WEBVTT
1
00:00:00.080 --> 00:00:03.799
POWERBI, the golden promise of self service analytics and the
2
00:00:03.839 --> 00:00:07.519
silent destroyer of data consistency. Everyone loves it until you
3
00:00:07.559 --> 00:00:10.960
realize your company has forty versions of the same sales dashboard,
4
00:00:11.119 --> 00:00:13.240
each claiming to be the truth. You laugh, I can
5
00:00:13.279 --> 00:00:15.039
hear it, but you know it's true. It starts with
6
00:00:15.039 --> 00:00:17.920
one quick insight, and next thing you know, the marketing
7
00:00:17.960 --> 00:00:21.920
interns spreadsheet is driving executive decisions. Congratulations, you've built a
8
00:00:21.960 --> 00:00:25.000
decentralized empire of contradiction. Now let me clarify why you're here.
9
00:00:25.079 --> 00:00:27.399
You're not learning how to use powerbi, you already know
10
00:00:27.480 --> 00:00:29.640
that part. You're learning how to plan it, how to
11
00:00:29.800 --> 00:00:34.799
architect control into creativity, governance into flexibility, and confidence into chaos.
12
00:00:36.039 --> 00:00:39.280
Defining the power BI wild West, the problem of duplication.
13
00:00:39.719 --> 00:00:43.320
Picture this. Every department in your company builds its own report.
14
00:00:43.520 --> 00:00:47.960
Finance has revenue, Sales has revenue. Operations apparently also has revenue.
15
00:00:48.479 --> 00:00:53.719
Same word, three definitions, none agree, And when executives ask
16
00:00:54.159 --> 00:00:57.359
what's our revenue this quarter, five people give six numbers.
17
00:00:57.759 --> 00:01:01.039
It's not incompetence. It's entropy disguised as an powerment. The
18
00:01:01.079 --> 00:01:04.040
problem is that POWERBI makes it too easy to build fast.
19
00:01:04.319 --> 00:01:06.799
The moment someone can connect an Excel file, they're suddenly
20
00:01:06.840 --> 00:01:09.719
a data modeler. They save to one drive, share links,
21
00:01:09.760 --> 00:01:12.400
and before you can say version control, you have dashboards
22
00:01:12.439 --> 00:01:15.480
breeding like rabbits, and because everyone thinks their version is
23
00:01:15.480 --> 00:01:18.400
the good one, no one consolidates, no one even remembers
24
00:01:18.400 --> 00:01:21.439
which measure came first. In the short term, this seems empowering.
25
00:01:21.519 --> 00:01:24.680
Analysts feel productive. Managers get their charts, but over time
26
00:01:24.719 --> 00:01:27.439
you stop trusting their numbers. Meetings devolve into crime scenes.
27
00:01:27.480 --> 00:01:30.879
Everyone's examining conflicting evidence. The CFO swears the trend line
28
00:01:30.879 --> 00:01:33.560
shows growth. The head of sales insists its decline. They're
29
00:01:33.560 --> 00:01:37.040
both right because their data slices come from different refreshers, filters,
30
00:01:37.200 --> 00:01:40.719
or strangely named tables like data final V three figx fixed.
31
00:01:41.120 --> 00:01:44.640
That's the hidden cost of duplication. Every report becomes technically
32
00:01:44.640 --> 00:01:47.760
correct within its own microcosm, but the organization loses a
33
00:01:47.799 --> 00:01:50.920
single version of truth. Suddenly, your self service environment isn't
34
00:01:51.000 --> 00:01:54.519
data driven, it's faith based, and faith while inspirational, isn't
35
00:01:54.560 --> 00:01:58.040
great for auditing. Duplication also kills scalability. You can't optimize
36
00:01:58.040 --> 00:02:00.480
refresh schedules when twenty similar models have are the same.
37
00:02:00.560 --> 00:02:05.920
Database performance tanks, gateways crash, and somewhere an IT engineer
38
00:02:06.120 --> 00:02:09.479
silently resigns. This chaos doesn't happen because anyone's lazy. It
39
00:02:09.520 --> 00:02:13.719
happens because nobody planned ownership, certification or lineage. The tools
40
00:02:13.759 --> 00:02:18.159
outgrew the governance, and Microsoft's convenience doesn't help. My workspace
41
00:02:18.240 --> 00:02:21.400
might as well be renamed my dumpster of unmonitored reports.
