Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:00,670 --> 00:00:00,960
All right.
2
00:00:00,980 --> 00:00:07,230
It's time to talk about my all time favorite A.I. visual and powered by the decomposition tree.
3
00:00:07,250 --> 00:00:09,410
Now this one still a preview feature.
4
00:00:09,500 --> 00:00:13,760
So you're going to want to make sure you have the latest version of power by desktop installed.
5
00:00:13,880 --> 00:00:20,510
And once you open it up you need to head to your options and settings open up the options window and
6
00:00:20,510 --> 00:00:22,560
then we can do is drill into your preview features.
7
00:00:22,580 --> 00:00:27,410
Just give that checkbox at check where it says decomp position treat visual and press.
8
00:00:27,430 --> 00:00:33,680
OK so you'll probably be prompted to restart power by go ahead and do that and you'll see this nice
9
00:00:33,740 --> 00:00:34,880
little icon here.
10
00:00:34,880 --> 00:00:39,860
Kind of looks like a flowchart or a process diagram with that a light bulb.
11
00:00:39,920 --> 00:00:41,900
That's our decomposition tree.
12
00:00:41,900 --> 00:00:47,360
So if you'd like to follow along with me go ahead and open up the visuals file and we're gonna create
13
00:00:47,390 --> 00:00:53,860
a new page here called D composition tree.
14
00:00:54,220 --> 00:00:59,000
Let's go ahead and give it a click drop in under a canvas resized things a bit.
15
00:00:59,060 --> 00:01:01,690
Looks good now you know me.
16
00:01:01,700 --> 00:01:06,470
Before we start tinkering I always think it's important to talk about the why.
17
00:01:06,500 --> 00:01:08,810
So why are we using this visual.
18
00:01:08,810 --> 00:01:15,530
What's the objective and what I love about the decomposition tree is that it's great for ad hoc data
19
00:01:15,530 --> 00:01:16,800
exploration.
20
00:01:16,850 --> 00:01:22,700
It's a great way to understand how your data is distributed and it's an excellent tool for root cause
21
00:01:22,790 --> 00:01:23,900
analysis.
22
00:01:24,110 --> 00:01:30,260
And what I love about it is that it's smart it's intelligent it's a.i. Driven but it's also really intuitive
23
00:01:30,320 --> 00:01:31,720
and really user friendly.
24
00:01:32,210 --> 00:01:37,370
So let's go ahead and give it a click and you'll see kind of two fields that should look familiar from
25
00:01:37,370 --> 00:01:42,480
our key influencers visual we've got and analyze well and then explain by well.
26
00:01:42,530 --> 00:01:48,770
So it's important to note is that anything we pull in the analyzed field here needs to be a quantitative
27
00:01:49,010 --> 00:01:55,640
metric so we can pull a categorical field in here and project outcome but it's going to get aggregated
28
00:01:55,820 --> 00:01:58,360
either as account or distinct count.
29
00:01:58,370 --> 00:02:05,660
So really what this visual is designed for is for continuous or quantitative fields things like the
30
00:02:05,660 --> 00:02:08,630
amount pledged in U.S. dollars for instance.
31
00:02:08,660 --> 00:02:10,050
You can see that by default.
32
00:02:10,160 --> 00:02:14,390
This was aggregated to a sum which in this case makes sense.
33
00:02:14,390 --> 00:02:17,030
So not really impressive quite yet.
34
00:02:17,130 --> 00:02:17,360
Right.
35
00:02:17,360 --> 00:02:23,510
We've got a little bar here showing the total amount pledged five hundred and forty four million now
36
00:02:23,540 --> 00:02:29,390
where it gets interesting is when we start populating the explained by fields and think of these explained
37
00:02:29,390 --> 00:02:35,180
by fields as the ways that we'd like to break down these dollar amounts this amount pledged.
38
00:02:35,240 --> 00:02:39,980
So I kind of like to think about this like a hierarchy right where I want to break things down at a
39
00:02:39,980 --> 00:02:46,850
high level maybe like category first and then a little bit deeper like subcategory and then we'll go
40
00:02:46,850 --> 00:02:51,700
all the way down to the project name and you can actually drop these in in any order.
