Test your eye.

1. A field-name title. "Average of Pay Rate by Department" names the data, not the finding. Rule 15, Titles.
2. Alphabetical order. The ranking is the point, and alphabetical order hides it. Rule 9.
3. No unit, two decimals. "31.90" of what? Dollars an hour, to one decimal. Rule 14.
4. Labels and an axis. They say the same thing twice, and the axis titles repeat the chart title. Rule 11.
5. Nothing to look at. Every column is the same blue. Rule 4.

Fixed: sorted, in dollars, with the two departments the title compares in colour. Executive Office is one person, and the subtitle says so.

1. 22 slices. Nobody can rank angles that small. A pie works for two or three parts of a whole. Pie and donut charts, Rule 3.
2. A 22-item legend. The reader's eye goes back and forth between slice and key for every source. Rule 10.
3. 22 colours that mean nothing. Colour should point at the answer. Rule 4.
4. No numbers and a field-name title. The reader can't tell that the top two sources bring in 60 people. Rule 15.

Fixed: a sorted bar of the top eight, with the other 14 sources grouped at the bottom, where "other" always goes.

1. The callout describes the chart. "This chart shows…" adds nothing. A callout says what it means and what to do. Callouts.
2. It's far from the evidence. It sits in a wide box under the chart, away from the referral bar it should explain. Put it beside that bar. Callouts.
3. Its colour doesn't match. The heading is orange, the highlighted bar is blue, so nothing links words to data. Rule 16.
4. No comparison. Is 12.90% good? The company rate (34%) belongs on the chart. Rule 5.
5. A label title and two-decimal rates. Titles, Rule 14.

Fixed: the title states the finding, a dashed line shows the company rate, and the callout sits next to the referral bar in the same blue.

1. Field names as labels. "Sum of Left Company" tells the reader a machine made it. Rule 15, Cards.
2. Wrong formats. 102.00 people, and a rate shown as 0.34 instead of 34%. Rule 14.
3. No comparison. 34%: up or down? Better or worse than what? Rule 5.
4. Cards that don't serve the question. Two cards show the same fact (102 left, 34% left), and headcount answers nothing about turnover. Three cards are plenty. Cards.

Fixed: three cards, plain labels, each with a comparison. Only the gaps are coloured, with an arrow so colour isn't the only cue.

1. The wrong page type. Leaders deciding in ten minutes need one answer, not a tool to explore. Rule 1, Presentations vs. dashboards.
2. A question for a title. Right for a dashboard, wrong here: state the answer, with a number and a comparison. Titles.
3. Slicers. Someone will filter mid-meeting and the numbers on screen will change under the discussion. A presentation page has none. Presentations vs. dashboards.
4. No "so what". Nothing tells the room what to do, and three cards compete with the chart for the one message. Rule 2, Callouts.

Fixed: the presentation version, with an action title, one chart, a takeaway, and the source and date.

1. Seven colours that mean nothing. Grey everything, then colour the scores the title is about. Rule 4.
2. A legend that repeats the axis. The names are already under the columns. Rule 10.
3. A rating scale in alphabetical order. "Exceeds" sits between "90-day meets" and "Exceptional". Use the scale's own order. Rule 9.
4. False precision. 53.33% rests on 15 people, and 44.44% on 9. Round to whole percentages and say how small the groups are. Rule 14.
5. A field-name title. Rule 15.

Fixed: best rating to worst, low ratings in blue, and the two "too early" groups combined (26 of 68 left, 38%).

1. Ink that carries no data. The border, gridlines, both axis titles and a one-series legend. Rule 6.
2. Labels and an axis. Keep the labels, drop the axis. Rule 11.
3. "84.00" hires. People come in whole numbers. Rule 14.
4. A field-name title. Rule 15.
5. No emphasis. The 2011 peak is the story; colour it. Rule 4.

Fixed: only the data and its labels remain. The subtitle warns that 2016 is a part year that includes future starts.

1. The axis doesn't start at zero. Someone set its minimum to 33.84%, so a 0.05-point gap looks like a near four-to-one difference. Bar length must be proportional to the value. Rule 8.
2. Two decimals. They invite the reader to see a difference that doesn't matter. Rule 14.
3. The title misses the finding. The finding is that there is no gap. Say so. Titles.

Fixed: bars from zero, the counts behind each rate, and a title that states the finding.

1. Text sorted as text. Power BI sorts "Under $20/hr" after "$45+/hr", so the lowest band ends up last and the pattern breaks. Sort by a band-order column. Rule 9.
2. A label title. The finding is that turnover halves from the lowest band to the highest. Titles.
3. No emphasis. Colour the two bands the title compares. Rule 4.
4. Labels, axis and two decimals. Rule 11, Rule 14.

Fixed: lowest pay to highest. In Power BI, add a Pay Band Order column (1-4), select Pay Band, and use Column tools > Sort by column > Pay Band Order.

1. Cents on a budget. "$10,980.00" is seven digits where three do the job: $11.0K. Rule 14.
2. Truncated names. Long category names belong on a bar chart, where they read across. Rule 3.
3. Alphabetical order. Rule 9.
4. Labels and an axis, both in cents. Rule 11.
5. "Sum of Total by Source". Rule 15.

Fixed: sorted bars in thousands, with the two biggest costs in blue. Together they are $21.0K of the $84.5K spent.

Model answer: "A third of employees have left, and most of them chose to go: 88 of 102"

It gives the size (a third), the split that matters (voluntary), and both numbers.

Model answer: "All 25 people who left unhappy or for more money worked in Production"

"All" is the finding. It points straight at where to act.

Model answer: "IT/IS lost 7 people in 2015, after none in 2014"

A change over time needs both ends of the change in the title.

Model answer: "15 of the 25 people hired through search engines have left, nearly twice the company rate"

60% against the company's 34% is the comparison that makes the number mean something.

Model answer: "Production employs 208 of 301 people, so its turnover drives the company's"

The second half is the "so what": why a reader should care about headcount on a turnover report.