Scatter and bubble charts are where you stop just “reporting numbers” and start showing how things relate: which customers are attractive, which businesses are off the curve, which levers move together, and where there are real trade-offs. They are more demanding than bars and lines, both analytically and visually, but when used well they can carry an entire strategy discussion.
This chapter covers when to use scatter-type charts, how to design the main variants we use in consulting, and what to avoid when relationships and segmentation are the main story.
9.1 When to use scatter charts (relationships, clusters, outliers)
Scatter charts exist to answer one core question: how do two numeric variables move together across many observations?
You use a scatter when:
- Each point is an entity (customer, product, market, store, project).
- You care about how two continuous metrics relate (e.g., margin vs growth; NPS vs resolution time; price vs volume).
- You want to see clusters, patterns, and outliers, not just averages.
They are especially powerful when:
- You suspect a trade-off (higher margin but lower growth, better service but higher cost).
- You want to show segments visually: different colors for regions, customer types, or strategies.
- You need a portfolio view of many items at once to support “focus here, exit there” decisions.
Do not use scatter charts when one or both axes are purely categorical (“Business Unit A, B, C”) or when you only have a handful of items and no real notion of relationship. In those cases, bars or a simple 2×2 table are almost always clearer.
9.2 Simple scatter chart (two-variable relationships)
The simple scatter is your starting point: two axes, many points, no quadrants, no bubbles. Just a clean view of how two metrics relate.
Examples:
- Customer-level: annual revenue per customer (x) vs gross margin per customer (y).
- Store-level: footfall (x) vs conversion rate (y).
- Product-level: discount level (x) vs volume sold (y).
Good practice:
- Choose axes that are meaningful to the decision. If leadership is choosing where to invest, “growth vs margin” is more actionable than “revenue vs number of SKUs.”
- Scale both axes to cover the relevant range without over-zooming. You want the full cloud of points visible, including outliers.
- Label only the critical points: your client, key competitors, or notable outliers. Do not try to label 80 customers.
- If there is a clear trend, you can add a light reference line (e.g., an average or a simple fitted line) but keep it understated; the cloud of points is the main actor.
A good scatter does three things fast: shows whether there is any relationship at all, whether there are meaningful clusters, and who/what sits off the pattern and deserves attention.
9.3 Four-quadrant scatter chart (strategy maps, positioning)
Most “strategy maps” you see in board decks are really four-quadrant scatter plots: same point cloud, but now with reference lines that divide the space into four regions with different strategic meaning.
You add this structure when the decision logic is genuinely 2×2:
- Growth (x) vs margin (y) → “Stars, cash cows, question marks, dogs.”
- Market attractiveness (x) vs competitive position (y).
- Satisfaction (x) vs economic value (y).
Design guidelines:
- Set meaningful thresholds for the vertical and horizontal lines (e.g., 10% growth, 15% margin, industry average NPS). These lines should be grounded in logic or benchmarks, not arbitrary midpoints.
- Label or lightly annotate the quadrants (“High growth / high margin,” “Low growth / low margin”) to make interpretation immediate.
- Make sure the client can easily find their own points. Use a distinct color or border for client entities vs peers.
- Keep the number of points manageable. If you have hundreds of items, consider aggregating (e.g., by segment) rather than plotting every micro-SKU.
A well-designed four-quadrant scatter lets an executive say, in seconds: “Most of our profit is in the upper-left; we’re under-represented in the attractive upper-right; and we should question why we still own everything in the lower-left.”
9.4 Segmented scatter chart (color-coding segments)
In many cases, the real story is not just where the points are but which points belong together. Segmented scatters add an extra encoding—usually color, sometimes shape—to show segments or types.
Examples:
- Customers colored by region or channel.
- Products colored as core, selective, and tail.
- Projects colored by business unit.
Use segmentation when:
- You want to see if certain segments occupy particular zones of the chart (e.g., online customers are high NPS but lower margin).
- You are heading toward segment-specific strategies (“this cluster is prime for upsell; that one needs cost reduction”).
Keep it clean:
- Limit the number of segment colors. Four to six is usually the upper bound for a main slide.
- Use a consistent legend and use the same colors for the same segments across pages.
- Still highlight the client or key entities with a secondary effect (border, label) so they do not get lost in the colored cloud.
Segmented scatters are incredibly powerful for revealing patterns you would never see in averages, but they depend on restraint: too many colors and labels, and the page collapses into noise.
9.5 XY bubble chart (adding a third variable with bubble size)
Bubbles extend the scatter concept by using area to encode a third numeric variable (usually size: revenue, value, volume, population). Each bubble’s x and y coordinates carry two metrics; its area carries “size of prize.”
Classic consulting uses:
- Portfolio maps: x = growth, y = margin, bubble size = revenue or profit.