41
00:02:51,770 --> 00:02:56,330
It's just kind of the way that I think about it my head and you'll notice nothing really changed here
42
00:02:56,330 --> 00:03:01,640
except this tiny little plus icon appeared which was so subtle that you might have missed but what that
43
00:03:01,640 --> 00:03:08,610
does is allows us to build our composition tree and we've got two options here with light bulbs.
44
00:03:08,660 --> 00:03:14,450
These are the A.I. driven options where we can find out what makes our metric the highest or the lowest
45
00:03:14,960 --> 00:03:18,990
or we can manually select one of the fields that we dragged in ourselves.
46
00:03:19,040 --> 00:03:25,130
So I'm going to start there because it's really clear really intuitive and also a really powerful option.
47
00:03:25,280 --> 00:03:28,940
So I'm going to take my amount pledged I mean to break it down by category.
48
00:03:29,210 --> 00:03:30,260
Boom there you go.
49
00:03:30,260 --> 00:03:38,000
Instantly I can see that game's design and technology products make up the bulk of my amount raised
50
00:03:38,000 --> 00:03:39,260
or amount pledged.
51
00:03:39,270 --> 00:03:42,860
There's a big big drop off before film and video publishing.
52
00:03:43,070 --> 00:03:47,270
And then I can actually scroll down a bit to see dance here at the bottom.
53
00:03:47,330 --> 00:03:50,040
So from here I can keep building my tree.
54
00:03:50,090 --> 00:03:54,380
I can select any one of these nodes and break it down at another level.
55
00:03:54,380 --> 00:03:56,870
So subcategory for instance.
56
00:03:56,870 --> 00:03:59,330
Here we go within games tabletop games.
57
00:03:59,330 --> 00:04:01,380
Clearly the biggest driver here.
58
00:04:01,550 --> 00:04:04,970
And let's drill one level deeper to project name.
59
00:04:04,970 --> 00:04:05,780
And there you have it.
60
00:04:05,840 --> 00:04:10,810
We've built a decomposition tree in about 30 seconds and it looks great.
61
00:04:10,810 --> 00:04:13,190
And check this out it's totally interactive.
62
00:04:13,190 --> 00:04:17,300
You can select any of these nodes totally dynamic.
63
00:04:17,340 --> 00:04:22,660
You can see exactly how your data is distributed where you'll find the bulk of the volume.
64
00:04:22,710 --> 00:04:25,730
It's just a great very user friendly tool.
65
00:04:25,920 --> 00:04:30,840
And honestly if they left it at that it would still be a pretty valuable addition to the visualizations
66
00:04:30,840 --> 00:04:34,580
pain but where it gets really cool is with the A.I. features.
67
00:04:35,010 --> 00:04:38,460
So let's close those out and let's go back to a plus sign.
68
00:04:38,460 --> 00:04:41,870
Now what if we choose the A.I. driven high value.
69
00:04:41,950 --> 00:04:42,170
OK.
70
00:04:42,180 --> 00:04:47,910
We see the same result that we did manually pulled in the category field and it's showing games at the
71
00:04:47,910 --> 00:04:48,410
top.
72
00:04:48,420 --> 00:04:51,910
Same order here kind of descending in volume.
73
00:04:51,920 --> 00:04:58,320
Now what we didn't really realize happened is that power b I looked at all of the available fields in
74
00:04:58,320 --> 00:05:05,130
order to find the one field with the biggest chunk of volume that happened B category with games here
75
00:05:05,130 --> 00:05:07,560
which drove one hundred and forty four million dollars.
76
00:05:07,680 --> 00:05:15,270
But if for instance we had country as well now all the sudden US is the top driver and it's really just
77
00:05:15,270 --> 00:05:20,180
based on the ratio of that four hundred fifty four million out of the total.
78
00:05:20,190 --> 00:05:26,130
And that's what determines the rankings here as well as which field was selected to be displayed.
79
00:05:26,130 --> 00:05:29,910
And that's nice but it's not really artificial intelligence.
80
00:05:29,910 --> 00:05:35,730
It's not that smart or that impressive but where it gets more interesting is if we go into formatting
81
00:05:35,730 --> 00:05:41,940
pain drill into analysis you can see that by default the analysis type is absolute.
82
00:05:41,940 --> 00:05:44,080
In other words it's based on volume.
