Download "Роман Бунин, Яндекс Go - Развитие BI-системы компании с помощью продуктовых подходов"
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FlIQgP86zWJo%252Fmaxresdefault.jpg&w=1139&q=75)
Please wait. We're preparing links for easy ad-free video watching and downloading.
Similar videos from our catalog
%3Aformat(webp)%2Fi.ytimg.com%252Fvi_webp%252F-SpQucAJ3A8%252Fhqdefault.webp&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi_webp%252FQwdADqIB__o%252Fmaxresdefault.webp&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi_webp%252FVFmPoOYTEs8%252Fmaxresdefault.webp&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi_webp%252FW831PpzOn_E%252Fmaxresdefault.webp&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FYmBhAn-xqHc%252Fhqdefault.jpg%253Fsqp%253D-oaymwEmCOADEOgC8quKqQMa8AEB-AG-AoACtAGKAgwIABABGGUgXShPMA8%253D%2526rs%253DAOn4CLAxfR_ltmaJ55HhYP4iufL8fDiBcg&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi_webp%252F_ORZxe8gS-Y%252Fmaxresdefault.webp&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi_webp%252FiAbIGS7H5g4%252Fmaxresdefault.webp&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FpDFRU-TaL9A%252Fmaxresdefault.jpg&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FrZPxvW_nxic%252Fmaxresdefault.jpg&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FShwWmVUA7Zo%252Fmaxresdefault.jpg&w=1139&q=75)
Video tags
Subtitles
Russian
00:00:01
[music]00:00:11
[music]00:00:14
colleagues hello my name is Roma Bunin and00:00:17
today I’ll tell you about how we00:00:19
are developing our system as a product of the00:00:21
company Yandex google Yandex head this is00:00:24
everything that is related to00:00:27
the logistics of the city this is a taxi drive and yes and a00:00:30
shop00:00:31
I really believe that visualization00:00:34
ebay data are very important tools for any00:00:38
company because although we live00:00:41
in the age of ronin-guy machines,00:00:43
most of the decisions are00:00:46
actually still made on00:00:48
dash boards on some of the solid00:00:50
plates and so on, so00:00:52
biosystems on my opinion, sucking is00:00:55
one of the most00:00:56
important components of practical or00:00:59
decision-making is the real steering wheel00:01:01
that everyone uses when managing00:01:03
a company,00:01:05
we use ourselves with a sleep board when the00:01:07
system is not only on it I have worked with you for00:01:10
a long time and worked as a00:01:13
report developer and as a00:01:15
biosystems architect and Now I lead the00:01:18
visualization team in the google index I also run00:01:21
a blog and a river channel for data if you are00:01:23
interested in reading about the design of dashboards00:01:25
ebay join00:01:28
what will my report be about today00:01:31
we will report from the series and what about you I00:01:35
will talk about how we are actually00:01:37
developing our system and I’ll tell you what00:01:41
approaches we use for this, well, I myself00:01:43
love this type of report, so I00:01:45
decided to do it in this style. I’ll tell you00:01:47
exactly what we do, I’ll show all our00:01:49
various internal things and what’s most00:01:52
important, I’ll show what we can do and what does00:01:55
n’t work, so I think this story00:01:58
will be It’s useful for everyone to drive00:02:00
analytics and ultimately creates some kind of00:02:02
reporting that, for the company,00:02:07
let’s start by simply00:02:09
telling you how analytics works for us in00:02:11
Yandex Goal. In general, we have quite a lot of00:02:14
analysts, or those people who00:02:16
do something close to analytics00:02:18
enamel engineers and so on these people in00:02:20
my world are superman and they know how to do00:02:24
everything they can prepare their own00:02:25
data they can write python scripts00:02:29
that process that data and build00:02:31
some kind of models and these guys00:02:33
do it from boards, that is, full stack00:02:37
analysts who can prepare00:02:39
the data and close the whole cycle, and also00:02:42
quite a lot of users who00:02:44
actually look at this reporting and00:02:45
use the results that the00:02:47
guys get,00:02:49
naturally, these supermen, well, they can’t00:02:51
just physically be spokes in each00:02:54
of the areas, so there were a00:02:55
structural team which at each00:02:57
stage help them00:03:00
do their job better, there are the guys from game00:03:03
pe who help make cool00:03:05
showcases or somehow set up these00:03:07
processes, there are teams that are responsible00:03:10
for developing internal Python00:03:12
libraries that work great with00:03:14
our new technologies, and there is my00:03:16
team which It helps the00:03:18
analyst to do great reporting. The00:03:21
task of my team is to create00:03:24
standards for China from assembly that can00:03:27
speed up the work on ticks, we also00:03:30
train and support both users00:03:32
and developers, and we also put things in00:03:35
order somewhere, the volume of our analysts on00:03:37
hand so that they don’t create00:03:39
dirt on our server and we also make the00:03:42
biggest reports that00:03:45
I use as a company, and in general, we hand one of00:03:47
our biosystems as a product, what does00:03:49
this even mean, why is the product00:03:53
where it comes from, why is it, well, of course, first of all,00:03:56
just a little tribute to the00:03:58
hype engine that is now00:04:00
around product development of products00:04:02
and so on, well, if you’re not joking, but00:04:06
seriously, I just think that00:04:08
any approaches are systematized and they00:04:12
allow you to do your job better, let00:04:14
it be called Lina Giles, it doesn’t00:04:16
matter, all this is actually common00:04:18
sense, it’s just that it’s systematized and00:04:20
applied on a regular basis Basically, for00:04:23
me, approaches in product development00:04:26
converge in three broad directions. The00:04:29
first is that00:04:33
in order to develop a system as a00:04:36
product, you need to live iteratively and00:04:39
conduct research for this, that is,00:04:41
you can’t do it blindly, we develop00:04:43
our product with the help of some iterations00:04:46
we look at what happens and adjust00:04:48
these actions along the way,00:04:49
