운 원격 접속 모델이 필요한 3가지 이유
새로운 원격 접속 모델이 필요한 3가지 이유Akamai Technologies
 
更新遠端存取模式的 3 大理由
更新遠端存取模式的 3 大理由更新遠端存取模式的 3 大理由
更新遠端存取模式的 3 大理由Akamai Technologies
 
应该采用全新远程访问模式的 3 大原因
应该采用全新远程访问模式的 3 大原因应该采用全新远程访问模式的 3 大原因
应该采用全新远程访问模式的 3 大原因Akamai Technologies
 
3 つの理由 今こそ新しいリモート・アク セス・モデルを採用すべきと き
3 つの理由 今こそ新しいリモート・アク セス・モデルを採用すべきと き3 つの理由 今こそ新しいリモート・アク セス・モデルを採用すべきと き
3 つの理由 今こそ新しいリモート・アク セス・モデルを採用すべきと きAkamai Technologies
 
3 razões chegou a hora de um novo modelo de acesso remoto
3 razões chegou a hora de um novo modelo de acesso remoto3 razões chegou a hora de um novo modelo de acesso remoto
3 razões chegou a hora de um novo modelo de acesso remotoAkamai Technologies
 
3 motivi per cui è necessario un nuovo modello di accesso remoto
3 motivi per cui è necessario un nuovo modello di accesso remoto3 motivi per cui è necessario un nuovo modello di accesso remoto
3 motivi per cui è necessario un nuovo modello di accesso remotoAkamai Technologies
 
3 raisons de changer votre modèle d'accès à distance
3 raisons de changer votre modèle d'accès à distance3 raisons de changer votre modèle d'accès à distance
3 raisons de changer votre modèle d'accès à distanceAkamai Technologies
 
3 motivos por los que ahora es el momento perfecto para adoptar un nuevo mode...
3 motivos por los que ahora es el momento perfecto para adoptar un nuevo mode...3 motivos por los que ahora es el momento perfecto para adoptar un nuevo mode...
3 motivos por los que ahora es el momento perfecto para adoptar un nuevo mode...Akamai Technologies
 
3 Gründe für eine neue Art des Remotezugriffs
3 Gründe für eine neue Art des Remotezugriffs3 Gründe für eine neue Art des Remotezugriffs
3 Gründe für eine neue Art des RemotezugriffsAkamai Technologies
 
Chicago Tech Day Jan 2015: Foundry - HTTP2
Chicago Tech Day Jan 2015: Foundry - HTTP2Chicago Tech Day Jan 2015: Foundry - HTTP2
Chicago Tech Day Jan 2015: Foundry - HTTP2Akamai Technologies
 
Chicago Tech Day Jan 2015: Hidden Features
Chicago Tech Day Jan 2015: Hidden FeaturesChicago Tech Day Jan 2015: Hidden Features
Chicago Tech Day Jan 2015: Hidden FeaturesAkamai Technologies
 
Customer Technology Day Chicago 2015
Customer Technology Day Chicago 2015Customer Technology Day Chicago 2015
Customer Technology Day Chicago 2015Akamai Technologies
 
Edge 2014: Maintaining the Balance: Getting the Most of Your CDN with IKEA
Edge 2014: Maintaining the Balance: Getting the Most of Your CDN with IKEAEdge 2014: Maintaining the Balance: Getting the Most of Your CDN with IKEA
Edge 2014: Maintaining the Balance: Getting the Most of Your CDN with IKEAAkamai Technologies
 
Edge 2014: Increasing Control with Property Manager with eBay
Edge 2014: Increasing Control with Property Manager with eBayEdge 2014: Increasing Control with Property Manager with eBay
Edge 2014: Increasing Control with Property Manager with eBayAkamai Technologies
 
Edge 2014: Bypass Surgery - Akamai's Heartbleed Response Case Study
Edge 2014: Bypass Surgery - Akamai's Heartbleed Response Case StudyEdge 2014: Bypass Surgery - Akamai's Heartbleed Response Case Study
Edge 2014: Bypass Surgery - Akamai's Heartbleed Response Case StudyAkamai Technologies
 

Más de Akamai Technologies (20)

Replacing recovery with resilience
Replacing recovery with resilienceReplacing recovery with resilience
Replacing recovery with resilience
 
