Cosa contiene

Un file SKILL.md (la guida operativa) + un modello di partenza. Un'anteprima:

Stacked Column Chart

When to use

• You have numbers broken into parts — sales by channel per quarter, hours by project per week, budget by category per month — and a plain table hides which totals moved and why
• A slide, report, or post needs one chart that shows both the whole (how tall each column is) and the mix (what each column is made of) at a glance
• A spreadsheet's default stacked chart came out with a rainbow palette, a legend nobody can match to the segments, and labels too small to read
• You need to compare a handful of groups (say, 3–8 columns) across a handful of categories (say, 2–6 segments each) without resorting to a wall of pie charts
• An audience needs to see one story in the data — which segment grew, which shrank, which total changed — rather than raw numbers to decode themselves

Workflow

1. Shape the table (input: raw numbers → output: a tidy chart table)
   Lay the data out as one row per column and one column per segment, with every cell a number in the same unit — no blanks, no mixed units, no percentages mixed with counts. Put the segments in a meaningful order: largest-to-smallest, or the order the story needs (for example, a baseline segment at the bottom so its changes are easy to track). If a segment is zero or near-zero in every column, fold it into an "Other" row or drop it — a sliver nobody can see adds noise, not information. Check that every column's segments add up to its total before drawing anything.
2. Choose the segment order and palette (input: tidy table → output: segment order + colors)
   Stack the segments so the one the reader should compare first sits on the baseline (the bottom), because only the bottom segment shares a common starting line across columns. Order the rest by size or by story, and keep the same order in every column. Give each segment one flat, distinct color — silvers and greys with one accent color for the segment that matters most, never a rainbow — and make sure neighboring segments differ in lightness as well as hue so the chart survives greyscale printing and color-blind readers. Write the palette down as a legend table (segment → color) before drawing.
3. Draw the columns (input: table + palette → output: the chart)
   Draw one column per group, all the same width, with even gaps between them, scaled so the tallest column fits with headroom for its total label. Stack each column's segments bottom-to-top in the chosen order, each segment's height proportional to its value. Keep the baseline flat and the axis honest: the vertical axis starts at zero, with a few evenly spaced gridlines and plain number labels — no truncated axes, no 3D depth, no shadows, no rounded segment corners that distort heights. If a segment is too thin to label inside, label it outside with a short leader line.
4. Label everything on the chart (input: chart → output: a self-reading chart)
   Put the value on or beside every segment big enough to hold it, and put the total on top of every column in bold. Label the category under each column and give the chart a title that states the finding — "Online overtook in-store in Q3" — not "Sales by channel." Put the legend in reading order matching the stack (bottom segment first), and add a one-line note under the chart for the unit, the period, and the source ("Units sold, Jan–Dec, company records"). A reader should understand the chart without hunting for a caption in the surrounding text.
5. Check the read (input: labeled chart → output: a verified chart)
   Test the chart at the size it will actually be seen — slide, page, or thumbnail — and ask three questions: can I read every label, can I compare any two totals, can I name the story in one sentence. Re-add the numbers from the chart and compare them against the source table; a single mis-scaled segment silently corrupts the whole story. If the chart fails the thumbnail test, cut segments or split it into two charts rather than shrinking the type — a chart nobody can read is worse than no chart.

Anteprima del modello

Stacked Column Chart — Data + Layout Template

One dataset = one chart. Fill in the table first, then draw.

Chart title (states the finding): {e.g. Online overtook in-store in Q3}

Unit: {e.g. units sold} | Period: {e.g. Jan–Dec} | Source: {e.g. company records}

Tidy data table (one unit, no blanks, totals verified)

Check: [ ] every row's segments add up to its total [ ] segments in story order, same in every column

Segment order + palette (bottom segment first = the one to compare)

Palette check: [ ] flat colors, no rainbow [ ] neighbors differ in lightness [ ] survives greyscale

Layout

Il pacchetto della skill (SKILL.md + modello) è scritto in inglese.

Acquista il pacchetto completo della skill — $4.99 Pro — tutte le 511 skill