42
00:02:21.479 --> 00:02:24.400
When every user operates in isolation, the organization becomes a
43
00:02:24.400 --> 00:02:27.120
collection of private data islands. You get faster answers in
44
00:02:27.159 --> 00:02:30.039
the beginning, but slower decisions in the end. That contradiction
45
00:02:30.120 --> 00:02:32.599
is the pattern of every power BIA environment gone rogue.
46
00:02:32.639 --> 00:02:35.080
So what's the fix? Not more rules, not less freedom.
47
00:02:35.319 --> 00:02:38.479
The fix is structure, specifically, a structure that separates stability
48
00:02:38.479 --> 00:02:43.319
from experimentation without killing either, introducing hub and spoke architecture.
49
00:02:43.439 --> 00:02:45.520
The hub and spoke design is not a metaphor, it's
50
00:02:45.560 --> 00:02:49.000
an organizational necessity. Picture power BI as a city. The
51
00:02:49.080 --> 00:02:52.319
hub is your city center, the infrastructure, utilities, and laws
52
00:02:52.319 --> 00:02:55.960
that make life bearable. The spokes are neighborhoods, creative, adaptive,
53
00:02:56.039 --> 00:02:59.159
sometimes noisy, but connected by design. Without the hub, the
54
00:02:59.159 --> 00:03:02.599
neighborhoods these into chaos. Without the spokes, the city stagnates.
55
00:03:03.120 --> 00:03:07.479
In power bi terms, the hub holds your certified semantic models,
56
00:03:07.599 --> 00:03:11.520
shared data sets, and standardized measures the official truth. The
57
00:03:11.560 --> 00:03:16.159
spokes are your departmental workspaces. Sales, finance, HR build for exploration,
58
00:03:16.360 --> 00:03:19.520
local customization, and quick iteration. They consume from the hub,
59
00:03:19.520 --> 00:03:22.479
but don't redefine it. This model enforces a beautiful kind
60
00:03:22.479 --> 00:03:25.199
of discipline. Everyone still moves fast, but they move along
61
00:03:25.240 --> 00:03:28.199
defined lanes. When finance build a dashboard, it references the
62
00:03:28.240 --> 00:03:31.800
certified financial data set. When sales creates a pipeline tracker,
63
00:03:32.039 --> 00:03:35.520
it uses the same revenue definition as finance. No debates,
64
00:03:35.639 --> 00:03:38.960
no duplicates, just different views of a shared reality. Planning
65
00:03:39.000 --> 00:03:43.000
a hub in spoke isn't glamorous. Its maintenance of intellectual hygiene.
66
00:03:43.199 --> 00:03:46.479
You define data ownership by domain. Who maintains the sales model,
67
00:03:46.520 --> 00:03:50.199
who validates the HR metrics. Each certified data set should
68
00:03:50.280 --> 00:03:53.280
have both a business and technical owner. One ensures the
69
00:03:53.280 --> 00:03:56.439
measure's logic is sound, the other ensures it actually refreshes.
70
00:03:56.719 --> 00:04:01.080
Then there's life cycle discipline, dev test, PROD shocking. I know,
71
00:04:01.400 --> 00:04:05.240
governance means using environments development happens in the spoke. Testing
72
00:04:05.240 --> 00:04:08.840
happens in a controlled workspace. Production gets only certified artifacts.
73
00:04:08.919 --> 00:04:13.000
This simple progression eliminates midnight heroics, where someone publishes final
74
00:04:13.120 --> 00:04:16.000
dashboard minutes before the board meeting. The genius of hubin
75
00:04:16.079 --> 00:04:20.040
spoke is that it balances agility with reliability. Departments get
76
00:04:20.079 --> 00:04:22.759
their self service, but it's anchored in enterprise trust. It
77
00:04:23.040 --> 00:04:26.639
keeps oversight without becoming a bottleneck. Analysts innovate without reinventing
78
00:04:26.720 --> 00:04:30.560
KPIs every week. The chaos isn't eliminated, it's domesticated. From
79
00:04:30.639 --> 00:04:35.959
this foundation, true enterprise analytics is possible. Consistent performance, predictable refreshes,
80
00:04:35.959 --> 00:04:39.079
and metrics everyone can actually agree on. And yes, that's
81
00:04:39.160 --> 00:04:42.240
rareer than it should be. The Hub. Let's get serious
82
00:04:42.279 --> 00:04:45.480
for a moment, because this is where most organizations fail spectacularly.