secondly, that in order to manage and the00:04:52
product we need some metrics so that00:04:54
we can calmly understand00:04:56
whether we are going in the right direction or not, that00:04:58
is, we need to somehow measure our system00:05:00
and thirdly, we use00:05:03
structuring as a product, we00:05:05
use division into streams, we have a00:05:09
dedicated product direction for each00:05:11
piece of our product, there is a separate00:05:14
backlog, we can manage this00:05:16
structure exactly as a product00:05:17
structure, we can go through each of them00:05:20
and improve either one of them or00:05:22
improve them all together and understand how00:05:24
this affects the product began to00:05:27
give let's start with the first direction, I00:05:30
'll talk about how we use00:05:32
iteration research and here I probably00:05:35
took a step back I'll talk about the fact that we00:05:37
appeared as a company as a whole not so long ago00:05:39
a year ago and the first thing we did was we00:05:41
developed a mission mission for It00:05:43
was very important for our team that we00:05:44
understand exactly what tasks we will00:05:47
solve and we can, based on this00:05:49
mission,00:05:51
structure our work and00:05:52
prioritize it00:05:54
as a whole; we check the mission every00:05:57
six months and sit down to see how it00:05:59
corresponds to current goals. Well, at the00:06:01
last moment we realized what we00:06:02
need, for example In addition to Messi, we also have00:06:05
such an understanding of basic stability,00:06:07
what our product should do exactly,00:06:10
what we cannot do without, and these are00:06:12
the expectations that users have00:06:14
without relatively in general, products in general,00:06:19
and00:06:20
we will further transform this mission into00:06:24
this mission and further transform it into goals00:06:26
that is, we set some kind of big goal00:06:29
that we have in the form of a mission and00:06:31
then we divide it into different goals,00:06:33
we build this tree of goals every00:06:36
six months and then from this tree of00:06:38
goals we look at how we can understand these goals,00:06:41
have we achieved them or no, and00:06:43
we form it with our family, but here on the slide of the00:06:45
tree of goals that we had a year00:06:47
ago and here on the slide is the00:06:50
tree of goals that we have now,00:06:52
when we built this tree of goals, we00:06:55
conduct an expert assessment of which of00:06:58
these tasks we will do00:07:00
first and00:07:01
just with a simple brin curtain lingam we hold00:07:06
this session and decide what is important for us00:07:08
to do first of all,00:07:10
in addition to the goals themselves, here, probably00:07:12
now this year we decided to00:07:14
organize a process and regular ones are00:07:16
marked violet nicks, this is something that we00:07:19
want to do on a regular basis in00:07:22
order so that our new system is00:07:24
in a stable state,00:07:28
this tree of goals is then formed by a00:07:30
classic road map and here is an00:07:33
example of our kind, but of course we don’t00:07:35
always have time to do tasks, something is00:07:37
moving, something is moving, what do you mean, we00:07:39
understand what you don’t need to do, it’s completely00:07:41
normal the usual development of this00:07:44
kind, but these problems go further into00:07:46
the sprint and we work as individual00:07:48
students, there is no magic here, there is simply a00:07:51
Nevsky combo board on which00:07:53
everything is already detailed, and yet they are visualizing the00:07:57
problems, mother of the problems we just take and00:07:59
do some kind of magic here, probably you00:08:01
don’t have tons of image I want to show that such a00:08:03
classic quite00:08:06
process of developing products of anyone can be00:08:09
used and for example for the system,00:08:11
in addition to again forming these cases00:08:14
tree of goals, we also conduct00:08:15
research, but we are researching, we call it a00:08:19
custom journey map, we actually conduct them00:08:21
not directly then the methodologies00:08:23
that are supposed under this00:08:25
name, we conduct an in-depth00:08:28
interview and also look at00:08:30
usage statistics, how the reports are used, and00:08:32
just go to physically see00:08:35
how people use them to see what00:08:37
problems they face from such00:08:40
research, we get ideas on how00:08:43
something can be improved here, for example, on the00:08:45
slide presented00:08:48
our average is how one of the roles of00:08:52
employees working with our reporting is00:08:55
indicated where they receive this00:08:57
information, what problems they00:08:58
face and what ideas from this we00:09:01
can put into our goal tree in order to00:09:03
somehow work with them00:09:06
in general, this is true, everything I can00:09:08
tell you about iteration research, it’s00:09:11
really quite a classic00:09:13
approach of 1 six months, we work with00:09:15
conditional releases once every six months, we00:09:17
update our goal tree and from it they00:09:20
already go for sensors and so on,00:09:22
how do we understand that we are doing this tree of00:09:26
goals for a reason, she will leave and00:09:28
take us to the right one direction for this00:09:30
we use metrics, we have two00:09:34
key metrics and quite a large00:09:36
number of other auxiliary metrics00:09:38
that we also use, but 2 metrics are the00:09:40
most important one metric for us00:09:42
answers the question of the quality of our00:09:44
product the other on quantitative00:09:46
metrics how it is used how00:09:48
actively to measure the quality00:09:51
we use surveys once every six months we00:09:54
create a questionnaire that does not change and00:09:58
send it to all our users00:10:00
who we have on the server they00:10:02
answer it from this we get the00:10:04
results here is a fairly classic00:10:06
questionnaire we simply have on a 10-point00:10:08
scale assessment of the general condition as00:10:11
how much do you like the benefits of reporting and00:10:13
also there is a breakdown into more00:10:16
detailed layers, how convenient it is for them00:10:19
to work in terms of speed, design,00:10:21
visual part, and so on,00:10:23
