- Use a scatterplot to explore how two measures relate and to shape better follow-up questions.
- Begin with the axes, labels, and what each dot represents; the reading section covers the evidence-backed checks.
- Treat visible patterns as prompts for investigation, and review the causation section before making causal claims.
- Choose added visual cues only when they answer your question and remain easy to explain.
When two metrics seem connected, a scatterplot helps you check the relationship without flattening the data into a single average. Each dot represents one observation with two numeric values. You can see whether the values move together, form groups, curve, or produce a few surprises.
The chart works best when you begin with a focused question. Once you know what each dot represents and why the two variables belong together, the pattern becomes much easier to read and explain.
What Is a Scatterplot and When Should You Use One?
A scatterplot is a chart of paired numeric values plotted along two axes. One value determines a dot's horizontal position on the x-axis, while the other determines its vertical position on the y-axis. If one dot represents a store, for example, x might show weekly foot traffic and y might show weekly sales.
Why plot the values rather than compare two columns in a table? The chart gives you a compact view of how the observations relate. You can quickly look for a general direction, a curved or straight form, tight groups, gaps, and points far from the rest.
Scatterplots are especially useful when your question involves a possible relationship between two measures. An analyst might compare campaign spend with qualified leads, while an operations manager might compare order volume with fulfillment time. A data-visualization practitioner can also use the chart to check whether a proposed trendline fits the actual pattern.
The chart helps you examine four common features. Use this checklist before drawing conclusions:
Relationships: Do higher values of one variable tend to appear with higher or lower values of the other?
Correlation: How consistently do the points follow a shared direction or form?
Clusters: Do observations gather into groups that may deserve separate analysis?
Outliers: Which points sit apart from the main pattern and may need context?
You will often hear scatterplots discussed alongside correlation, but the visual does more than hint at a straight-line relationship. It can reveal curves, changing spread, and separate groups that one correlation value could blur. That broader view is one reason scatterplots remain a useful part of a visual analytics guide.
Consider another chart when your question centers on categories, a time trend, or parts of a whole. A scatterplot is most useful when each observation has two meaningful numbers and your question depends on their relationship.
How Do You Read a Scatterplot?
A busy field of dots can look like confetti with a spreadsheet habit. Give it a minute. A consistent reading order keeps you from chasing the most dramatic point before you understand the chart.
Start with the axes and paired values
First, read both axis labels, units, ranges, and scales, then ask what one dot represents. According to CDC’s scatter plot guidance, each dot needs paired values; label the variables clearly, and explain visible outliers when context matters. These checks help readers understand what the pattern represents.
Clear labels matter because the same shape can support very different interpretations. A chart comparing tenure with spending differs from one comparing age with spending, even when the dots look similar. Units also shape meaning, so check whether values are percentages, counts, dollars, minutes, or another measure.
Next, check whether the displayed ranges help you read the observations in relation to the question. If a scale choice could surprise readers, explain it in plain language.
Finally, confirm that the paired values belong to the same observation and time frame. Mixing one store's January traffic with its February sales would create a dot, but the pairing would not answer the intended question.
Read direction, form, strength, and outliers
Start with direction. An upward pattern suggests that higher x values tend to appear with higher y values, while a downward pattern suggests the opposite. A point cloud with no obvious direction may show little visible association, or it may contain a pattern that requires another view.
Then inspect form. Does the pattern look roughly linear, curved, segmented, or clustered? A curve can show a consistent relationship even when a straight trendline tells a weak story, so let the points speak before adding a line.
Strength describes how closely the dots follow the apparent form. Tightly grouped points suggest a more consistent association, while a wide spread suggests more variation. Avoid universal labels based on fixed numerical cutoffs because the practical meaning depends on the data and the decision you face.
Last, investigate outliers rather than deleting them on sight. An unusual point could reflect a data-entry problem, a special event, a distinct segment, or a valid rare case. Add a note when readers need that context, and document any exclusion from the analysis.
Correlation Is Useful, but It Is Not Causation
A scatterplot can show two variables moving together, but it cannot tell you that one produced the change in the other. The National Institute of Standards and Technology (NIST) describes scatterplots as tools for revealing association in NIST’s scatter plot reference. NIST also states that association alone does not establish causation.