83
00:04:45.519 --> 00:04:48.360
The Hub isn't a powerbi workspace. It's a philosophy wrapped
84
00:04:48.360 --> 00:04:51.000
in a folder. It defines who owns reality. When people
85
00:04:51.040 --> 00:04:53.680
ask where do I get the official revenue number, the
86
00:04:53.680 --> 00:04:56.839
answer should never be depends who you ask. It should
87
00:04:56.839 --> 00:04:59.959
be the certified finance model in the Hub, one place,
88
00:05:00.160 --> 00:05:03.040
one truth, one data set to rule them all. A
89
00:05:03.079 --> 00:05:06.920
shared data set is basically your organization's bloodstream. It carries clean,
90
00:05:06.959 --> 00:05:10.120
standardized data from the source to every report that consumes it.
91
00:05:10.519 --> 00:05:13.959
But unlike human blood, this data set doesn't circulate automatically.
92
00:05:14.040 --> 00:05:16.399
You have to control its flow. The minute one rogue
93
00:05:16.439 --> 00:05:20.040
analyst starts building direct connections to the underlying database in
94
00:05:20.079 --> 00:05:23.600
their own workspace, your bloodstream develops a clot, and clots
95
00:05:23.720 --> 00:05:26.759
in both analytics and biology cause strokes. So the golden
96
00:05:26.839 --> 00:05:30.480
rule the hub produces the spokes consume. That means every
97
00:05:30.519 --> 00:05:33.680
certified model, your finance model, your HR model, your sales
98
00:05:33.720 --> 00:05:36.600
performance model, lives in the hub. The spokes only connect
99
00:05:36.639 --> 00:05:39.319
to them. No copy paste imports, no local tweaks to
100
00:05:39.360 --> 00:05:42.160
fix it temporarily. If you need a tweak, propose it
101
00:05:42.199 --> 00:05:44.639
back to the owner. Because the hub is not a museum,
102
00:05:44.680 --> 00:05:48.199
it's a living system, it evolves, but deliberately. Now governance
103
00:05:48.240 --> 00:05:51.680
begins with ownership. Every shared data set must have two parents,
104
00:05:51.720 --> 00:05:54.000
a business owner and a technical one. The business owner
105
00:05:54.040 --> 00:05:56.800
decides what the measure means, what qualifies as active customer
106
00:05:56.959 --> 00:06:00.839
or gross margin. The technical owner ensures the model actually functions.
107
00:06:01.079 --> 00:06:05.160
Refreshed schedules, DAX Performance Gateway reliability. Both names should be
108
00:06:05.279 --> 00:06:07.920
right there in the data set description, because when that
109
00:06:07.959 --> 00:06:10.519
refresh fails at two am or the CFO challenge is
110
00:06:10.519 --> 00:06:12.639
a number at nine a M, you shouldn't need a
111
00:06:12.680 --> 00:06:15.959
company wide scavenger hunt to find who's responsible. Documenting the
112
00:06:16.000 --> 00:06:18.439
Hub sounds trivial until you realize memory is the least
113
00:06:18.480 --> 00:06:21.439
reliable form of governance In the Hub. Every data set
114
00:06:21.439 --> 00:06:24.680
deserves a read me, short, human readable and painfully clear.
115
00:06:25.040 --> 00:06:27.839
What are the data sources? What's the refresh frequency? Which
116
00:06:27.879 --> 00:06:31.319
reports depend on it? You're not writing literature, you're preventing archaeology.
