Day 11 — Building the report — visuals that get read

The measures are done. Now make something a manager will actually look at.

Choosing the visual

The chart type is decided by the question, not by which one looks impressive.

Question	Visual
How is this changing over time?	Line chart
Which category is biggest?	Bar chart, sorted
What is the single headline number?	Card
Numbers by two dimensions at once	Matrix
Is this number good or bad?	KPI or card with a comparison
How do parts make a whole?	Stacked bar — rarely a pie

Pie charts and their relatives People cannot compare angles accurately. Above three or four slices, a pie becomes decoration. A sorted bar chart answers the same question in a fraction of the time. Donut charts, gauges and funnels have the same problem — they look like a dashboard rather than communicating one.

Cards need a comparison

A number on its own means nothing. "30,37,800" — is that good?

17,51,700 in 2025, up 4,65,600 on the year before — 36.2% growth. Now the number means something. Every headline card should carry a comparison: last year, target, or previous period.

Layout — top left is the most valuable space

Eyes start top left. Put the headline numbers there, supporting detail below, and the deepest detail last.

One page, one question A page that answers "how are we performing this month?" is useful. A page with nineteen visuals answering nothing in particular is a data dump. If a page needs scrolling, it is probably two pages.

Slicers

Slicers change the filter context — which means every measure you wrote reacts automatically. That is why the measures were written to be context-aware.

Imagine a Category slicer on this page. Both columns would recalculate — but the percentage still divides by the all-region total, because ALL(Region) only removes the Region filter and leaves Category alone. Precision in the ALL is what makes a percentage behave sensibly under slicing.

Slicers are not free Every slicer forces a recalculation of every visual on the page. Four or five is plenty. Beyond that, use the filter pane, which is cheaper and does not consume canvas.

Formatting that earns its place

Sort bar charts by value, not alphabetically. The order is the message.
Use the same colour for the same thing on every page.
Reserve red and green for good and bad — never as ordinary category colours.
Round hard. Nobody needs 30,37,800.00 — 30.4 lakh is easier to hold in mind.
Say the units. "Sales (₹ lakh)" prevents an expensive misunderstanding.
Delete every gridline and border that is not doing a job.

The performance rules

Five habits that keep a report fast

Remove unneeded columns in Power Query, not in the report
Prefer measures to calculated columns
Avoid FILTER when a plain CALCULATE filter would do
Keep the star schema — avoid chains of related tables
Limit visuals per page; every one is a separate query

Before you publish

Do the parts add up to the total? Check one region against the grand total by hand.
Does every measure still work with no filters, and with every slicer set?
Are there blanks that should be zeros, or zeros that should be blanks?
Does a total row show anything meaningless — like the growth percentage on Day 9?
Would someone who has never seen the report understand it without you in the room?

Four regions, four measures, every number checkable by hand. That is a report worth publishing.

Try these yourself

Choose the right visual for "sales by month over two years".
Choose the right visual for "which category sells most".
Explain why a card should carry a comparison.
Give two reasons to avoid a pie chart with eight slices.
List three things you would check before publishing.

← PreviousDay 10 — Iterators, ranking and top-N Next →Day 12 — 12 practice questions