we do this and aggregate everything and get00:10:26
some results, here on the slide the00:10:30
change over the last six months is presented as we have00:10:32
two just worked and it’s clear that in general00:10:36
we are increasing our score, we’ve grown from 6 to 200:10:40
families, 2 in this half year, which I’m00:10:43
certainly happy about, but here you can also see where00:10:45
we need to strive, it’s clear that there are00:10:47
quite a lot of users who are00:10:49
uncomfortable with some of the parts, for example,00:10:52
of course Unfortunately,00:10:53
performance still suffers and such things are00:10:56
highlighted to us and we understand00:10:59
what is worth paying attention to before00:11:01
building our goal tree,00:11:03
here is a cool life hack that we00:11:07
use if you have a user from survey to survey if your00:11:12
score dropped, then this is the one you00:11:14
should use first and00:11:15
foremost and see what happened, that00:11:18
is, the main point and we pay00:11:20
more attention to those who worsened their00:11:23
score and we conduct a more detailed analysis with them,00:11:26
see what went wrong00:11:29
in general, the results of this00:11:31
We use these surveys when preparing a tree of00:11:33
goals when there is a Kobrin storm, when we00:11:35
form00:11:38
a number we measure using00:11:41
statistics the scoreboard server allows you to00:11:44
use00:11:46
track the number of uses of00:11:48
each report and we use it and00:11:50
here we came up with a rather complex00:11:53
synthetic kipiani, maybe even00:11:55
a little too clever with it but It00:11:57
seems to me that it allows us00:11:59
to make decisions quite conveniently on what it00:12:02
consists of. The00:12:04
first thing we did was we divided all the00:12:06
reports that we have as a company into00:12:08
good reports and not so good. A00:12:10
good report is those reports that are00:12:13
certified, that is, they have been audited00:12:15
by my team and we got ticks for all three00:12:18
hearts, that is, the information is shown00:12:22
on the slide, it can be seen that it is quite00:12:24
short, the loading time should be the00:12:27
report must be formatted correctly,00:12:29
it must have metadata and so on,00:12:31
and00:12:33
in fact there are reports that have00:12:34
not passed certification and then they are not00:12:37
taken into account and we how It is00:12:39
very important for the visualization team to ensure that00:12:42
the number of good reports on00:12:44
the server grows and is not used by as00:12:45
many people as possible. For this we00:12:48
use the following. Now we look00:12:50
at00:12:52
how many people view00:12:55
some reports once a day, that is, we00:12:57
learn unique visits as an employee00:13:00
in the report if he logged into this report for 1 day and adds00:13:04
one point for himself,00:13:06
we add this one point and always00:13:10
weigh it by the level of this00:13:12
employee in our hierarchy, it is clear that we will00:13:16
probably have more users,00:13:18
so to speak, line managers00:13:20
who use the reporting, but We00:13:23
believe that the00:13:24
greatest contribution, well, the greatest among00:13:27
all decisions is made naturally00:13:28
at the top of the company and therefore if these00:13:31
employees use the reporting, then their00:13:33
visit should bring us a little more of00:13:35
our00:13:38
views, due to the fact that this00:13:40
review most likely allowed us to take00:13:42
more important ones for the companies management00:13:44
decision, but at the same time there is a balance here00:13:47
if we have a lot of people looking at reports, but there are a00:13:51
lot of them or one person with a00:13:52
lot of weight, they equalize each other00:13:54
and we, that is, here we do not fall into the00:13:57
trap of focusing either00:13:59
only on the top management, or vice versa,00:14:00
only on the most visited reports, that00:14:03
is, such a weighted one allows us to00:14:04
evenly pay our attention to00:14:07
different types of reports, to those that are very00:14:09
important for the company, and to those00:14:11
reports that are simply used by a00:14:13
lot of people and the whole thing is weighed by00:14:16
these days of viewing by a person we divide00:14:19
by the number of employees this is necessary so00:14:21
that ours now simply does not grow with the00:14:23
growth of the company because in general the00:14:25
number of people who have been in parity since March00:14:26
is approximately proportional to00:14:28
the size of the company and here we simply00:14:31
equal this so that ours now will be00:14:34
more pure for ourselves00:14:38
this now we are monitoring on the dashboard,00:14:42
we have made a dashboard for ourselves00:14:44
on which we monitor what is00:14:46
happening with our server, here there is a00:14:49
main part that is actually00:14:51
dedicated to this, now the black00:14:54
lines show the value that we00:14:57
have for our China at the moment, the00:14:59
blue lines show theoretically00:15:01
possible value, that is, these are all the00:15:03
reports that are on the server, how much00:15:06
does it give them, so ours, now we have indicated00:15:08
our goals and some planned00:15:10
value is00:15:11
very convenient and I really enjoy this thing00:15:14
while you rest, prominent times we had it00:15:16
hanging on a large monitor and we are for00:15:19
we watched this every day with the team, we saw00:15:20
what was happening, often we just get messages in00:15:23
the form of messages and each messenger is also00:15:25
quite convenient, we just look at what is00:15:27
happening in00:15:28
addition to this main thing, now00:15:30
we measure our magic parrots,00:15:32
we have a breakdown of this now which, in00:15:34
general, is a little bit more human readable00:15:37
and you can understand what is happening, that00:15:40
is, we are just looking at the number of00:15:41
unique views, this indicates00:15:43
the number of active users00:15:44
that we have on the server, the number of00:15:47
books, and so on, the whole breakdown of our00:15:51
boil is there and you can see00:15:53
exactly00:15:54
what influenced it that ours has00:15:58
now fallen or increased, and right here00:16:01