Competitive EDGE - Data Driven Differentiation
Competitive EDGE - Data Driven DifferentiationCompetitive EDGE - Data Driven Differentiation
Competitive EDGE - Data Driven Differentiation
 
3 Reasons You Need Proactive Protection Against Malware
3 Reasons You Need Proactive Protection Against Malware3 Reasons You Need Proactive Protection Against Malware
3 Reasons You Need Proactive Protection Against Malware
 
3 Reasons It's Time for a New Remote Access Model
3 Reasons It's Time for a New Remote Access Model3 Reasons It's Time for a New Remote Access Model
3 Reasons It's Time for a New Remote Access Model
 
새로운 원격 접속 모델이 필요한 3가지 이유
새로운 원격 접속 모델이 필요한 3가지 이유새로운 원격 접속 모델이 필요한 3가지 이유
새로운 원격 접속 모델이 필요한 3가지 이유
 
更新遠端存取模式的 3 大理由
更新遠端存取模式的 3 大理由更新遠端存取模式的 3 大理由
更新遠端存取模式的 3 大理由
 
应该采用全新远程访问模式的 3 大原因
应该采用全新远程访问模式的 3 大原因应该采用全新远程访问模式的 3 大原因
应该采用全新远程访问模式的 3 大原因
 
3 つの理由 今こそ新しいリモート・アク セス・モデルを採用すべきと き
3 つの理由 今こそ新しいリモート・アク セス・モデルを採用すべきと き3 つの理由 今こそ新しいリモート・アク セス・モデルを採用すべきと き
3 つの理由 今こそ新しいリモート・アク セス・モデルを採用すべきと き
 
3 razões chegou a hora de um novo modelo de acesso remoto
3 razões chegou a hora de um novo modelo de acesso remoto3 razões chegou a hora de um novo modelo de acesso remoto
3 razões chegou a hora de um novo modelo de acesso remoto
 
3 motivi per cui è necessario un nuovo modello di accesso remoto
3 motivi per cui è necessario un nuovo modello di accesso remoto3 motivi per cui è necessario un nuovo modello di accesso remoto
3 motivi per cui è necessario un nuovo modello di accesso remoto
 
3 raisons de changer votre modèle d'accès à distance
3 raisons de changer votre modèle d'accès à distance3 raisons de changer votre modèle d'accès à distance
3 raisons de changer votre modèle d'accès à distance
 
3 motivos por los que ahora es el momento perfecto para adoptar un nuevo mode...
3 motivos por los que ahora es el momento perfecto para adoptar un nuevo mode...3 motivos por los que ahora es el momento perfecto para adoptar un nuevo mode...
3 motivos por los que ahora es el momento perfecto para adoptar un nuevo mode...
 
3 Gründe für eine neue Art des Remotezugriffs
3 Gründe für eine neue Art des Remotezugriffs3 Gründe für eine neue Art des Remotezugriffs
3 Gründe für eine neue Art des Remotezugriffs
 
Chicago Tech Day Jan 2015: Foundry - HTTP2
Chicago Tech Day Jan 2015: Foundry - HTTP2Chicago Tech Day Jan 2015: Foundry - HTTP2
Chicago Tech Day Jan 2015: Foundry - HTTP2
 
Chicago Tech Day Jan 2015: RWD
Chicago Tech Day Jan 2015: RWDChicago Tech Day Jan 2015: RWD
Chicago Tech Day Jan 2015: RWD
 
Chicago Tech Day Jan 2015: Hidden Features
Chicago Tech Day Jan 2015: Hidden FeaturesChicago Tech Day Jan 2015: Hidden Features
Chicago Tech Day Jan 2015: Hidden Features
 
Customer Technology Day Chicago 2015
Customer Technology Day Chicago 2015Customer Technology Day Chicago 2015
Customer Technology Day Chicago 2015
 
Edge 2014: Maintaining the Balance: Getting the Most of Your CDN with IKEA
Edge 2014: Maintaining the Balance: Getting the Most of Your CDN with IKEAEdge 2014: Maintaining the Balance: Getting the Most of Your CDN with IKEA
Edge 2014: Maintaining the Balance: Getting the Most of Your CDN with IKEA
 
