WEBVTT

00:00:00.000 --> 00:00:31.270 align:middle line:90%


00:00:31.270 --> 00:00:34.640 align:middle line:84%
The last three years, we
built a data warehouse

00:00:34.640 --> 00:00:37.480 align:middle line:84%
on Hadoop technologies
and big data.

00:00:37.480 --> 00:00:42.160 align:middle line:84%
And then we started doing
the simple analytics,

00:00:42.160 --> 00:00:47.080 align:middle line:84%
basically just
reporting what happened.

00:00:47.080 --> 00:00:50.950 align:middle line:84%
But in those years,
the game matured,

00:00:50.950 --> 00:00:55.450 align:middle line:84%
and now we have a need
to basically understand

00:00:55.450 --> 00:00:58.180 align:middle line:90%
more in-depth analytics.

00:00:58.180 --> 00:01:03.740 align:middle line:84%
So we created dedicated
teams of data scientists

00:01:03.740 --> 00:01:08.580 align:middle line:84%
that are data mining our data,
and they're creating models

00:01:08.580 --> 00:01:12.800 align:middle line:84%
that in the future we hope
will help us to customize

00:01:12.800 --> 00:01:18.010 align:middle line:84%
the offers, to, hopefully
as players stay longer,

00:01:18.010 --> 00:01:21.190 align:middle line:84%
address all of those not
only advanced analytics,

00:01:21.190 --> 00:01:24.370 align:middle line:84%
but also prescriptive
analytics questions.

00:01:24.370 --> 00:01:29.980 align:middle line:84%
We can collect data from
even anything we do in a game

00:01:29.980 --> 00:01:33.130 align:middle line:84%
to obviously your financial
data, your transaction data,

00:01:33.130 --> 00:01:35.870 align:middle line:90%
your chart data.

00:01:35.870 --> 00:01:40.990 align:middle line:84%
And also, we have started
collecting social media data

00:01:40.990 --> 00:01:45.040 align:middle line:84%
on what our players say
and on the social networks.

00:01:45.040 --> 00:01:48.730 align:middle line:84%
The main thing that open
source solutions don't give you

00:01:48.730 --> 00:01:52.300 align:middle line:90%
is scalability.

00:01:52.300 --> 00:01:56.560 align:middle line:84%
Once you're ready to put your
automation of your models

00:01:56.560 --> 00:02:01.200 align:middle line:84%
on a massive scale, the
open source solutions,

00:02:01.200 --> 00:02:06.730 align:middle line:84%
so just take too long
and a lot of manual

00:02:06.730 --> 00:02:11.240 align:middle line:84%
work around them to
put it in production.

00:02:11.240 --> 00:02:13.750 align:middle line:84%
So once we realize
that we're going

00:02:13.750 --> 00:02:17.860 align:middle line:84%
to be running hundreds or
even thousands of models

00:02:17.860 --> 00:02:21.130 align:middle line:84%
for all of our games,
all of our regions,

00:02:21.130 --> 00:02:25.990 align:middle line:84%
all of our time
frames, we started

00:02:25.990 --> 00:02:30.760 align:middle line:84%
looking for the solution that
can make it scalable for us.

00:02:30.760 --> 00:02:35.830 align:middle line:84%
Once our data is prepared and
once our modeling methodology

00:02:35.830 --> 00:02:42.160 align:middle line:84%
is established, multiplying
those hundreds and thousands

00:02:42.160 --> 00:02:48.220 align:middle line:84%
of models becomes a one-person
job, versus probably

00:02:48.220 --> 00:02:51.850 align:middle line:84%
10 to 20 people just manually
doing it and maintaining it.

00:02:51.850 --> 00:02:54.580 align:middle line:90%
And they'll make mistakes.

00:02:54.580 --> 00:02:57.550 align:middle line:84%
Automated production
environments, like SAS,

00:02:57.550 --> 00:02:59.380 align:middle line:90%
does not make mistakes.

00:02:59.380 --> 00:03:02.920 align:middle line:84%
The main feature for us
is the scalability of SAS

00:03:02.920 --> 00:03:05.700 align:middle line:90%
that no one else can offer.

00:03:05.700 --> 00:03:11.851 align:middle line:90%