Conviva SPI Calculation in Session Source Data
This document describes how to calculation SPI data based on Session Source Data (SSD) reports.
The Conviva SPI provides a visual indicator of your KPIs performance so you can quickly determine the number and percent of impacted streams and performance level. A stream is impacted when it fails to meet one or more of the defined KPI settings.
Conviva formulates a unified streaming performance KPI based on the percentage of streaming sessions with good or best viewing experience. This KPI represents the Conviva Streaming Performance Index, and is based on the percentage of streams with:
-
No errors (VSF-T or VPF-T)
-
No or very low Rebuffering (using CIRR)
-
Acceptable picture quality based on average bitrate for different screen sizes
-
Acceptable Video Start Time
-
No EBVS if the viewer was waiting a long time before exiting.
Conviva provides KPI settings based on Good and Best performance. You can also set custom KPI settings to match your specific performance goals.
Good Performance KPI Settings
-
No Errors (VSF-T or VPF-T)
-
VST-T < 10sec
-
EBVS wait time < 10sec
-
Avg. Peak Bitrate > 800Kbps for TV screens
-
Avg. Peak Bitrate > 400Kbps for desktop or tablets
-
Avg. Peak Bitrate > 200Kbps for mobile devices screens
-
CIRR < 0.4%
-
CIRT < 2sec
Best Performance KPI Settings
-
No Errors (VSF-T or VPF-T)
-
VST < 10sec
-
EBVS wait time < 8sec
-
Avg. Peak Bitrate > 2Mbps for TV screens
Note: Encoding compression are not considered in the bit rates. -
Avg. Peak Bitrate > 800Kbps for desktop or tablets
-
Avg. Peak Bitrate > 400Kbps for mobile devices screens
-
CIRR < 0.02%
-
CIRT < 1sec
This example shows one approach for deriving SPI from your SSD data.
-
Select and compute fields necessary for SPI calculation, such as startup time, average bitrate, and rebuffering.
create or replace temporary view simple_ssd as
select
case when 'startup_error' = 0 and 'startup_time_ms' = -1 then true else false end as ebvs,
end_time_unix_time - start_time_unix_time as session_time_sec,
VSF_T,
VPF_T,
case when ended_status > 0 and playing_time_ms > 0 then true else false end as ended_play,
startup_time_ms,
playing_time_ms,
rebuffering_time_ms,
connection_induced_rebuffering_time_ms,
average_bitrate_kbps,
session_tags['Device Hardware Type'] as device_hardware_type
from
ssd-
Determine if each session is a valid SPI stream. SPI streams do not include non-ended plays, failures due to business errors, and non-abandonment EBVS.
create or replace temporary view base as
select
*,
round(connection_induced_rebuffering_time_ms / (playing_time_ms + rebuffering_time_ms), 4) as cirr,
case
when ebvs and session_time_sec > 10 then true
when VSF_T then true
when VPF_T then true
when ended_play then true
else false
end as good_spi_stream,
case
when ebvs and session_time_sec > 8 then true
when VSF_T then true
when VPF_T then true
when ended_play then true
else false
end as best_spi_stream
from
simple_ssd-
Compute whether the SPI sessions qualify against the SPI Good and Best threshold settings.
create or replace temporary view prep as
select
good_spi_stream,
best_spi_stream,
case
when ebvs and session_time_sec > 10 then "Bad"
when VSF_T then "Bad"
when VPF_T then "Bad"
when ended_play
and
(
startup_time_ms > 10000
or
(cirr > 0.004 and connection_induced_rebuffering_time_ms > 2000) -- calculate percentages as values between 0 and 1 instead of 0 and 100, so threshold is 0.004 instead of 0.4
or
(
(average_bitrate_kbps > 0 and average_bitrate_kbps <= 200)
or
(device_hardware_type in ("Desktop","Tablet") and average_bitrate_kbps > 0 and average_bitrate_kbps <= 400)
or
(device_hardware_type in ("TV","Set Top Box","Games Console") and average_bitrate_kbps > 0 and average_bitrate_kbps <= 800)
)
) then "Bad"
end as good_spi_bad,
case
when ebvs and session_time_sec > 8 then "Bad"
when VSF_T then "Bad"
when VPF_T then "Bad"
when ended_play
and
(
startup_time_ms > 10000
or
(cirr > 0.0002 and connection_induced_rebuffering_time_ms >= 1000) -- calculate percentages as values between 0 and 1 instead of 0 and 100, so threshold is 0.0002 instead of 0.02
or
(
(average_bitrate_kbps > 0 and average_bitrate_kbps <= 400)
or
(device_hardware_type in ("Desktop","Tablet") and average_bitrate_kbps > 0 and average_bitrate_kbps <= 800)
or
(device_hardware_type in ("TV","Set Top Box","Games Console") and average_bitrate_kbps > 0 and average_bitrate_kbps <= 2000)
)
) then "Bad"
end as best_spi_bad
from
base-
Aggregate and total the session performance and compute the SPI scores.