83
00:05:44,130 --> 00:05:49,990
That's why we see always the kind of descending by volume we see the biggest categories pop up first.
84
00:05:50,070 --> 00:05:54,030
If we change this to relative watch what happens.
85
00:05:54,030 --> 00:05:55,510
Everything changed.
86
00:05:55,650 --> 00:06:03,000
Now that A.I. driven high value is showing project name which is the lowest most granular field that
87
00:06:03,000 --> 00:06:10,020
we have and it's showing this top project The Seventh Continent has the biggest relative factor that
88
00:06:10,020 --> 00:06:11,960
contributes to the amount pledged.
89
00:06:11,970 --> 00:06:18,390
So the simplest way to explain kind of the rationale here is that basically power by saying OK.
90
00:06:18,420 --> 00:06:25,580
Given all of the project names what would I expect a project to drive in terms of pledge amount.
91
00:06:25,590 --> 00:06:30,050
In other words it's the average across all projects in the entire table.
92
00:06:30,220 --> 00:06:36,840
And the reason why this project here The Seventh Continent is appearing at the top and why project name
93
00:06:36,840 --> 00:06:44,070
is appearing at all is because the difference between what this project drove and the average among
94
00:06:44,220 --> 00:06:49,960
all projects that difference was bigger than any other difference in the dataset.
95
00:06:49,980 --> 00:06:55,530
It's bigger than what the US drove as opposed to the average among all countries and it's bigger than
96
00:06:55,530 --> 00:06:59,670
what games drove compared to the average among all categories.
97
00:06:59,670 --> 00:07:06,240
And that's why A.I. has determined that project name should appear here and why The Seventh Continent
98
00:07:06,300 --> 00:07:11,460
is technically the biggest relative driver towards the amount pledged.
99
00:07:11,460 --> 00:07:17,070
So really interesting findings here that would be very tough to kind of realize without doing some pretty
100
00:07:17,070 --> 00:07:22,410
heavy analysis on your own so that let's go ahead and clear that out.
101
00:07:22,410 --> 00:07:27,360
I'm going to kind of recreate that initial tree that we built because I think it's a pretty helpful
102
00:07:27,360 --> 00:07:31,440
one category subcategory and project name.
103
00:07:31,440 --> 00:07:35,790
Note that if you hover over the headers here you can lock specific categories.
104
00:07:35,910 --> 00:07:42,390
So if you always want to start at the category level but you'd like users to be able to rearrange these
105
00:07:42,390 --> 00:07:49,140
or pull them out you can go ahead and lock one category or one column here of your tree but leave the
106
00:07:49,140 --> 00:07:51,790
others flexible now.
107
00:07:51,810 --> 00:07:56,490
Last thing I'll call out here that makes these visuals so great is that you can treat them just like
108
00:07:56,550 --> 00:07:57,790
any other visuals.
109
00:07:57,900 --> 00:07:59,360
You can format them.
110
00:07:59,490 --> 00:08:05,730
In fact Microsoft just released a lot of new formatting options for this composition tree visual specifically
111
00:08:06,210 --> 00:08:09,630
so you can change the look of your data bars if you want.
112
00:08:09,630 --> 00:08:15,120
You can change the connector lines all sorts of formatting options here and perhaps the best thing of
113
00:08:15,120 --> 00:08:22,110
all is that you can filter and cross filter these visuals based on things like sliders or other visuals
114
00:08:22,140 --> 00:08:23,260
in your report.
115
00:08:23,370 --> 00:08:24,620
So check this out.
116
00:08:24,630 --> 00:08:33,070
It's as simple as dropping in a map for instance or use a filled map here put country in and now check
117
00:08:33,070 --> 00:08:33,370
it out.
118
00:08:33,370 --> 00:08:43,900
R D composition tree is now dynamic it's tied to our map we can look at the UK France Canada the US
119
00:08:44,290 --> 00:08:46,400
and everything updates dynamically.
120
00:08:46,600 --> 00:08:53,140
Again really powerful tool when it comes to ad hoc data exploration data distribution and root cause
121
00:08:53,140 --> 00:08:54,230
analysis.
122
00:08:54,280 --> 00:08:58,540
So there you have it the new a.i. Driven decomposition tree visual.
13059
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.