Most consulting charts focus on central tendencies: averages, medians, totals. But many real client problems live in the spread of the data: the long tail of unprofitable customers, the few plants with very high downtime, the extreme service delays that damage NPS despite a decent average.
Distribution and density charts help you show how a variable is spread out, not just where the middle is. They make variation, skew, and outliers visible and force more honest conversations about risk and performance gaps.
10.1 When the story is about distribution, not just averages
You should reach for distribution charts when the client’s questions sound like:
- “Why are some branches so much worse than others?”
- “Is this risk really small and rare, or does it happen often?”
- “Is performance consistently improving, or just in a few pockets?”
Key signals you’re in “distribution territory”:
- The average looks fine, but complaints, losses, or risks still feel high.
- Stakeholders talk about “tails” or “extremes” (“a few very late deliveries kill us”).
- You are segmenting into quartiles or deciles (top 10%, bottom 10%) to drive actions.
In these cases, a single bar or line on an average can be misleading or incomplete. You want to show:
- The shape of the data: symmetric vs skewed, single peak vs multiple peaks.
- The range: minimum to maximum, typical spread.
- The concentration: where most observations sit.
- The outliers: rare but impactful extremes.
Once you see that the story is about spread, not just center, histograms, density charts, and range-style visuals become natural tools.
10.2 Histograms (frequency distributions)
A histogram is the simplest way to show a distribution. You divide a continuous variable into bins (ranges) and count how many observations fall into each bin. Each bar shows the frequency (or sometimes percentage) of observations in that range.
Examples:
- Order value: how many orders fall into $0–$10, $10–$20, etc.
- Resolution time: how many tickets are solved in <1 day, 1–3 days, 3–7 days, >7 days.
- Loan size, project duration, on-site time per visit.
Use histograms when:
- You have a large number of observations (dozens to thousands).
- The variable is continuous or quasi-continuous (time, money, size).
- You want the audience to see where the bulk of cases sits and how long the tails are.
Design choices that matter:
- Bin width:
- Too few bins → distribution looks over-smoothed and hides structure.
- Too many bins → charts looks noisy and random.
- In practice, pick a bin width that yields roughly 8–20 bins across the range, then adjust for readability.
- Scale:
- Use counts when the audience cares about absolute numbers (“how many customers?”).
- Use percentages when comparing two groups or samples of different sizes.
- Labelling:
- Make it clear what each bin represents (“Days to resolve: 0–1, 1–3, 3–7, >7”).
- Highlight the tail segments if they are the story (“these 5% of tickets take >14 days and damage NPS disproportionately”).
A clean histogram turns a vague concern (“we have some long delays”) into a precise framing (“15% of deliveries take more than 10 days”).
10.3 Density charts (smoothed distributions)
Density charts are a smoothed version of histograms. Instead of bars, you see a continuous curve that estimates how likely different values are. Technically these are kernel density estimates, but you do not need to use that term with clients.
Use density charts when:
- The audience can handle a slightly more technical visual.
- You want to compare distribution shapes between groups (e.g., before vs after, segment A vs segment B).
- You care less about the exact count in each range and more about the overall shape (e.g., one peak vs two peaks, where the mass is shifting).
Typical applications:
- Comparing lead time distributions before and after a process improvement.
- Comparing transaction size distributions across customer segments.
- Comparing salary distributions across functions or regions (when appropriate).
Design guidance:
- Use one or two density curves on a chart; three is usually the upper limit for clarity.
- Pick a smoothing level (often controlled by a “bandwidth” parameter in tools) that does not introduce artificial bumps or hide real ones. If the curve looks too jagged or too flat, adjust.
- Clearly label which curve is which and highlight the one that carries your story.
Because density charts are less familiar to some executives, pair them with a strong title that explains the key takeaway (“Service times are less volatile and more centered around 2–3 days after redesign”). If needed, a small histogram inset can help ground the concept, but avoid overcomplicating the slide.
10.4 Box-and-whisker and range-style alternatives (where appropriate)
Box-and-whisker plots (boxplots) and related range charts are powerful summary views of distributions, especially when you need to compare multiple groups side by side.
A standard boxplot shows:
- The median (line inside the box).
- The interquartile range (IQR: 25th to 75th percentile = the box).
- The whiskers, which usually extend to a defined range of the data (often 1.5 × IQR or to min/max).
- Sometimes outliers, plotted as individual points beyond the whiskers.
Use boxplots or range-style charts when:
- You are comparing many groups (e.g., 10 regions, 15 plants) and want a compact view of central tendency and spread.
- The audience is at least somewhat data literate, or you have time to explain the chart once.
- The questions are about variation between and within groups: “Which plants are consistently good? Which are volatile?”
If your client audience is less familiar with boxplots, consider simpler range-style alternatives:
- Min–max ranges with a marker for median or mean (e.g., a vertical line from min to max with a dot or bar at the middle).
- Band charts: showing a shaded band for the range (or 10th–90th percentile) over time with a line for the median.
These alternatives trade some statistical richness for interpretability, which is often a good bargain in executive settings.
Whichever style you choose, make sure you:
- Explain in one sentence in your script what the elements mean (“the box shows the middle 50% of observations; whiskers show the full range; dots are outliers”).
- Order groups meaningfully, not alphabetically (e.g., by median performance or by size).
- Focus the title and your narration on contrast (“plants A and B have similar average yield but very different consistency”).
10.5 Population pyramids as distribution charts
Population pyramids are a special case of back-to-back column charts that show a distribution across ordered bands, split by two groups. While they are commonly used for age and gender in demographics, the pattern is more general.
Think of a population pyramid as a distribution mirror:
- The vertical axis is an ordered variable (age band, tenure band, size band, grade level).
- Each horizontal bar shows the count or percentage in that band for one group, to the left or right.
Consulting-friendly uses:
- Workforce structure: grade/tenure distribution by region or function (e.g., “EMEA is top-heavy; APAC is junior-heavy”).
- Customer size distribution: small vs large customers across revenue bands.
- Any two-group comparison where the shape, not just the total, matters.
Design tips:
- Keep the band definitions clear and consistent (“0–2 years,” “2–5 years,” etc.).
- Use the same scale on both sides so lengths are comparable.
- Use related but distinct colors for the two sides to show they are paired.
When you present a population pyramid, explicitly frame it as a distribution story: “This chart shows where our people sit by grade; notice how one region has far more senior staff in the upper bands than the other.”
10.6 Common pitfalls and good practices for distribution charts
Distribution charts are easy to overcomplicate. Before sending one to a client, check for a few classic issues.
Common pitfalls:
- Only showing averages when the debate is clearly about variability or extreme cases.
- Choosing bin sizes in histograms that hide or exaggerate structure (e.g., one bin covering almost the entire range).
- Drawing density charts so smooth that important bumps disappear, or so jagged they look like noise.
- Using boxplots with an audience that has never seen them and providing no explanation.
- Overplotting too many groups or segments on one density or range chart.
- Mixing different definitions of spread (e.g., standard deviation on one slide, IQR on another) without making it explicit.
A quick checklist you can apply:
- Is the question truly about distribution (spread, tails, consistency) rather than just levels?
- Does the chart make the shape and range of the data obvious in a second or two?
- Have I limited the number of groups or overlays to what a human can parse quickly?
- Is the choice of binning, smoothing, and range reasonable and not misleading?
- Could I explain this chart’s structure to a senior stakeholder in one short sentence?
If those answers are yes, you are using distribution and density charts the way we intend in the Firm: to show the real variability behind the averages, and to focus attention where it matters—on the tails, the outliers, and the structural differences that often drive the recommendations you are about to make.