I want to tell you that it seems to me that00:16:03
this metric has turned out to be quite convenient00:16:05
because it allows us in00:16:06
completely different ways and to increase, for00:16:09
example, we can increase the number of00:16:11
certified nicknames on the server, just00:16:13
my team can go through the reports and improve them,00:16:16
or we can simply attract00:16:19
users to those reports that we have already00:16:21
made good, that is, we can do00:16:23
internal promotions, do mailings,00:16:25
attract new users00:16:27
who, for example, in this report,00:16:29
simply did not know yet, or we can, for00:16:31
example, train our analysts so that00:16:33
they immediately We did great reporting and00:16:35
used our stlg.00:16:37
In addition to the main metrics, we also have00:16:40
a lot of auxiliary history,00:16:41
which we also monitor, the manager monitors the00:16:44
average loading time and then00:16:46
how often does a crash occur in00:16:48
data updating, if it happens00:16:49
often there and we notice it and see what what00:16:53
happens with this00:16:54
in general, I repeat that00:16:57
I really like this approach to monitoring00:16:59
and I really enjoy it every day when I get the00:17:02
results of this report, it seems that it00:17:05
allows you to take it very conveniently, that is,00:17:07
not like a hand, if it helps more00:17:09
to understand what to do tomorrow if anything happens -00:17:11
is going wrong or where to pay attention to the00:17:16
next point or we will move on and00:17:18
here it will be the most00:17:20
massive presentation this is00:17:23
what components we have our biosystems00:17:25
what it consists of and what we do00:17:28
for each of them00:17:29
here they are not located in some00:17:31
order of importance, just a photo, a logical00:17:34
order of this that I will tell you,00:17:36
well, in general, and it is not being developed00:17:40
in approximately the same order as the00:17:44
first component is not indicated in which we will00:17:46
talk, there was a guide on it, the00:17:48
plate of the post came to the company, it was00:17:50
clear what needed to be developed the visual00:17:52
part of our reports because many of the00:17:56
reports were made very high quality in00:17:57
terms of graphic design and00:17:59
ease of use,00:18:00
in order to fix this we decided to make a stl00:18:02
guide;00:18:04
in general, it was definitely needed simply00:18:08
because we have a lot of reports on the server,00:18:11
we really have quite a large00:18:12
number of people with desktop licenses00:18:15
that can create reports and they00:18:18
publish reports every day, literally00:18:19
every day, new reports appear on our server00:18:21
that must00:18:23
meet our internal00:18:24
quality criteria so that it would be00:18:27
easier for an analyst to do this, we decided to do just that,00:18:29
there was a guide that would allow them to00:18:32
do this quickly and without problems, that is,00:18:36
here it is also an interesting point that I00:18:38
would like to emphasize that here the purpose of00:18:40
creating the guide was not so much to00:18:42
make it easy to use for the00:18:45
end user, although we00:18:46
certainly achieved this goal and the most important thing is00:18:49
that the developer of the reports would be comfortable and00:18:52
quickly do it well It seems to me that00:18:55
this approach allowed us to distribute this thing00:18:56
quite well in the00:18:58
company. I often hear feedback from analysts00:19:00
that this is a very cool00:19:02
product that allows them to quickly00:19:04
solve their problems, that is, if you want to00:19:06
distribute your company,00:19:09
try to provide some kind of guide its00:19:11
convenience is not for business users, but to make it00:19:14
convenient for your developers00:19:16
to use this template style guide and00:19:18
so that it helps them speed up their work,00:19:20
and of course, in addition to these two goals, we00:19:23
also wanted to achieve that our00:19:25
reports were simply cool and simply the00:19:28
best on market are the most beautiful00:19:31
why I think this is an important point,00:19:33
again about reporting developers,00:19:35
that they need help with this part, the fact is00:19:38
that most of the guys are from00:19:40
analytics who work in mystical00:19:43
departments and these are usually technical guys00:19:45
who rarely and rarely00:19:47
encounter any design, if00:19:49
you look at the skills of an analyst00:19:51
who now requires that00:19:52
the device be given comfort, then this is a fairly large stack of00:19:55
skills and skills that are aimed at00:19:58
completely different00:20:00
[music]00:20:01
areas of knowledge and one of these00:20:04
shoes building data visualization, it00:20:07
seems to me that demanding from an analyst that00:20:08
he is still cool he knew graphic design,00:20:11
information design and x00:20:13
telling stories and was a professional in the taiga00:20:15
system he works with, but this is just too00:20:17
much and therefore he became needed00:20:21
primarily to help00:20:23
develop these skills or so that he would00:20:25
level out the lack of these skills00:20:27
or their smallness quantity therefore00:20:30
I think that such things are really really00:20:32
very necessary because to meet a00:20:34
person who knows all this and00:20:36
knows everything is super cool it’s like00:20:38
meeting a00:20:39
unicorn00:20:41
our lie consists of three large00:20:43
components 1 component of a presentation with00:20:46
recommendations is quite a classic00:20:48
I think this is exactly what Usually00:20:50
people understand that the guide is a big00:20:52
presentation where it is written what and how it is00:20:54
done, the second part is instructions with a00:20:57
video on how to use it, it’s00:20:59
just educational, and 3, this book is a template00:21:02
that is actually the heart of the00:21:04
guide, which allows you to00:21:06
quickly make new reports from this template,00:21:09
and so do we we train and conduct00:21:10
consultations because how to use it00:21:13
self-presentation shoots it’s just a00:21:16
large set of slides of 60 pieces and more is00:21:18
probably not aimed at developers;00:21:20