117
00:06:31.639 --> 00:06:36.040
Without documentation, every analyst becomes Indiana Jones, digging through measure
118
00:06:36.040 --> 00:06:39.040
definitions that nobody's updated since twenty twenty two. Then there's
119
00:06:39.079 --> 00:06:42.959
certification POWERBI gives you two signals, promoted and certified. Promoted
120
00:06:43.000 --> 00:06:46.800
means someone thinks this is good. Certified means the Data
121
00:06:46.800 --> 00:06:49.319
Governance Board has checked it, blessed, and you may trust
122
00:06:49.319 --> 00:06:52.360
your career to it. In the Hub, certification isn't decorative,
123
00:06:52.399 --> 00:06:56.120
it's contractual. The certified status tells every other department, use this,
124
00:06:56.480 --> 00:06:59.399
not your homegrown version hiding in one drive. Certification also
125
00:06:59.480 --> 00:07:01.600
comes with a cut ontability if the logic changes. There's
126
00:07:01.639 --> 00:07:04.279
a change log. You don't silently swap a measure definition
127
00:07:04.319 --> 00:07:07.319
because someone panicked before a meeting. Lineage isn't optional either.
128
00:07:07.560 --> 00:07:11.079
A proper hub uses lineage view like a detective uses fingerprints.
129
00:07:11.279 --> 00:07:13.879
Every data set connects visibly to its sources and all
130
00:07:13.920 --> 00:07:17.480
downstream reports. When your CTO asks, if we deprecate that
131
00:07:17.560 --> 00:07:20.399
sequel table, what breaks, you should have an instant answered,
132
00:07:20.399 --> 00:07:22.439
not a hunch, not a guess, a lineage map that
133
00:07:22.439 --> 00:07:24.680
shows exactly which reports cry for help the moment you
134
00:07:24.680 --> 00:07:27.639
pull the plug. The hub turns cross department dependency from
135
00:07:27.680 --> 00:07:32.000
mystery into math. Version control comes next. No POWERBI isn't GIT,
136
00:07:32.120 --> 00:07:34.839
but you can treat it as code export PBIP files
137
00:07:34.920 --> 00:07:38.399
store them in a repot tag releases. When analysts break something,
138
00:07:38.519 --> 00:07:40.720
because they will, you can roll back to stability instead
139
00:07:40.720 --> 00:07:43.920
of re engineering from memory. Governance without version control is
140
00:07:43.959 --> 00:07:47.040
like driving without seat belts and insisting your reflexes are enough.
141
00:07:47.240 --> 00:07:49.879
Capacity planning also lives at the hub level. Shared data
142
00:07:49.879 --> 00:07:53.160
sets run on capacity. Capacity costs money. You don't put
143
00:07:53.160 --> 00:07:55.720
test models or one of prototypes there. The hub is
144
00:07:55.759 --> 00:08:00.000
production grade only optimized models, incremental refresh compressed columns the word.
145
00:08:00.439 --> 00:08:04.360
Every refresh must be scheduled deliberately to avoid collision. Refreshing
146
00:08:04.399 --> 00:08:07.399
fifteen models at eight am is not governance. It's CPU
147
00:08:07.480 --> 00:08:12.160
arsen Now, let's address the political side. Governance means saying no, strategically, calmly,
148
00:08:12.199 --> 00:08:15.560
and repeatedly. When a manager insists on adding a column
149
00:08:15.600 --> 00:08:18.399
because they need it right now, the HUB team evaluates
150
00:08:18.399 --> 00:08:21.360
it through impact, not emotion. How many reports depend on
151
00:08:21.360 --> 00:08:24.639
this measure? Does it align with business definitions? Adding one
152
00:08:24.680 --> 00:08:29.000
casual column might corrupt thirty downstream visuals? The Hub provides guardrails,
153
00:08:29.000 --> 00:08:32.879
not customer service, but authority alone isn't enough. Visibility is
154
00:08:33.200 --> 00:08:38.039
publish internal dashboards that track data set, health, refresh successes, failures,
155
00:08:38.320 --> 00:08:42.600
refresh duration, data set size, number of connected reports. Let
156
00:08:42.679 --> 00:08:46.559
leadership see governance in action. When executives visually witness uptime
157
00:08:46.600 --> 00:08:49.279
at ninety nine percent, governance stops looking like red tape
158
00:08:49.279 --> 00:08:52.000
and starts smelling like competence. Let's talk security. The Hub
159
00:08:52.080 --> 00:08:55.320
enforces invisible discipline. That means row level security and object
160
00:08:55.399 --> 00:08:58.000
level security are modeled here, not duct taped later by
161
00:08:58.000 --> 00:09:01.240
the spokes. You define the filters one by region, division
162
00:09:01.320 --> 00:09:04.759
or role, and every consuming report inherits them. No one copies,
163
00:09:04.799 --> 00:09:08.639
ducks filters across ten workspaces like medieval scribes reproducing scripture.
