A platform strategy slide shows one bubble twice as wide as another. A reader may assume that the larger platform is twice the size. But a circle with twice the radius has four times the area. If the designer mapped store counts directly to radius, the slide has visually amplified the difference before anyone reads the numbers.
EcomToolkit recommends treating chart construction as part of the statistical analysis. A bubble chart can combine platform scale with two other measures, but only when its encodings and population are clear. The fictional examples in this guide demonstrate chart mechanics; they are not current market-share estimates for any named vendor.
Table of Contents
- Decide whether a bubble chart answers the question
- Map value to area rather than radius
- Keep the scale consistent across panels and dates
- Define the population behind platform size
- Prevent overlap from hiding the smaller platforms
- Handle zero, missing and negative values separately
- The EcomToolkit view
Decide whether a bubble chart answers the question
A bubble chart usually places two numerical variables on the axes and encodes a third through circle area. A platform analysis might use adoption growth horizontally, operating cost vertically and observed store count as bubble size. Each dimension needs a defined unit, reporting period and source population.
If the only question is which platform has more stores, a sorted bar chart or table is often easier to read. Bubbles earn their space when the relationship among several variables matters. Adding a third visual dimension without a decision behind it creates complexity rather than insight.
Avoid treating chart position as a recommendation by itself. A low-cost, fast-growing platform may not support the workflows a merchant needs, and the operating-cost estimate may depend on store size. Write the commercial question above the chart so readers know what the display can and cannot help decide.
The same discipline applies to a packed-bubble graphic without axes. That display may communicate broad scale, but it makes precise comparisons difficult. Keep exact values available and resist using attractive circles as the only evidence for a migration recommendation.
Map value to area rather than radius
Circle area equals pi times radius squared. To make area proportional to a positive value, radius must be proportional to the square root of that value. If 100 stores receive a radius of 10 pixels, 400 stores should receive a radius of 20 pixels, producing four times the area.
If radius were scaled directly with count, 400 stores would receive a radius of 40 pixels and sixteen times the area. The data increased fourfold, but the display increased sixteenfold. This is a geometric distortion, not merely an aesthetic disagreement about which bubble looks more balanced.
| Fictional observed stores | Correct proportional radius | Radius mapped directly to count | Misleading area relative to 100 stores |
|---|---|---|---|
| 100 | 10 pixels | 10 pixels | 1 times |
| 400 | 20 pixels | 40 pixels | 16 times |
| 900 | 30 pixels | 90 pixels | 81 times |
The D3 power-scale documentation describes square-root scales. When implementing proportional bubble area, use a zero-based mapping such as radius equals a constant times the square root of value. Adding a positive radius offset changes the area ratios, so a square-root function alone is not sufficient if its range starts above zero.
Keep the scale consistent across panels and dates
Suppose a report shows two countries side by side. If each panel independently scales its largest platform to the same maximum bubble radius, equally sized bubbles no longer represent equal store counts. That may be intentional for within-country comparison, but the visual needs an explicit warning and separate legends.
For direct comparisons, use a common size scale across the panels. Include reference bubbles with labelled values, and verify that the legend uses the same function as the data marks. A manually drawn legend can look reasonable while disagreeing with the actual chart.
Apply the same rule to a monthly animation or slide sequence. Recomputing the maximum radius each month can make a growing platform appear unchanged in size. Either preserve the scale over the comparison period or tell the reader that the view is rescaled and provide the underlying values.
Do not add a minimum visible bubble without accounting for the distortion. A floor can help small observations remain selectable, but it breaks strict proportionality for those values. Prefer a clearly explained marker treatment or an accompanying table rather than pretending that all circle areas remain exact.
Define the population behind platform size
A correctly scaled chart can still mislead if bubble values mix incompatible definitions. Detected websites, active merchants, storefront domains and transaction volume are different measures. A platform with many small stores can dominate one metric while looking smaller under another.
State the unit directly in the size legend: for example, observed eligible storefront domains in a defined dataset and month. If a provider cannot detect every technology equally, describe that limitation. Do not silently relabel an observed technology count as a complete census of businesses using the platform.
Use the platform market-share denominator guide before combining counts and shares. If stores can use multiple technologies, the categories may overlap. Circle areas representing overlapping counts should not be presented as mutually exclusive pieces of one market.
When size represents monetary value, make the currency and time basis explicit. Annual transaction volume and monthly subscription revenue cannot share a common size legend merely because both are money. The meaning of the variable matters before its geometry can be evaluated.
Prevent overlap from hiding the smaller platforms
Large circles can conceal smaller ones positioned nearby. Transparent fills may help reveal overlap, but they do not guarantee readability and can create confusing blended colours. Inspect the chart at the size people will actually use, including a mobile screen or a projected presentation.
Consider drawing larger circles first so smaller ones remain visible, while providing a way to inspect each platform’s exact values. Labels should identify the point without suggesting that the label’s position is the actual data coordinate. If labels need connectors, keep those connectors visually separate from the data marks.
Do not move bubbles arbitrarily to avoid collisions if their positions encode numerical axes. Jittering or repelling the circles changes the displayed values unless the displacement is clearly handled. A label can move while its anchor remains fixed; the measured point should not casually drift to a more attractive location.
If the view becomes unreadable, simplify it. Filter by a stated eligibility rule, use small multiples with consistent scales, or replace the chart with a table. Excluding inconvenient platforms without disclosing the rule is not a legitimate way to improve visual clarity.
| Review question | Acceptable evidence | Reason to revise |
|---|---|---|
| Does area match the stated variable? | Checked radius formula and reference values | Radius is directly proportional to value |
| Are panels comparable? | Shared scale or explicit separate legends | Identical circles mean different counts |
| Can every platform be inspected? | Readable labels and value table | Large marks completely hide smaller ones |
| Is the population defined? | Unit, date and eligibility note | Generic market size label |
| Are zero and missing distinct? | Separate documented states | Missing values displayed as zero |
Handle zero, missing and negative values separately
A zero observation has zero proportional area, so it may disappear from the plot. That is mathematically consistent but potentially confusing. Keep the platform in the accompanying table and explain whether zero means no observed eligible stores or a rounded value below the publication threshold.
A missing value is not zero. If the provider has no measurement, display an unavailable state in the table and avoid inventing a bubble. Likewise, a negative growth rate can be represented on an axis, but a negative count cannot be represented as ordinary circle area.
Avoid using signed growth as bubble size unless the visual has a carefully explained separate encoding for sign and magnitude. A reader should not need to guess whether a large bubble means strong growth, a severe decline or merely a large absolute number.
The platform growth-rate guide explains another trap: a small base can produce a large percentage increase. Showing growth on an axis and scale through area can help reveal that distinction, provided both measures describe the same population and compatible dates.
The EcomToolkit view
A trustworthy platform chart should survive a simple audit with a calculator and its source table. Confirm the area ratios, scale consistency and population before interpreting the pattern. Then decide whether the display helps a real platform decision or whether a clearer table would serve the reader better.
Bring a platform comparison slide and its underlying values to an EcomToolkit platform analysis review. The useful outcome is a defensible interpretation of the data, with the visual helping the evidence rather than magnifying it.