Edge 2014: Increasing Control with Property Manager with eBay
Edge 2014: Increasing Control with Property Manager with eBayEdge 2014: Increasing Control with Property Manager with eBay
Edge 2014: Increasing Control with Property Manager with eBay
 
Edge 2014: Bypass Surgery - Akamai's Heartbleed Response Case Study
Edge 2014: Bypass Surgery - Akamai's Heartbleed Response Case StudyEdge 2014: Bypass Surgery - Akamai's Heartbleed Response Case Study
Edge 2014: Bypass Surgery - Akamai's Heartbleed Response Case Study
 

Último

Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Enterprise Knowledge
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking MenDelhi Call girls
 
08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking MenDelhi Call girls
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationRadu Cotescu
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024The Digital Insurer
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationMichael W. Hawkins
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsMaria Levchenko
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxKatpro Technologies
 
Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024The Digital Insurer
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Allon Mureinik
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Drew Madelung
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processorsdebabhi2
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024Results
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc
 
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure serviceWhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure servicePooja Nehwal
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonAnna Loughnan Colquhoun
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024Rafal Los
 
08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking MenDelhi Call girls
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountPuma Security, LLC
 

Último (20)

Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
 
08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organization
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day Presentation
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed texts
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
 
Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processors
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure serviceWhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024
 
