HomeAboutProjectsHow I workContact
05Storytelling + UX

CommunicateIt’s understood without me there.

I guide the story with good narrative, good design and good experience. The title states the conclusion, one metric leads and every number has context.

Power BISketchesData storytelling
01The title

The titledoes half the work.

It’s the most-read text on the page. It can be a label or it can be the answer.

LabelSales by region
With contextSales by region, September
ConclusionThe north falls 8% and explains the whole drop
00Why it matters

Being right isn’t enough.It has to be understood.

On average, only one in four employees uses BI tools in their daily work.

When…
1It’s published
2Monday comes
3Someone needs a number
4Someone new joins
5Three months go by
6Another dashboard is requested
✗Only published
12 visualsAll the same size
Nobody opens itThey bring their Excel
They ask meI am the dashboard
One hourExplaining how to read it
Few use itAnd fewer each month
Starting from zeroThe last one didn’t work
✓Well communicated
One metric leadsThe rest explains it
It opens the meetingEveryone looks at the same
They find itThe title says it
10 minutesIt explains itself
It’s still usedChecked at 30 and 90 days
It gets extendedOn what already works
The usage figure is from BARC, The BI & Analytics Survey 26 (more than 1,000 professionals). The rest are example figures.
00What I do and don’t

What I don’t do,and what I do.

1Before building
✗I don’t

Build and see what happens

✓I do

Sketch: 10 minutes

2The title
✗I don’t

“Sales by region”

✓I do

States the conclusion

3The page
✗I don’t

12 visuals, no conclusion

✓I do

One metric leads

4Color
✗I don’t

3D and competing backgrounds

✓I do

One accent, for what matters

5Delivery
✗I don’t

Email the link and done

✓I do

I check usage at 30 and 90 days

00Who reads it

Three readers.Each one gets their depth.

The same analysis, at three depths. The conclusion always goes first.

30 secondsDecides. Reads the title and one number.
1
When they ask……they get
  • The answerAre we OK?One sentence and one number
  • The askWhat do you need from me?One decision, with its options
3 minutesManages. Reads the page.
2
When they ask……they get
  • The placeWhere do I act this week?The ranking, worst first
  • The referenceCompared with what?Target, history and peers
  • The ownerWho takes it?A name next to each action
30 minutesChecks. Reads the detail.
3
When they ask……they get
  • The sourceWhere does this come from?The table and the definition
  • The repeatCan I run it again?The query, with its date
00The chart

The question picks the chart.Not taste.

Seven questions cover almost everything a dashboard gets asked.

The question→What it compares→The chart
“How much, against the target?”
It’sMagnitude

✓Bars from zero, with the target as a line.

“How did it change?”
It’sChange over time

✓A line. Bars only for a few periods.

“Who comes first?”
It’sRanking

✓Horizontal bars, sorted.

“What share is it?”
It’sPart of a whole

✓One stacked bar. A pie only with two or three parts.

“How is it spread?”
It’sDistribution

✓A histogram. The average alone hides it.

“Do they move together?”
It’sCorrelation

✓A scatter plot, saying what each dot is.

“What is the exact figure?”
It’sLookup

✓A table. Not everything is a chart.

The categories follow the Financial Times Visual Vocabulary.
00The decisions

One dashboard,four decisions.

Pick one and see where it sits on the dashboard.

  1. 1

    One metric leads

    On-time delivery goes first: it is what the customer feels.

  2. 2

    One point, one explanation

    Only the week that changes the story is marked.

  3. 3

    Where to look next

    The ranking says which warehouse to check first.

  4. 4

    Context, not decoration

    Every number comes with a comparison.

Operations · week 42Simulated
OTIF92.4%
Fill rate96.1%
Cycle3.8 d
Stock-outs14
vs target 95%: −2.6 ppvs 4 weeks: −1.1 pp
00Common mistakes

Six common mistakesand how they get fixed.

The title

“Sales by region”

“The north falls 8%”

A label leaves the work to the reader. A conclusion does it for them.
Eight categories

A pie

Sorted bars

Angles are hard to compare. Lengths aren’t.
Color

One per category

Gray and one accent

If everything stands out, nothing does.
Good and bad

Only red and green

Color and a mark

About 1 in 12 men can’t tell those colors apart well.
The axis

Starts at 90

Starts at zero

A cut bar exaggerates the difference.
The page

12 visuals

One leads, three explain

A page is read in one look, or it isn’t read.
00For AI

The dashboard and the AIgive the same answer.

If Copilot or Claude reads the model, it has to find the same words as the page.

✗
Not like thisDashboard: 92.4%. AI: 94.1%.

Two definitions of “on time”.

✓
Like thisDashboard: 92.4%. AI: 92.4%.

A single measure, with its description.

  1. 1Names people useIn the model

    “On-time delivery”, not “otif_flag_v2”.

    I askWhat do they call it in the meeting?
    I leave with✓Tables and measures renamed
  2. 2A description per measureIn the model

    What it counts, what it leaves out and its unit.

    I askWould someone new understand it?
    I leave with✓One line per measure
  3. 3Verified answersIn Power BI

    The most asked questions, each tied to an approved visual.

    I askWhat do they ask every week?
    I leave with✓Answers a person already reviewed
  4. 4The same question, three waysMe

    I ask the AI and compare with the page.

    I askDoes it match the dashboard?
    I leave with✓A note of what didn’t match
In Power BI this is done in Prep data for AI, which is in preview: it reduces ambiguity, it doesn’t guarantee the answer.
03After publishing

Publishingis day 1 of 90.

A dashboard nobody uses costs twice. This plan turns it into a team habit.

WhenWhat I doWhat I check
Day 030-minute launch: how to read each page in 3 minutes.Everyone can get in
Week 1Monday subscription for the main person.Active users
Day 30I review usage: what isn’t opened gets changed or removed.Pages nobody opens
Day 90I measure against the brief from phase 1.What the brief agreed
00What I get asked

What I get asked,and what’s behind it.

The request is about the page. The need is about a decision.

What I get asked→What it is→What I do
“Add one more chart.”
It’sA new question

✓I ask which decision it supports. If none, it stays out.

“Make it look nicer.”
It’sA reading problem

✓Fewer things and more order. Then, color.

“Nobody uses it.”
It’sAn adoption problem

✓I look at usage and sit down with three users.

“Put everything on one page.”
It’sA summary

✓One page with what decides. The rest, one click away.

“I need it for the board.”
It’sA presentation

✓Three slides: the conclusion, the evidence and the ask.

“Can it come by email?”
It’sA subscription

✓The same page, sent every Monday.

“Can I just ask Copilot?”
It’sA question to the model

✓Yes, once names and definitions are ready.

00How I know I’m done

They decide without calling me.And nobody builds their own Excel.

Communicating ends when the dashboard is used every week, without me in the room.

I’m done when…
  1. ✓
    The title states the conclusionMargin falls on 7 products: raise price or drop them.
  2. ✓
    One metric leadsThe rest explains it.
  3. ✓
    It reads in 30 secondsTested with someone who wasn’t in the project.
  4. ✓
    It’s still used after day 30What nobody opens gets changed or removed.
Margin falls on 7 products: raise price or drop them
31.8%
Day 90: still in use and nobody builds their own Excel.

One phase left: document, so everything makes sense without me.