select
round((1 - sum(case when good_spi_bad = "Bad" then 1 else 0 end) / sum(case when good_spi_stream then 1 else 0 end)), 4) * 100 as good_spi,
round((1 - sum(case when best_spi_bad = "Bad" then 1 else 0 end) / sum(case when best_spi_stream then 1 else 0 end)), 4) * 100 as best_spi
from
prep-- Select and compute fields necessary for SPI calculation from SSD
create or replace temporary view simple_ssd as
select
case when 'startup_error' = 0 and 'startup_time_ms' = -1 then true else false end as ebvs,
end_time_unix_time - start_time_unix_time as session_time_sec,
VSF_T,
VPF_T,
case when ended_status > 0 and playing_time_ms > 0 then true else false end as ended_play,
startup_time_ms,
playing_time_ms,
connection_induced_rebuffering_time_ms,
average_bitrate_kbps,
session_tags['Device Hardware Type'] as device_hardware_type
from
ssd
-- Determine whether each session is an SPI stream.
-- SPI streams do not include non-ended plays, VPF-B, VSF-B, and non-abandonment EBVS.
-- This is because we cannot definitively say whether such sessions are good or bad.
create or replace temporary view base as
select
*,
round(connection_induced_rebuffering_time_ms / (playing_time_ms + rebuffering_time_ms), 4) as cirr,
case
when ebvs and session_time_sec > 10 then true
when VSF_T then true
when VPF_T then true
when ended_play then true
else false
end as good_spi_stream,
case
when ebvs and session_time_sec > 8 then true
when VSF_T then true
when VPF_T then true
when ended_play then true
else false
end as best_spi_stream
from
simple_ssd
-- Compute whether a session is good or bad based on Good and Best SPI thresholds.
create or replace temporary view prep as
select
good_spi_stream,
best_spi_stream,
case
when ebvs and session_time_sec > 10 then "Bad"
when VSF_T then "Bad"
when VPF_T then "Bad"
when ended_play
and
(
startup_time_ms > 10000
or
(cirr > 0.004 and connection_induced_rebuffering_time_ms > 2000) -- Percentages as values between 0 and 1 instead of 0 and 100, so threshold is 0.004 instead of 0.4
or
(
average_bitrate_kbps <= 200
or
device_hardware_type in ("Desktop","Tablet") and average_bitrate_kbps <= 400
or
device_hardware_type in ("TV","Set Top Box","Games Console") and average_bitrate_kbps <= 800
)
) then "Bad"
end as good_spi_bad,
case
when ebvs and session_time_sec > 8 then "Bad"
when VSF_T then "Bad"
when VPF_T then "Bad"
when ended_play
and
(
startup_time_ms > 10000
or
(cirr > 0.0002 and connection_induced_rebuffering_time_ms >= 1000) -- calculate percentages as values between 0 and 1 instead of 0 and 100, so threshold is 0.0002 instead of 0.02
or
(
average_bitrate_kbps <= 400
or
device_hardware_type in ("Desktop","Tablet") and average_bitrate_kbps <= 800
or
device_hardware_type in ("TV","Set Top Box","Games Console") and average_bitrate_kbps <= 2000
)
) then "Bad"
end as best_spi_bad
from
base
-- Compute aggregate SPI scores.
select
round((1 - sum(case when good_spi_bad = "Bad" then 1 else 0 end) / sum(case when good_spi_stream then 1 else 0 end)), 4) * 100 as good_spi,
round((1 - sum(case when best_spi_bad = "Bad" then 1 else 0 end) / sum(case when best_spi_stream then 1 else 0 end)), 4) * 100 as best_spi
from
prepType to search. Results appear as you type.