08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path Mount
 

Streaming Video Quality & User Engagement Whitepaper: IDC & Akamai

  • 1. WHITE P APER Streaming Video Quality and User Engagement Sponsored by: Akamai Karsten Weide July 2011 IDC OPINION www.idc.com In an industry first, IDC conducted a statistical analysis of the server log files of six major 2010 sports events that were streamed live to consumers in both North America and Western Europe with a total of more than 2 million users. The analysis found that both user engagement (measured as session length) and, F.508.935.4015 consequently, unique user numbers were influenced by video quality. Several factors were shown to have an impact:  Higher bit rates do increase user engagement. For each event, after a certain bit rate threshold, a further increase of bit rates had no additional positive effect on P.508.872.8200 user engagement anymore.  An important factor negatively impacting user engagement was the number of rebuffering events per hour.  Other, less influential negative factors were the share of time the video player Global Headquarters: 5 Speen Street Framingham, MA 01701 USA spent rebuffering during users' sessions and the number of dropped frames per hour. Our research suggests that measuring and monitoring key performance indicators (KPIs) for video quality is of critical importance for publishers because they affect user engagement and audience reach and therefore publishers' revenue and competitiveness. METHODOLOGY Akamai tasked IDC with a research project to explore the impact of different aspects of online streaming video quality on user engagement. To that end, Akamai provided IDC with the server log files of six major 2010 sports events that were streamed live to consumers via Akamai's HD Network employing HTTP streaming, using adaptive bit rate technology as outlined in Table 1. For all of the following analyses, keep in mind that the bit rates offered by publishers were vastly different between events. The FIFA World Cup soccer events in particular were offered at comparatively low bit rates because of the massive crowds expected to watch.
  • 2. TABLE 1 Event Overview Event Region Number of Users Bit Rates Served Soccer A: FIFA World Cup 2010 North America 235,052 400, 750, 1000, 1300, 1800 kbps Soccer B: FIFA World Cup 2010 Western Europe 76,843 700, 1300, 2200, 3000 kbps Soccer C: FIFA World Cup 2010 Western Europe 125,244 400, 800, 1200, 1600 kbps Sports Event A: Major 2010 sports event North America 1,376,727 564, 1064, 1564, 2200 kbps Sports Event B: Major 2010 sports event North America 25,644 564, 1064, 1564, 2200 kbps Sports Event C: Major 2010 sports event North America 469,876 Four bit rates under 2000 kbps Source: IDC, 2011 The log files were cleaned up before statistical analysis as follows:  Sessions were consolidated by user ID. Each log file entry originally represented one viewing session. Where there were two or more separate sessions for the same user ID, these sessions were consolidated so that log entries represented the complete viewing experiences for each user ID for each event.  Logs were cleaned up. We removed any entry where it was clear from the data that it was either impossible for the respective user to have seen the video (average playback bit rate was zero, number of average frames per second [FPS] was zero) or where more than 20% of total session time was spent rebuffering (with the picture frozen), making it unlikely for the user to have endured that bad of a viewing experience. We also ignored cases where the total aggregated session time was less than one minute, assuming that shorter sessions could not be counted as "viewing" a live video. We also assumed that each user ID related to one person, even though several persons or different persons at different times may have watched the video. Furthermore, we assumed that users had spent the entire total aggregated session time watching the video. In practice, users might have walked away from their PC or could have had the video run in the background. For both assumptions, there was no way for us to determine from the log files whether they held true. IDC then conducted a statistical analysis of the remaining cases using the statistical software package SPSS. The approach was to correlate user engagement (measured as the total aggregated session time per user ID per event [short: session length or session duration]) with certain measurements of streaming video quality for that user's session during that event (see the Correlation section for details on the statistical method):  Average playback bit rate: The average bit rate at which the video was rendered on the user's screen as reported to the server by the user's video player 2 #229083 ©2011 IDC
  • 3.  Rebuffering events per hour: The number of times the buffer ran out of data and had to be replenished, possibly with the picture frozen if the rebuffering event was long enough to be noticeable by the user  Percent of time spent rebuffering: The share of the total aggregated session time that was spent rebuffering  Dropped frames per hour: The number of frames that the user's video player did not show (The project did not analyze the impact of video start-up times on user engagement because the log files did not include that information.) The hypotheses were that:  Where positive KPIs such as playback bit rate or average FPS were higher (i.e., video quality was better), user engagement would also be higher (i.e., session times would be longer). The expected correlation coefficient would be > 0.  Where positive KPIs were lower (i.e., video quality was worse), we expected user engagement to also be lower (i.e., session times would be shorter). The expected correlation coefficient would be < 0.  Conversely, where negative KPIs were higher (i.e., video quality was worse), we expected user engagement to also be lower (i.e., session times would be shorter). The expected correlation coefficient would be < 0.  Where negative KPIs were lower (i.e., video quality was better), user engagement would also be higher (i.e., session times would be shorter). The expected correlation coefficient would be > 0. All correlation coefficients reported in this document were significant at the 0.01 level. This means that mathematically, there is only a 1% likelihood that the reported correlation occurred by chance. Correlation Correlation is a statistical method that analyzes the relationship between two sets of data and expresses the closeness of their relation in a "correlation coefficient," a single number between 1 and -1. For instance, we compared the average bit rates at which thousands of users watched a video and the total time they spent watching the video.  If higher bit rates in each case translate into longer viewing times in a certain proportion, the correlation coefficient would be 1.  If there was no relation at all between bit rates and viewing times, the coefficient would be 0.  If higher bit rates in each case translate into shorter viewing times in a certain proportion, the correlation coefficient would be -1.  Values between 0 and 1 and 0 and -1 would express varying degrees of relationship. ©2011 IDC #229083 3