164
00:09:09.080 --> 00:09:13.080
Security is consistent inherited and auditible metadata. Hygiene rounds it out.
165
00:09:13.200 --> 00:09:16.600
Sensitivity labels, and data loss prevention policies should originate in
166
00:09:16.639 --> 00:09:19.840
the hub as defaults. Every certified data set carries its
167
00:09:19.879 --> 00:09:24.159
classification like an ID badge, public, internal, confidential. When those
168
00:09:24.240 --> 00:09:26.879
data sets flow into Excel or Outlook, the labels travel
169
00:09:26.919 --> 00:09:32.000
with them. Governance isn't about blocking, It's about making trust portable. Finally, culture,
170
00:09:32.399 --> 00:09:35.000
the hub is only as strong as the behavior it normalizes,
171
00:09:35.320 --> 00:09:38.879
so institutionalized short show and tell sessions where departments present
172
00:09:38.960 --> 00:09:42.120
improvements made to their reports or measures derived from Hub data.
173
00:09:42.720 --> 00:09:46.159
Little rituals like that remind everyone that governance is collaboration,
174
00:09:46.399 --> 00:09:49.360
not surveillance. The hub feeds the spokes, The spokes give
175
00:09:49.360 --> 00:09:52.039
feedback to the hub. It's an ecological loop, not a monarchy.
176
00:09:52.440 --> 00:09:55.360
When designed properly, the hub turns powerbi from a zoo
177
00:09:55.399 --> 00:09:58.519
into a zoo with fences, feeding schedules, and a veterinarian.
178
00:09:58.759 --> 00:10:01.919
The animals still roam, but nobody gets mauled. That is
179
00:10:01.960 --> 00:10:07.320
shared governance done right. Predictable refreshes, defined ownership, certified semantics,
180
00:10:07.600 --> 00:10:10.480
and a lineage so clear even auditors smile. That's the
181
00:10:10.519 --> 00:10:13.720
infrastructure of trust you build before the chaos begins, and
182
00:10:13.759 --> 00:10:17.720
once it exists, your spokes can finally innovate freely because
183
00:10:17.720 --> 00:10:21.240
they know the hub has their back the spokes. Now
184
00:10:21.279 --> 00:10:23.639
that the hub is keeping your data cleaned, certified, and
185
00:10:23.720 --> 00:10:26.200
under control, it's time to talk about the parts everyone
186
00:10:26.240 --> 00:10:29.759
actually sees the spokes. The spokes are where creativity lives,
187
00:10:29.960 --> 00:10:33.519
where analysts experiment, and where business decisions happen at speed.
188
00:10:33.879 --> 00:10:37.519
But freedom without structure is chaos. With better lighting, building
189
00:10:37.519 --> 00:10:41.639
in the spoke means operating inside lanes that protect performance, consistency,
190
00:10:41.720 --> 00:10:44.320
and user trust. A common temptation in the spokes is
191
00:10:44.360 --> 00:10:47.679
to rebuild what's already in the hub. Duplicate measures, import
192
00:10:47.679 --> 00:10:51.000
tables just in case, or tweak logic to match someone's
193
00:10:51.000 --> 00:10:54.840
anecdotal truth. Don't. The purpose of the spoke is to
194
00:10:54.879 --> 00:10:58.759
consume shared data, not reinterpret it. A thin report connects
195
00:10:58.799 --> 00:11:01.600
life to certified data sets. It doesn't drag entire models
196
00:11:01.600 --> 00:11:05.000
into memory. You're visualizing, not remodeling. Thinness is a virtue here,
197
00:11:05.159 --> 00:11:09.639
fewer dependencies faster refreshes, lighter performance load. Think of each
198
00:11:09.759 --> 00:11:13.679
spoke workspace as a showroom floor, elegantly displaying vehicles engineered
199
00:11:13.720 --> 00:11:16.919
in the hub's factory. Optimization starts with model connections. Every
200
00:11:16.960 --> 00:11:20.080
spoke should use live connections or direct links to semantic models,
201
00:11:20.120 --> 00:11:23.960
not copies. That keeps data consistent and refreshed schedules centralized.