Suppose stores with more staff also report higher sales. The pattern may reflect a staffing effect, but larger stores could have both more employees and more customers. Location, store size, season, promotions, or other variables could also shape the relationship.
Observed associations can reflect chance, bias, confounding, or limits in study design, execution, or analysis. CDC’s epidemiology guidance notes that statistical significance alone does not indicate cause and effect. Treat these as general possibilities, not diagnoses of your data.
Reverse direction can matter too: higher demand may lead managers to schedule more staff, rather than staffing creating the demand. A causal conclusion calls for evidence beyond the chart, such as a suitable study design or analysis that addresses credible alternatives. You should also state the assumptions and limits clearly.
Correlation remains useful because it tells you where to look next. It can challenge an assumption, suggest a segment to investigate, or show that a simple average misses an important pattern. Following sound data visualization principles helps you present that evidence without asking the chart to claim more than it can support.
How Do You Create a Scatterplot That Answers a Real Question?
Good scatterplots begin before you choose a color or marker shape. Start with the decision or uncertainty behind the chart, then build the visual around what the reader needs to compare.
Frame the question. Name the observations, two measures, and relationship you want to inspect. Ask, “How does delivery distance relate to delivery time?”
Choose paired numeric variables. Each dot needs two values from one observation. Use matching time frames, populations, and levels of detail.
Select the axes. Put a defensible explanatory variable on x and the outcome on y. Otherwise, explain your arrangement without implying cause.
Plot points and labels. Add specific axis titles, units, and a title that states the question. Use a clear legend for added encodings.
Inspect, then annotate. Check direction, form, spread, clusters, and unusual points. Label only details that help readers interpret the pattern.
After plotting, step back and ask whether the chart answers your original question. If the view feels crowded or ambiguous, simplify the presentation or narrow the question before adding more detail. Avoid polishing the chart into a confident story before you have checked the data behind it.
The most useful data visualization best practices come back to the same idea: make the reader's task obvious. If someone cannot tell what a dot represents or why the variables appear together, visual polish will not rescue the analysis.
When Should You Use a Variation Instead?
A basic scatterplot gives every observation an x position and a y position. Sometimes your question also involves a third measure or an ordered sequence, so an added visual cue may help. Use that extra cue only when it helps answer the question and remains easy to explain.
The United Nations High Commissioner for Refugees (UNHCR) covers both variations in UNHCR’s chart guidance. A bubble chart extends a scatterplot by using bubble size to represent an additional quantitative variable. A connected scatterplot links sequential points, often time periods, to show how two measures move in relation to each other.
Bubble Charts
You might plot product price on x, units sold on y, and revenue as bubble size. Use this choice only when the third measure helps answer the question and its role remains easy to explain.
Connected Scatterplots
The connecting line gives readers a path through the observations. Use it only when sequence matters to the question and the path remains easy to explain.
Bubble charts and connected scatterplots are two options in a broader guide to chart types. Pick the variation that matches your question, then give readers enough guidance to understand the added cue.
Move from a Chart to a Useful Follow-Up Question
A scatterplot often answers the first question and raises a better one. You may notice that the relationship changes by region, customer segment, product line, or time period. You might also find an outlier worth checking against source records or business context.
That next question is where a static chart can feel limiting. With interactive data visualization, you can filter the plotted observations, inspect individual points, and drill into a segment without rebuilding the chart for every angle. The goal is focused exploration, not interaction for its own sake.
For example, a sales-versus-discount chart may show a loose downward pattern overall. Filtering by product category could reveal that one category behaves differently, while drilling into an unusual point could show a bulk order or a return. Each step should test a specific explanation rather than send you clicking through data at random.
A useful follow-up question names what changed and what you need to compare next. Try asking, “Does this pattern hold across regions?” or “What distinguishes the outliers from the main group?” Such questions connect the shape to business conditions.
Conclusion
A scatterplot is most useful when it helps you move from a broad hunch to a focused question. The cited reading section shows what to inspect, while the causation section explains why a visible association still needs context and supporting evidence. Keep those guardrails close whenever the pattern looks especially convincing.
Use the chart as a starting point for discussion, comparison, and deeper analysis. When you are ready to explore how search, charts, and follow-up analysis can work together, explore ThoughtSpot. Keep the question in charge, and let each view earn the next one.