  • 4. Correlation does not necessarily indicate causation (i.e., two sets of data might be shown to relate to each other statistically even though there is no relation between the two data sets in the real world). IN THIS WHITE P APER This IDC white paper explores the impact of different aspects of online streaming video quality on user engagement based on the statistical analysis of server log files of six major 2010 sports events that were streamed live to consumers. SITUATION OVERVIEW Statistical Analysis of the Impact of Streaming Video Quality on User Engagement IDC's statistical analysis of the server log files of six major sports events that were streamed live to users found that user engagement was influenced by video quality. We found the following two factors had a positive impact on session durations (i.e., they tended to improve user engagement):  Higher playback bit rates (which of course are based on higher transferred bit rates) had the greatest impact in terms of extending session lengths, but only up to a certain optimal bit rate threshold. If bit rates were further increased beyond that threshold, session durations were not further increased, or they were not increased as much.  Higher frame rates (frames per second) also had a positive impact on session lengths, but to a lesser extent than higher bit rates. The following factors had a negative impact on session lengths — that is, they tended to worsen user engagement (in sequence of their level of impact):  The number of rebuffering events  The share of the session time spent rebuffering  The number of dropped frames Of the preceding factors negatively impacting user engagement, one of the most important was the number of rebuffering events. The share of viewing time spent rebuffering and the number of dropped frames/s had less of an impact. Our research suggests that measuring and monitoring KPIs for video quality is of critical importance for publishers because they affect user engagement and audience reach and therefore publisher revenue and competitiveness. 4 #229083 ©2011 IDC
  • 5. Playback Bit Rate Online video publishers all adopt high-quality or high-definition video for competitive and branding purposes, also based on the experience in cable TV, where higher resolutions translated into greater user engagement. There has been a lot of discussion in the industry about whether increasing the bit rate available to the user and thereby improving video resolution has a positive impact on user engagement. Therefore, we began our analysis by correlating users' average playback bit rates and session lengths. We expected a positive correlation (i.e., that higher bit rates come with longer sessions). Our statistical analysis showed that users did watch the video streams for a longer time if they watched the event at higher playback bit rates (i.e., at higher video resolutions) — but only up to a certain bit rate. That is, for each event, if we analyzed only the cases up to that event's optimal bit rate threshold, correlation between playback bit rates and session length was positive, which means that higher bit rates tended to go with longer sessions (see Figure 1). The impact was slight, but statistically significant. FIGURE 1 Correlation Between Session Time and Playback Bit Rates for Cases up to the Optimal Bit Rate Threshold for Each Event Soccer A Soccer B Soccer C Sports Event A Sports Event B Sports Event C -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 (Correlation coef f icient) Note: For the optimal bit rate threshold (i.e., the playback bit rates up to which cases were analyzed for each of the above events), see Table 2. Source: IDC, 2011 ©2011 IDC #229083 5
  • 6. After that threshold, there was no additional positive effect on session times, or the effect decreased. For the six events analyzed, the threshold was at different levels (see Table 2). For publishers, this means that it is necessary to carefully measure and monitor the impact of bit rate on session lengths to establish the optimal bit rate range. TABLE 2 Maximum Playback Bit Rate Level Showing Positive Impact on Session Length Maximum Bit Rate (kbps) up to Which Higher Bit Rates Further Improved Impact on Session Lengths Correlation Coefficient Soccer A 1000 0.120 Soccer B 2500 0.148 Soccer C 1000 0.031 Sports Event A 2000 0.021 Sports Event B 3500 0.048 Sports Event C 1500 0.023 Source: IDC, 2011 It is difficult to arrive at a formula that would express how much user engagement (i.e., session lengths) increases as bit rate increases given the many factors that have an impact on video performance (see next paragraph). But based on the kind of performance increases we have seen in the data, we would expect to see, as a rule of thumb, an increase of 10% in session lengths per 500 kbps increase in average playback bit rate. Keep in mind that the bit rates offered by publishers were vastly different between events. This may be one reason why the cutoff is at different levels for different events. We also theorize that there may be other effects at work as well. For instance, those users who watch video at the highest bit rates also must have the infrastructure (e.g., a high broadband access speed) in place to be able to watch at these rates. Those users are also more likely to have incomes and busier lives, which could explain why they are more likely to watch for shorter periods of time. More research is needed, taking into account cultural, social, and situational factors. Rebuffering Rebuffering events are among the most frustrating experiences when watching a live video stream. We analyzed the impact of rebuffering events per hour. These are incidents where the buffer of the user's video player runs out of data and must be 6 #229083 ©2011 IDC