202
00:11:24.279 --> 00:11:26.559
If someone insists on adding a localized measure, say a
203
00:11:26.600 --> 00:11:30.399
region specific KPI, contain it within the report layer, documented
204
00:11:30.440 --> 00:11:33.120
clearly as local logics so downstream users don't confuse it
205
00:11:33.159 --> 00:11:37.159
with a certified field. Remember, traceability is oxygen. Without it,
206
00:11:37.240 --> 00:11:42.080
creativity suffocates under ambiguity. Good spoke design also demands performance empathy.
207
00:11:42.279 --> 00:11:45.519
When you visualize a data set, each slicer, card and
208
00:11:45.639 --> 00:11:48.720
matrix is a query waiting to pounds on capacity. Layering
209
00:11:48.720 --> 00:11:50.879
twenty filters on a single page may look clever, but
210
00:11:50.919 --> 00:11:54.240
it will turn interactive exploration into molasses. Use bookmarks to
211
00:11:54.279 --> 00:11:58.320
hide visual clutter, separate summary dashboards from deep explorations, paginate
212
00:11:58.320 --> 00:12:02.080
detail views. The more predictable your query pattern, the fewer
213
00:12:02.080 --> 00:12:06.360
support tickets you'll generate about powerbi being slow, which spoiler
214
00:12:06.360 --> 00:12:10.360
alert usually means the report designer ignored basic logic. Now,
215
00:12:10.440 --> 00:12:14.480
let's touch on consistency. Every spoke should follow shared UI standards,
216
00:12:14.519 --> 00:12:19.240
color palettes, typography, layouts. It's not about esthetics, it's cognitive efficiency.
217
00:12:19.360 --> 00:12:22.200
If finance users switch to a sales dashboard and instantly
218
00:12:22.240 --> 00:12:26.039
know where to click, you've succeeded. Branded templates reduce friction
219
00:12:26.080 --> 00:12:29.519
and support adoption. Establish a design system at the hub level,
220
00:12:29.799 --> 00:12:33.600
approved fonds, regional color rules, margin constraints, then lock those
221
00:12:33.600 --> 00:12:36.679
into the shared theme. Jason spokes should inherit style, not
222
00:12:36.840 --> 00:12:40.279
improvise it like amateur painters. Navigation matters more than most
223
00:12:40.320 --> 00:12:43.840
analysts admit. Users don't want detective work. They want direction.
224
00:12:44.519 --> 00:12:48.360
Keep home pages clean, KPIs on top, filter's obvious context clear,
225
00:12:48.759 --> 00:12:52.279
use tooltips and horver explanations to make every number self explanatory.
226
00:12:52.320 --> 00:12:55.279
Every extra click is a tiny tax on comprehension. Great
227
00:12:55.360 --> 00:12:59.080
user experience in powerbi is invisible. It feels obvious because
228
00:12:59.120 --> 00:13:02.600
someone agonized labels that nobody notices. In a well planned
229
00:13:02.679 --> 00:13:07.000
hub and spoke ecosystem, collaboration flows both ways. Analysts in
230
00:13:07.039 --> 00:13:09.679
the spokes aren't rogue agents their scouts. When they find
231
00:13:09.720 --> 00:13:11.919
a better calculated measure, they submit it back to the
232
00:13:11.960 --> 00:13:15.440
hub for standardization. That's evolution through shared intelligence. The hub
233
00:13:15.480 --> 00:13:19.919
then republishes the improved logic, instantly updating every department downstream.
234
00:13:20.159 --> 00:13:24.159
This iterative loop turns experimentation into enterprise level progress. Without it,
235
00:13:24.200 --> 00:13:27.440
every innovation dies in departmental isolation, like a lab experiment,
236
00:13:27.559 --> 00:13:30.600
never peer reviewed. And here's the part people forget. Thin
237
00:13:30.679 --> 00:13:34.600
reports are cheaper to maintain, They refresh faster, consume less capacity,