we have separate reporting for my team00:21:21
so that we understand what standards the00:21:26
dashboard developers themselves are currently working on. I think this00:21:28
presentation is even possible haven’t seen the00:21:31
second large part, this is on our00:21:34
internal page, there is00:21:36
a section dedicated to how to use the00:21:39
style guide, here there is a video on how you00:21:41
can use it, here there are examples,00:21:43
frequently asked questions, and so on, that00:21:46
is, we are trying to get into the topic of00:21:48
cool use of teamplay, it was00:21:51
as simple as possible00:21:52
Well, the third thing is that the book itself was compiled as a00:21:55
guide, or rather the template itself for which it is00:21:59
used, this is the most important component of00:22:01
our guide, now I will show it00:22:03
live, but I also have a report00:22:05
about it in some detail, if you want to know00:22:06
how it works technically inside, then00:22:09
follow the link00:22:11
yourself the book is a template, it consists of00:22:18
how to fry, okay,00:22:23
we have a technical overlay here,00:22:25
let’s do this and I won’t show it to him00:22:27
and let’s move on, maybe00:22:30
we’ll correct it a little because it’s impossible, but we00:22:31
still need to show the screen further and we need00:22:34
it to work, then I00:22:35
’ll come back to this and I’ll show you this book00:22:37
at the same time as it,00:22:42
in general, there is a book that allows you to00:22:45
quickly understand the reporting and quickly00:22:47
and at the same time00:22:52
make reports like this, for00:22:55
example, you can make a report like this, and00:22:57
what’s most important is that it’s cool, this00:22:59
reporting is done very quickly, that is, to00:23:00
develop such a report00:23:02
can actually be done in 15-20 minutes if the data is prepared, you00:23:05
can achieve such a pretty00:23:07
cool result. The00:23:10
next big section that we00:23:11
pay a lot of attention to is training,00:23:13
training our developers, I00:23:17
have developed three programs that I am currently00:23:18
running, we have a part on the basics of the00:23:20
scoreboard, there is a large part about00:23:23
data visualization and layout of dashboards00:23:25
and there are lessons about advanced functionality00:23:28
in general, it’s almost like such00:23:31
a system before00:23:33
professional training will scratch for 7000:23:34
academic hours00:23:36
on the basics of the board, classes are held 1 200:23:39
times a year, for the rest we00:23:41
conduct classes once a year, but00:23:43
of course there are recordings like but as an00:23:45
analyst, they are also a body of water to look at, we00:23:48
use conditional gamification, if00:23:51
our analyst goes through our training, then00:23:54
he receives these badges, we00:23:56
called our school the Jedi scoreboard school,00:23:58
and depending on the amount of00:24:00
training you have completed and what00:24:02
projects you have done, you can receive different badges00:24:06
for For our internal system of00:24:08
achievements,00:24:12
in addition to the visual part, it is also important to00:24:14
upgrade the technical part, it is00:24:17
also important to upgrade the logical00:24:19
structural part, so the next00:24:21
point is that we also often actively00:24:23
distribute this algorithm for creating00:24:25
a dashboard, and I called it the Kansas dashboard and00:24:27
this thing is similar to the canvas business model00:24:30
if here I heard with her00:24:33
this algorithm that allows you to00:24:36
structurally approach the very process of00:24:39
developing a dashboard aek, mainly to00:24:41
collecting requirements from the customer’s business, that00:24:43
is, here analysts using this00:24:45
form can collect all the data and00:24:49
all the water that is needed00:24:51
to make even the board cool00:24:53
in our own understanding, we go through00:24:55
it, it looks like it looked like a00:24:56
graphical form like on a slide, or00:24:59
we also just use it in paragraphs in the00:25:01
keta when these are just the headings of each00:25:04
paragraph, you can also read more in detail on00:25:06
my blog, while you00:25:10
and I have now talked about what we00:25:12
have been doing with our developers now I want to00:25:14
show what we do for our00:25:16
users, the00:25:17
first important component is our00:25:20
vision of Wikipedia and the system on warding,00:25:22
which allows us00:25:25
to look, then allows our Polish to00:25:28
quickly understand where, what, know and how to00:25:31
use it, we fly out this story in this00:25:33
way and now I’ll try to show00:25:35
again how it is it looks alive,00:25:55
it’s a cup of00:25:59
flour,00:26:01
okay, unfortunately, a little bit of a piece is00:26:04
actually missing, I’ll try it00:26:06
somehow separately, maybe there will be an00:26:08
indicator later,00:26:11
in general, we have an internal page00:26:13
that is divided and also under control00:26:15
just unidirectionally, part for00:26:17
reporting users and the other for00:26:19
developers and just for the user00:26:22
we have it boarding which is just a00:26:24
page with short videos that00:26:27
are recorded and about how to use00:26:28
the reporting like a finger server and how00:26:31
to get the access you need, we are00:26:34
now sending this thing out to all our colleagues00:26:36
who are new to us, come to work and00:26:39
When they receive a license to view, they00:26:41
also receive a letter with such a00:26:43
brief note about how to use the00:26:47
server. The00:26:50
next important story that00:26:53
we are developing is the issue of access, the fact is00:26:56
that00:26:58
we can’t just give access to everyone, after00:27:02
all, the gdp system stores quite00:27:04
a lot of all sorts of information. which00:27:07
can somehow be used incorrectly and according to00:27:09
us, it is therefore very important for00:27:12
business to delimit access for00:27:15
security, while you never want00:27:18
these accesses to limit business00:27:20
benefits for them to be an00:27:23
obstacle to00:27:26
employees receiving real information that00:27:28
they need in their work00:27:29
earlier the idea was to use the following00:27:32