  • 7. replenished by the server while the picture freezes. Rebuffering is caused either by a connection slowdown or by bad heuristics (i.e., when the player waits too long to switch to a lower bit rate). We expected a negative correlation between the number of rebuffering events and session durations (i.e., for more rebuffering events to go with shorter sessions) because with more interruptions of the video stream, users would become more frustrated and more likely to stop watching it. That is precisely what we found. The number of rebuffering events turned out to be one of the worst factors impacting user engagement (see Figure 2). FIGURE 2 Correlation Between Sessi on Time and Rebuffering Events per Hour Soccer A Soccer B Soccer C Sports Event A Sports Event B Sports Event C -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 (Correlation coef f icient) Source: IDC, 2011 Then we looked at the impact of the total share of the session time that users had to spend waiting for the rebuffering to complete and the video to resume (aggregating the waiting time incurred by all rebuffering events) on session lengths. Again, we expected a negative correlation (i.e., for higher rebuffering time shares to go with shorter session durations). And again, we found that to be the case (see Figure 3). Rebuffering duration was the second most important factor negatively impacting user engagement. This also means the negative impact of rebuffering time was smaller than that of the number of rebuffering events. Apparently, viewers find that disruption as such is worse than waiting for it to end. ©2011 IDC #229083 7
  • 8. FIGURE 3 Correlation Between Session Time and Percent of Time Spent Rebuffering Soccer A Soccer B Soccer C Sports Event A Sports Event B Sports Event C -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 (Correlation coef f icient) Source: IDC, 2011 Video Frames The number of frames per second (FPS) or frame rate expresses the number of consecutive images shown in a video transmission per second that create the illusion of motion. Higher FPS numbers translate into better video quality because the video plays more smoothly; lower FPS numbers conversely result in worse video streams. Dropped frames are images that are not displayed by the user's video player, either because local resources (CPU, graphics adapter, memory, etc.) are not sufficient or because there is a disruption in the video transmission. From the user's perspective, dropped frames translate into choppy video. For the number of dropped frames per hour, we expected a negative correlation with session durations (i.e., for more dropped frames to coincide with lower user engagement) because dropped frames disrupt the viewing experience. This is what we found in the numbers, too. Dropped frames per hour were the third most important factor negatively impacting session lengths. However, the impact was fairly minimal (see Figure 4). 8 #229083 ©2011 IDC
  • 9. FIGURE 4 Correlation Between Session Time and Average Number of Dropped Frames per Hour Soccer A Soccer B Soccer C Sports Event A Sports Event B Sports Event C -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 (Correlation coef f icient) Source: IDC, 2011 FUTURE OUTLOOK This research has established that there is an impact of streaming video quality on user engagement (i.e., session durations) and, therefore, on audience reach. As consumers embrace online video distribution as a viable alternative to cable and broadcast TV, their expectations of video quality will continue to increase. This will be even more so as online video makes its way into consumers' living rooms, where it ends up on high-definition television sets and will have to compete with the quality that cable routinely provides. Here, a reliable online transmission at a resolution of 720p is only the beginning. Publishers will need to embrace measuring and monitoring video quality such as buffering, drop-off, and bit rate consumption on an ongoing, routine basis to tune the experience and avoid dips in video quality and the resulting drop in user engagement in order to protect their financial performance and competitiveness. Given the wide range of bit rates offered in the six events analyzed, and the different infrastructures given for them, it is difficult to arrive at a universal bit rate benchmark. Two of the three soccer events in particular offered comparatively low bit rates because of the expected huge numbers of viewers. If one looked only at the United States, recommended bit rates would have to be set quite a bit higher. Based on the given events, for the bit rate provided by publishers, IDC suggests maintaining a level of at least 1200 kbps to attract the kind of audience numbers we saw in the events analyzed. If one wanted to increase audience reach and user engagement beyond that level, it would be prudent to increase the average bit rate to 1500 kbps. ©2011 IDC #229083 9
  • 10. The single most negative impact on engagement was the number of rebuffering events. We believe the best practice is not about keeping rebuffering events to a certain exact number; rather, it is about ensuring that the largest share of your audience experiences no buffering at all. Of course, only part of the occurrence of rebuffering events can be controlled by publishers. Again, an optimized distribution technology and measuring and monitoring rebuffering events are key to an optimized user engagement. The second most important negative impact on viewer engagement is the amount of time spent rebuffering. IDC recommends, as a rule of thumb, maintaining a level of video quality at which users experience rebuffering for a maximum of 1% of the time. This lines up with the experiences that publishers have had in practice. In an interview with IDC, Glenn Goldstein, MTV's VP, Video Technology Strategy, said, "Once rebuffering time hits 1% of the playback time, we know we're in trouble." More research is needed regarding the impact of start-up times on user engagement (which was not explored in this research) and the influence of demographic, psychographic, and situational factors on video consumption. Copyright Notice External Publication of IDC Information and Data — Any IDC information that is to be used in advertising, press releases, or promotional materials requires prior written approval from the appropriate IDC Vice President or Country Manager. A draft of the proposed document should accompany any such request. IDC reserves the right to deny approval of external usage for any reason. Copyright 2011 IDC. Reproduction without written permission is completely forbidden. 10 #229083 ©2011 IDC