1
00:00:06,302 --> 00:00:08,277
- Now let's review a use case

2
00:00:08,277 --> 00:00:10,607
where we perform business intelligence

3
00:00:10,607 --> 00:00:13,473
on a video subscription service.

4
00:00:13,473 --> 00:00:17,290
In this example, we have a
video subscription service,

5
00:00:17,290 --> 00:00:19,318
something similar to Netflix,

6
00:00:19,318 --> 00:00:23,393
where our customers are
out there watching videos

7
00:00:23,393 --> 00:00:27,311
on their laptops, their
desktops, their mobile devices,

8
00:00:27,311 --> 00:00:29,732
and as they're watching videos,

9
00:00:29,732 --> 00:00:33,503
as they're pressing Play,
as they are pressing Pause,

10
00:00:33,503 --> 00:00:35,631
as they're scrubbing forward, or backward,

11
00:00:35,631 --> 00:00:37,501
or replaying, or re-watching,

12
00:00:37,501 --> 00:00:40,727
or favoriting a particular
section of a video,

13
00:00:40,727 --> 00:00:45,036
or leaving comments, all of
these events are coming in

14
00:00:45,036 --> 00:00:46,869
to our Kinesis Stream.

15
00:00:48,620 --> 00:00:51,057
From that stream we have a number

16
00:00:51,057 --> 00:00:52,913
of different applications
that are doing things

17
00:00:52,913 --> 00:00:55,296
like Sliding Window analysis,

18
00:00:55,296 --> 00:00:59,301
how many videos are actively
being played right now,

19
00:00:59,301 --> 00:01:01,421
what videos are being played right now,

20
00:01:01,421 --> 00:01:05,389
what categories are
being watched right now.

21
00:01:05,389 --> 00:01:09,099
As new users actually
sign up we want to be able

22
00:01:09,099 --> 00:01:12,233
to see how many users
actually became new users

23
00:01:12,233 --> 00:01:15,742
in the last minute or the last hour.

24
00:01:15,742 --> 00:01:18,320
We're doing these Sliding Window analyses

25
00:01:18,320 --> 00:01:21,703
and those are being stored in DynamoDB.

26
00:01:21,703 --> 00:01:23,408
We have another application
that is just taking

27
00:01:23,408 --> 00:01:27,575
the raw event messages,
aggregating those per hour

28
00:01:29,153 --> 00:01:33,153
or per day and storing
those files in Amazon S3.

29
00:01:35,978 --> 00:01:39,898
From there we have our
User Account information,

30
00:01:39,898 --> 00:01:42,268
we have Video meta-data being stored

31
00:01:42,268 --> 00:01:46,101
in the Amazon RDS using
Amazon Aurora database

32
00:01:47,090 --> 00:01:50,028
which again is a MySQL compatible database

33
00:01:50,028 --> 00:01:54,195
but gives us a much greater
performance over stock MySQL.

34
00:01:55,058 --> 00:01:58,229
We also have our Finance
and Accounting team is using

35
00:01:58,229 --> 00:02:00,977
SQL Server on Amazon RDS

36
00:02:00,977 --> 00:02:04,942
to track how much we spend
on marketing campaigns,

37
00:02:04,942 --> 00:02:08,657
how much we spend to
produce video content.

38
00:02:08,657 --> 00:02:11,209
You can see that we have data stored

39
00:02:11,209 --> 00:02:13,431
in a lot of different areas.

40
00:02:13,431 --> 00:02:16,362
Each area, each particular database,

41
00:02:16,362 --> 00:02:20,815
whether it be RDS Amazon Aurora
or SQL Server or DynamoDB,

42
00:02:20,815 --> 00:02:23,519
they are serving a very
particular function

43
00:02:23,519 --> 00:02:25,306
but it still means that we have data

44
00:02:25,306 --> 00:02:27,359
in a lot of different areas.

45
00:02:27,359 --> 00:02:30,829
We're going to use AWS Data Pipeline

46
00:02:30,829 --> 00:02:34,329
to run queries on these various databases.

47
00:02:36,207 --> 00:02:39,720
We're going to pull in
information about user accounts.

48
00:02:39,720 --> 00:02:42,369
We're going to pull in
information about videos.

49
00:02:42,369 --> 00:02:44,605
We're going to pull in information

50
00:02:44,605 --> 00:02:47,642
about all of our financials.

51
00:02:47,642 --> 00:02:49,571
We're going to aggregate all of that data

52
00:02:49,571 --> 00:02:53,071
and store it over here in Amazon Redshift.

53
00:02:54,117 --> 00:02:57,688
We're also going to take the data

54
00:02:57,688 --> 00:02:59,948
from these Sliding Windows.