system, we had folders that were00:27:35
tied to each direction, that is,00:27:37
there was a00:27:39
folder for each direction that was different,00:27:42
and this inside the folder there were four00:27:44
levels of access:00:27:46
public, commission, secret, and the most00:27:49
secret, and actually00:27:52
why did we find such a system00:27:54
inconvenient why? We decided00:27:56
to change the fact that, surprisingly,00:27:59
employees from one case very often00:28:02
look at the reports of other departments and this is00:28:05
super cool, which means that we have00:28:06
some kind of data exchange, and it00:28:08
turned out that by giving access to a00:28:10
parallel department, a person got00:28:12
access to all reports which are there00:28:15
and this is not so cool anymore, so00:28:17
we have now replaced this system and we00:28:19
are trying to make access not by departments but00:28:23
by the roles that people need, that is,00:28:26
if we have some role, I don’t know00:28:28
there is a regional manager, then this is the00:28:30
regional manager king gets00:28:31
access not only to the folder dedicated to the00:28:33
regions, but gets to a set of reports00:28:36
that are needed for his work. How do we00:28:41
determine this set of reports that00:28:43
he needs to work? We do this00:28:45
using our internal00:28:47
analytics system for portal visits, we00:28:51
have a morning parity that allows00:28:52
us track how each department00:28:55
uses what reporting, if here00:28:57
in this report, for example, you select a00:28:59
specific service, then for this service you00:29:01
and I have this reporting00:29:03
that is used for00:29:06
those who do not use it, so by analyzing00:29:09
the history we understand that00:29:11
these reports were needed for this service and then we will00:29:16
immediately give new employees access00:29:18
to all these reports to the person who comes to this service, regardless of00:29:20
what folders they are in.00:29:23
What should we do if an employee needs00:29:26
some new report, a new report has appeared00:29:27
or something else for this we We are00:29:29
monitoring this and we have created a system of00:29:31
individual accesses, that is, now00:29:33
if you need access to someone, you00:29:35
just write it down through our00:29:37
internal system, specific access and00:29:39
why do you get it and then we00:29:41
track all these accesses, we also follow a00:29:44
special report com this is the case and00:29:47
when a fairly large00:29:49
number of individual accesses to a00:29:51
certain report are collected, these detailed accesses are00:29:53
formed at what price they relate to a00:29:55
certain business, then we simply turn00:29:58
this into a group role and this report now00:30:00
becomes available by default to all00:30:02
those who work in this department00:30:05
on my I think it’s a pretty successful system;00:30:08
it’s a little more complicated to manage,00:30:09
but at the same time it gives you greater flexibility, and00:30:12
of course our employees now, when they go00:30:15
to work, receive simply a minimum00:30:17
threshold of resistance in the sense that they00:30:19
immediately receive, on the first day of work, the00:30:22
maximum amount of information that00:30:24
will be useful to them: an00:30:27
important part that is00:30:30
at the interface of our developers who used the00:30:32
body is content management, we believe00:30:35
that our server on which all00:30:37
reporting is stored is a marketplace in general, we00:30:40
have content producers, analytics,00:30:42
we have business users who00:30:43
use this content, and here, like00:30:46
any new management system, In the York world with00:30:48
this content, we must check that00:30:51
our content is of high quality, that it is in the00:30:53
correct form, and so on, so we have00:30:56
now introduced a00:30:57
review system for all reports published on the server;00:31:00
at this moment we00:31:02
check that these reports are of high quality,00:31:03
load quickly, and the like with00:31:06
each publication on crying about the fact that a00:31:08
ticket falls that is less than my team and we00:31:12
sit down with the guys for 1 week just00:31:14
warming up this story, looking at what’s00:31:16
new on the server and actually00:31:18
what kind of reporting we can00:31:21
certify and what kind of reporting needs00:31:23
some work in order for it to00:31:24
turn out cool and00:31:26
here there is an example of exactly the report00:31:29
that we look at, each00:31:31
column is just a different00:31:33
certification criteria, this report that receives00:31:36
all the points, it becomes00:31:37
certified by it, then we00:31:39
close the ticket and00:31:41
continue to work, but if we00:31:45
need any additions, we simply00:31:48
write like once an employee said that it’s00:31:49
better to correct the name here, it’s better to00:31:52
make a different distance, it’s better00:31:53
to use the organization, and so on,00:31:56
in fact, for me personally,00:31:58
they told me this kind of command, it was just a00:32:00
discovery, this process allowed you to understand00:32:03
what’s happening on the server, you understand00:32:05
how much yours is used became a00:32:07
guide and so on, and now you00:32:10
go through it manually, look through the reports,00:32:11
they are published, this is a super useful story,00:32:14
well, it’s just involved in researching00:32:16
exactly what happens physically with the00:32:19
product, okay,00:32:22
about the external part,00:32:24
we talked to you about some part about searching, but it’s a common00:32:26
problem with any biosystem the fact that00:32:28
we have a dictionary of metrics, that the metrics in00:32:31
the calculations are considered somehow wrong, since00:32:33
analysts00:32:36
can calculate the same trick in two different reports using00:32:38
different formulas and00:32:40
this, accordingly, affects how they00:32:44
finger the business policy I can’t what -it’s a00:32:46
mistake to make because of this, in order to00:32:49
level this thing out, we have now developed00:32:51
the words of metrics, this is such a huge plate,00:32:54
our p pointee, which can be00:32:56
edited by those people who understand00:32:58