- Market maps: x = attractiveness, y = share, bubble size = market size.
- Customer maps: x = satisfaction, y = profitability, bubble size = revenue per customer or number of customers.
Design rules that matter:
- Bubble area, not radius, should be proportional to value. Many tools default to radius; check and correct this so that doubling the metric visually doubles the bubble area, not makes it four times as large.
- Avoid a huge range of sizes. If needed, cap or transform extreme outliers or group very small items into “Other,” so some bubbles are not microscopic and others gigantic.
- Provide a size key (small, medium, large example bubbles with labels) so viewers can anchor their perception.
- Limit the number of bubbles on a main slide; 10–20 is often a practical ceiling for a senior audience.
A bubble chart should support statements like: “The three largest, most attractive markets where we under-index are here—let’s focus there.” If you cannot make that kind of statement easily, reconsider the structure.
9.6 Dot matrix chart (grids of dots to compare magnitudes)
Dot matrix (or dot plot) charts use discrete dots in a grid to represent magnitudes, rather than continuous bars or bubbles. They can be thought of as a bridge between charts and iconography.
Two common patterns in consulting:
- Unit-based dot grids: each dot represents a fixed amount (e.g., $10M, 10 customers), and you show several groups side by side. This makes relative sizes and “how many units” visually clear without implying continuous precision.
- Category-by-category dot layouts: a matrix where rows are categories (e.g., segments) and columns are metrics (e.g., products), with the number or darkness of dots in each cell indicating magnitude.
Dot matrices are useful when:
- You want to convey relative scale and countability (“this group is about twice that group”) rather than precise numeric values.
- The “many small pieces” nature of a quantity matters (e.g., many small customers vs a few large ones).
- You are building a compact comparison grid where traditional bars would be too dense.
To keep them readable:
- Define what one dot represents and keep that constant on the slide.
- Avoid too many tiny dots in a single group; if you need dozens, move back to a bar or bubble chart.
- Use a simple, consistent layout with clear labels on rows and columns so the reader can navigate quickly.
Dot matrices are not your everyday chart, but for the right story—especially around countable units and relative load—they can be more intuitive than numbers alone.
9.7 Radar chart (multi-attribute profiles)
Radar charts (also called spider charts) plot multiple attributes on radial axes around a circle, connecting points to form a polygon. They are best for comparing profiles across several dimensions, not for precise measurement.
Typical consulting uses:
- Capability assessments across dimensions (strategy, operations, digital, people).
- Feature or performance comparisons of products across attributes (speed, reliability, usability).
- Organizational health or culture diagnostics across survey dimensions.
Use radar charts when:
- You have 4–8 attributes that form a logical set.
- You are comparing one entity vs a benchmark or at most two or three entities.
- The story is about shape differences: where one profile sticks out or lags, not small numeric gaps.
Design guidance:
- Keep the number of entities low. Two is ideal (client vs benchmark); three is often the upper limit. More than that becomes a tangle of overlapping polygons.
- Label axes clearly and, where useful, include a reference ring (e.g., “industry average”) to orient the reader.
- Use distinct but restrained fills or outlines, often with transparent fills so overlaps can still be seen.
- Emphasize the relative strengths and weaknesses in the title and perhaps with subtle call-outs on 1–2 key axes.
If you need your client to read exact scores off the chart, a radar is the wrong choice. Think of it as a visual “fingerprint” of an entity’s profile, not a precision instrument.
9.8 Common pitfalls and good practices for scatter and bubble charts
Because these charts are more complex, they are also easier to get wrong. Before sending a scatter, bubble, or radar-type chart to a client, run through a quick sanity check.
Common pitfalls:
- Overplotting: too many points or bubbles on one chart, turning it into static.
- Categorical misuse: using scatter with non-numeric or unordered axes, implying relationships where none exist.
- Bad scaling: using different, unlabeled scales across similar charts, or extreme zoom that makes minor differences look dramatic.
- Bubble area errors: scaling radius instead of area, which visually exaggerates big entities.
- Legend overload: too many colors, shapes, or size encodings to decode under time pressure.
- Spurious quadrant lines: arbitrary thresholds that do not map to real strategic cut-offs.
A short checklist for good practice:
- Am I clearly answering a relationship/segmentation/portfolio question—not a simple ranking or trend?
- Are axes and scales clear, meaningful, and consistent with related pages?
- Have I limited the number of points, segments, and encodings to what a busy reader can actually process?
- Can I identify the client or key entities in less than a second?
- Does the title state the main insight, not just describe the axes?
If you can say “yes” to those, your scatter or bubble chart is doing its job: helping a senior decision-maker see patterns in a messy landscape of options—and decide where to look, and act, next.