55
00:02:59,948 --> 00:03:02,578
We're going to leverage AWS Data Pipeline

56
00:03:02,578 --> 00:03:05,946
to pull data from DynamoDB and transfer

57
00:03:05,946 --> 00:03:08,884
it over to Redshift as well.

58
00:03:08,884 --> 00:03:12,817
From here, now we can use our
business intelligence software

59
00:03:12,817 --> 00:03:16,984
to run much more complex queries
across the entire dataset

60
00:03:17,862 --> 00:03:20,705
that would have normally
been very difficult

61
00:03:20,705 --> 00:03:22,450
because they were in all of these

62
00:03:22,450 --> 00:03:25,095
various disparate
databases but now that all

63
00:03:25,095 --> 00:03:27,927
of that data is in one cluster

64
00:03:27,927 --> 00:03:31,249
we can leverage the massive
parallel processing power

65
00:03:31,249 --> 00:03:35,416
of Amazon Redshift to answer
some very complex questions.

66
00:03:36,421 --> 00:03:40,891
We can also, if we wanted
to retroactively go back

67
00:03:40,891 --> 00:03:43,899
and ask questions that we weren't asking

68
00:03:43,899 --> 00:03:46,319
at the time the data was coming in,

69
00:03:46,319 --> 00:03:50,895
we can go back, pull these
Raw Log files out of S3,

70
00:03:50,895 --> 00:03:54,228
load them into Amazon Elastic MapReduce,

71
00:03:55,664 --> 00:03:58,417
leveraging the power of thousands

72
00:03:58,417 --> 00:04:02,334
of nodes running in
parallel to further process

73
00:04:03,795 --> 00:04:07,962
those logs and store those
results also in Redshift.

74
00:04:08,984 --> 00:04:12,067
With all of this data being pulled in

75
00:04:13,810 --> 00:04:15,808
and aggregated into Redshift,

76
00:04:15,808 --> 00:04:19,278
then we can start to get a
sense of what's happening

77
00:04:19,278 --> 00:04:20,517
in the big picture.

78
00:04:20,517 --> 00:04:23,136
We want to start to understand things

79
00:04:23,136 --> 00:04:26,205
like user engagement per video.

80
00:04:26,205 --> 00:04:30,372
We want to understand how well
a particular series is doing

81
00:04:32,174 --> 00:04:33,658
versus other series,

82
00:04:33,658 --> 00:04:36,370
how much are people interacting with,

83
00:04:36,370 --> 00:04:38,967
how much of a video are people watching.

84
00:04:38,967 --> 00:04:42,327
We want to see user
engagement again per video,

85
00:04:42,327 --> 00:04:45,561
per series, per category, and so on.

86
00:04:45,561 --> 00:04:48,727
We want to understand out conversion rate,

87
00:04:48,727 --> 00:04:51,474
as we're spending money
on marketing campaigns.

88
00:04:51,474 --> 00:04:54,004
People are coming back
to the site as a result

89
00:04:54,004 --> 00:04:56,907
but how many of those people come back?

90
00:04:56,907 --> 00:05:00,549
How many of the people that come back,

91
00:05:00,549 --> 00:05:03,285
how many of those
actually become customers?

92
00:05:03,285 --> 00:05:05,818
What are our conversion rates?

93
00:05:05,818 --> 00:05:09,387
We can answer those questions
given all of this data

94
00:05:09,387 --> 00:05:12,768
from our user accounts,
given the data that we spend

95
00:05:12,768 --> 00:05:15,786
on marketing we can
answer those questions.

96
00:05:15,786 --> 00:05:18,958
We can also figure out how much we spend

97
00:05:18,958 --> 00:05:20,420
per customer acquisition.

98
00:05:20,420 --> 00:05:25,370
How much does it actually cost
us to acquire new customers?

99
00:05:25,370 --> 00:05:28,453
We also can look at our user accounts

100
00:05:29,873 --> 00:05:32,623
and the money that we're spending

101
00:05:33,591 --> 00:05:36,383
and determine what are
our user churn rates.

102
00:05:36,383 --> 00:05:39,324
This month versus last month,
how many users signed up

103
00:05:39,324 --> 00:05:42,494
versus how many users
canceled their account.

104
00:05:42,494 --> 00:05:45,586
Again, you can see this
is just one more example

105
00:05:45,586 --> 00:05:48,985
of the many, many ways
that we can possibly

106
00:05:48,985 --> 00:05:52,077
put all of these various services to use

107
00:05:52,077 --> 00:05:55,868
to analyze a very large amount of data,

108
00:05:55,868 --> 00:05:59,118
very easily within Amazon Web Services.

109
00:06:00,640 --> 00:06:04,807
That is just one example of
video subscription service

110
00:06:06,189 --> 00:06:09,565
business intelligence
leveraging multiple services

111
00:06:09,565 --> 00:06:11,482
in Amazon Web Services.