how this metric is calculated, that is, here00:33:00
we have a dictionary of metrics for one hundred and sixty00:33:03
plus lines, each line is this metric00:33:06
and we are now gradually describing how00:33:09
each metric should be considered, what00:33:11
it consists of, and so on,00:33:14
while this is a very useful tool00:33:16
not only for us as00:33:19
content holders, but even before the00:33:21
analysts themselves, many analysts now00:33:23
look at how some metric is calculated00:33:25
such oooh, it should have been considered differently,00:33:28
how this thing sprouts for our00:33:31
users for users, we00:33:33
replicate this Shirk pointer sign, we put together a00:33:36
plan of 2 mpla, we00:33:37
pull it into the scoreboard and it is built into00:33:40
our style guide became a guide it looks00:33:42
like such a button and at the top dashboard00:33:47
which stores all the information, if you00:33:49
click on it you can see the metrics00:33:51
that are used in this dashboard, that00:33:53
is, these metrics when designing00:33:55
dashboards, the analyst should click to00:33:57
say that in my dashboard these are the00:33:59
metrics used and a00:34:01
description of them and calculation00:34:03
formulas will automatically appear and the analyst himself will appear here00:34:05
checks himself that he considers it to be the correct00:34:06
formula, and the business user can00:34:08
actually see what exactly was00:34:10
meant by a particular metric,00:34:16
we also support our00:34:18
users, both developers and00:34:20
business users, we do this00:34:22
using the messenger, we have two00:34:23
channels 20 where we We answer questions,00:34:27
we don’t have any strict goals for00:34:29
response time, we don’t create tickets, but00:34:31
it seems that Chad works best in this sense,00:34:33
we also track them with this00:34:35
activity, see how many people write,00:34:38
how often we answer here, we00:34:40
use the ratio from this list00:34:42
the number of messages from users and00:34:44
this Minsk team from the00:34:46
support team, in general, there’s00:34:49
probably no magic here, but I want to tell you00:34:52
that in this sense, it’s the00:34:54
chateau themselves that save us and it seems that for all00:34:57
users this is the easiest way00:34:58
to get us some kind of approval help00:35:01
and feedback and therefore we think that00:35:04
Chad is a very successful solution for this, although of00:35:07
course sometimes there is a lack of some kind of00:35:08
structure as it is, we worked00:35:11
there in trickles, but00:35:12
I think that compensation for speed and00:35:15
convenience for users is more important here than the00:35:18
next hour, which I can also00:35:21
pay attention to is communications we00:35:23
need to tell our users about00:35:25
what happened, we now have a guide, we have a00:35:27
new report, we have some00:35:29
changes in the data about all these00:35:31
things, we report through different channels, for00:35:33
this we use all channels, everything00:35:37
possible, everything we can we have and00:35:39
simply depending on the conditional00:35:42
occasion that happens to us, we00:35:44
use a suitable channel for this and a00:35:46
suitable format, we have developed our00:35:48
own format for each event, we00:35:52
store it and in this way we simply00:35:54
use them when we need to00:35:55
communicate something if a new one has opened00:35:57
you copy the report what is there, paste it into00:36:00
messenger, send it and all users00:36:03
receive information that a00:36:07
new report has appeared, and so on. I just00:36:09
want to tell you that for example,00:36:11
some of this did not work for us, we at00:36:14
one time planned to use00:36:16
these to you with Jer and which00:36:18
would allow you to directly00:36:21
show00:36:24
answer quickly to user questions00:36:26
some things to carry with such 100:36:28
card, but it turned out that in this00:36:31
story it didn’t take off very much and the people of the clique did00:36:34
n’t receive much, so they have now refused the energy00:36:36
and are not using00:36:39
the latter, but probably not00:36:43
one of the most important plasterers we00:36:46
have infrastructure00:36:47
infrastructure name in mind both the structure of00:36:49
our team and purely technical00:36:51
history00:36:52
in order to manage our system00:36:55
we use the infrastructure that we00:36:57
ourselves develop this information will melt from00:37:00
three large pieces the first is a set of00:37:01
internal reports 2 this00:37:03
robot automation, for example, we have robots00:37:05
that make sure that some00:37:07
reports are not used, they are rarely looked at,00:37:10
such reports are deleted by the server and00:37:14
there is also just technical support for the00:37:16
server itself, we have a rather complex00:37:17
server display configuration which, I00:37:19
admit, in different data centers,00:37:21
this also has to be done maintain it,00:37:23
make sure it doesn’t break and see00:37:25
that it works well, how we00:37:27
use it, for example, we have a lot of00:37:29
reports in our internal almost00:37:31
every area that is supervised by us, we00:37:33
have our own dashboard,00:37:35
which is probably not surprising for the00:37:38
visualization team and00:37:39
based on these dashboards we can,00:37:41
for example, draw cool conclusions,00:37:44
for example, we have a report on connections00:37:47
that shows how users00:37:48
move between samples, that is, if a00:37:50
person in a field with one report then00:37:52
moved to some other one, he moved in the00:37:54
sense of not from this miracle, he just went00:37:56
for 10 minutes and used another00:37:58
report, then we can link these00:38:00
reports from saying that these reports are connected,00:38:02
most likely people use them in00:38:04
some logical sequence,00:38:06
from this we get such00:38:08
connection maps and then use them00:38:10
to create real links, that is,00:38:12
we now have one of the jobs which we are00:38:13
doing right now is creating blood00:38:16
links between readings00:38:17
thanks to this connection diagram,00:38:20
it’s really cool to see that there are clusters of00:38:21
reports that are very connected, there is one00:38:24
