Same colour per measure, zero-based bars, per-unit not totals

Three rules that hold across panels

Module 8 taught honesty for one chart. A dashboard has six, and the reader's eye moves between them, so three rules apply across the panels:

• One measure, one colour. If sales are blue in the province bar, they are blue in the monthly line and blue in the KPI tile. Show sales in orange on one panel and the reader assumes it is a different measure — cost, or units — and reads the dashboard as saying something it does not. Colour is a label, not a decoration, and a label means the same thing everywhere.
• Zero-based bars and columns. Bar length is the value. The monthly sales column chart in the first draft had its axis starting at 1,500: Jan 1,610 stood 110 tall and Feb 2,560 stood 1,060 tall — nearly ten times the height for a rise of 59 %. Start at zero and the columns are 1,610 and 2,560, which is the truth. A line chart may sometimes zoom to show a small movement, but it must say so; a bar never may.
• Same period length on both sides. Comparing March (one month) with Q1 (three months), or this year's twelve weeks with last year's ten, is a comparison of nothing. The dashboard states its period in every title — Q1 2025, Jan–Mar — and any pair it compares covers the same span. Truncating a trend to the months that flatter it is Module 8's misleading display, now in project form.

Worked example — the totals trap

The first draft's headline read: "Alberta is our biggest market." The evidence was the province total: AB 2,300 against ON 1,995. Both numbers are right and the sentence is wrong, because it is a population statement about provinces that do not have the same population of orders — AB placed two orders in Q1, Ontario four. Totals reward whoever has more of whatever is being counted; they answer which province did we sell most to, not where are orders worth more, and Renata's question was the second.

A fair comparison divides by a base and says which base. Two are available:

• Per order. AB 2,300 / 2 = 1,150.00; ON 1,995 / 4 = 498.75. AB is ahead by 651.25 — on two orders.
• Per unit. AB shipped 8 + 6 = 14 units, ON 3 + 5 + 4 + 2 = 14 units — the same base by coincidence. AB 2,300 / 14 = 164.29; ON 1,995 / 14 = 142.50. AB is ahead by 21.79 per unit, about 15 %.

The rewritten headline: "Per unit sold, AB $164.29 vs ON $142.50 (14 units each); per order, AB $1,150 vs ON $499 on two AB orders." Longer, and it states the base twice — which is the rule: note the base. A per-unit or per-capita figure with no base stated is a total in disguise.

Fixing the draft, panel by panel

The three corrections, so you recognise each on a dashboard you did not build:

Fill the per-unit worksheet

The numeric item asks for both per-unit figures; the method is total amount ÷ total units per province.

CRISP-DM: fair comparison is evaluation deciding what the dashboard may claim, and deployment writing it so the claim is honest.