most important report, there are clusters of00:38:26
reports that live in general00:38:27
separately and you can see that you have to answer in a00:38:30
separate direction00:38:33
in general -that’s almost all I wanted to00:38:34
tell you today, what conclusions I think00:38:37
you should draw? Well, I personally would like00:38:39
you to start measuring your system00:38:41
if you haven’t done this yet, I suggest00:38:44
doing it first with surveys and secondly with00:38:46
some metrics if you don’t00:38:48
want bother hearing how we,00:38:50
like nismo, started looking for a synthetic00:38:51
metric, then in my opinion, the big left00:38:53
chooser is the most adequate00:38:56
metric for how your reporting is used,00:38:59
it seems that it is worth breaking that system00:39:01
into all those areas that we are currently00:39:03
developing, you can see00:39:06
if you are doing this and make sure that in this00:39:08
direction you have a plan for how you00:39:10
will improve it, well, for example, we are also00:39:13
called further, we want to upgrade,00:39:16
for example, the fact that now it is00:39:18
not well set up yet, we do not have enough00:39:20
regular cleaning processes, we will00:39:22
do this, we just need to00:39:24
go regularly see that everything is clean on the server,00:39:26
that the reports for our son are correct, and so on,00:39:28
we will work on speeding up00:39:31
our reports. There is a problem with this. It’s a cool00:39:33
idea. We now want to make hop reports.00:39:35
These are reports that will00:39:38
distribute our users in00:39:40
some direction and store them. contains00:39:42
basic information in some00:39:44
area and at the same time have links to00:39:47
more detailed reporting for each of00:39:49
them, we want to automate support and00:39:51
create working documentation because the00:39:53
world has a lot of documentation for each report,00:39:55
we haven’t taken off so now we will00:39:58
try to do as I did for you00:39:59
showed with metrics built inside the00:40:01
reporting, well, we’ll try to do it with00:40:04
a service, for example, we have such a classic00:40:06
analytics service for00:40:07
business users, it hasn’t taken off yet, we00:40:09
did a couple of projects on this, but00:40:12
the use of this story00:40:14
usually agrees no, in the end I want to give my00:40:18
team actually my team is quite00:40:20
a little mood, we are working on the00:40:22
visual part of our server and00:40:25
there are also Shevtsov who are the manager of it,00:40:29
namely the infrastructure part and the guys00:40:31
who help him do this and create00:40:33
all sorts of00:40:34
infrastructure things, and our00:40:37
managers who help us with00:40:39
priorities and strategies and guide00:40:42
us in our development00:40:44
that's all for me00:40:47
[music]Description:
Совсем скоро Матемаркетинг-22, 17-18 ноября Подробности: https://matemarketing.ru/ Программа: https://docs.google.com/spreadsheets/d/1Iyi-Wj75gmzC-R9cGk16DOldLhWzVtLUNwZJoTht5xU/edit?usp=sharing Купить билет: https://matemarketing.ru/
Preparing download options
* — If the video is playing in a new tab, go to it, then right-click on the video and select "Save video as..."
** — Link intended for online playback in specialized players
Questions about downloading video
How can I download "Роман Бунин, Яндекс Go - Развитие BI-системы компании с помощью продуктовых подходов" video?
http://unidownloader.com/ website is the best way to download a video or a separate audio track if you want to do without installing programs and extensions.
The UDL Helper extension is a convenient button that is seamlessly integrated into YouTube, Instagram and OK.ru sites for fast content download.
UDL Client program (for Windows) is the most powerful solution that supports more than 900 websites, social networks and video hosting sites, as well as any video quality that is available in the source.
UDL Lite is a really convenient way to access a website from your mobile device. With its help, you can easily download videos directly to your smartphone.
Which format of "Роман Бунин, Яндекс Go - Развитие BI-системы компании с помощью продуктовых подходов" video should I choose?
The best quality formats are FullHD (1080p), 2K (1440p), 4K (2160p) and 8K (4320p). The higher the resolution of your screen, the higher the video quality should be. However, there are other factors to consider: download speed, amount of free space, and device performance during playback.
Why does my computer freeze when loading a "Роман Бунин, Яндекс Go - Развитие BI-системы компании с помощью продуктовых подходов" video?
The browser/computer should not freeze completely! If this happens, please report it with a link to the video. Sometimes videos cannot be downloaded directly in a suitable format, so we have added the ability to convert the file to the desired format. In some cases, this process may actively use computer resources.
How can I download "Роман Бунин, Яндекс Go - Развитие BI-системы компании с помощью продуктовых подходов" video to my phone?
You can download a video to your smartphone using the website or the PWA application UDL Lite. It is also possible to send a download link via QR code using the UDL Helper extension.
How can I download an audio track (music) to MP3 "Роман Бунин, Яндекс Go - Развитие BI-системы компании с помощью продуктовых подходов"?
The most convenient way is to use the UDL Client program, which supports converting video to MP3 format. In some cases, MP3 can also be downloaded through the UDL Helper extension.
How can I save a frame from a video "Роман Бунин, Яндекс Go - Развитие BI-системы компании с помощью продуктовых подходов"?
This feature is available in the UDL Helper extension. Make sure that "Show the video snapshot button" is checked in the settings. A camera icon should appear in the lower right corner of the player to the left of the "Settings" icon. When you click on it, the current frame from the video will be saved to your computer in JPEG format.
What's the price of all this stuff?
It costs nothing. Our services are absolutely free for all users. There are no PRO subscriptions, no restrictions on the number or maximum length of downloaded videos.