CompTIA Data+ DA0-002 · Free study guide

Choosing the right chart, and the ways a chart lies

Chart selection begins with one thing: what question is the chart being asked to answer? Once that is clear, data type, audience, and accessibility constraints narrow the defensible choices.

Match the chart to the question

Nearly every selection question fits one of these.

Two of these deserve extra elaboration.

Line charts require an ordered dimension, normally time. Connecting unordered categories with a line implies a progression between them that does not exist — a classic wrong answer.

Bar charts compare along a common baseline, which is the most accurate comparison human vision performs. That is why they beat pie charts almost always, and why a strong answer for "compare 15 categories" is a sorted horizontal bar chart: sorting makes rank readable and horizontal orientation keeps long labels legible.

The pie chart rule

A pie chart is most defensible when it shows parts of one whole with only a few clearly distinguishable slices. As category count grows, angle comparison becomes difficult and a sorted bar chart is generally clearer. Use another visual when values are not components of one total or when the purpose is to show change over time.

Histogram is not a bar chart

They look similar and test differently. A bar chart compares discrete categories; gaps between bars are meaningful because the categories are separate. A histogram shows the distribution of one continuous variable by binning it; the bars touch because the underlying scale is continuous.

If the scenario says "shape of the distribution," "where values cluster," "is it skewed," or "does it have more than one peak," it wants a histogram. If it wants median, quartiles, and outliers — especially across several groups — it wants box plots.

Four common ways charts mislead

Truncated axes

A bar chart should generally use a zero baseline because bars encode magnitude through length. Starting the axis at 95 makes a change from 96 to 98 fill most of the plot and reads as dramatic. If a specialized chart breaks the axis, the break must be unmistakable and the resulting comparison still needs scrutiny.

The nuance worth carrying: line charts may legitimately use a narrower range, because a line encodes change and slope rather than magnitude. The rule is about the encoding, not about axes in general.

Unnormalized maps

Colouring US states by a raw count can largely reproduce population size. If the question is prevalence or relative incidence, normalize to a defensible rate such as per capita, per household, or per customer. Raw counts can still be appropriate when total workload or volume is the actual question.

Dual axes

Two series on independent vertical scales can be made to appear correlated or divergent almost at will, because the crossing points are artifacts of the scaling choices. Safer alternatives are indexing both series to a common baseline or using two aligned panels.

Selective omission

Plotting the three months that rose and omitting the two that fell, then labelling the chart "growth," misleads even though every plotted number is correct. Completeness of the window is part of honesty. Watch for scenarios where the data is accurate and the selection is the problem.

Colour, and the accessibility answer

Encoding a critical distinction in red versus green alone fails for viewers with red-green colour vision deficiency, the most common form. The robust fix is a redundant, non-colour cue — a label, an icon, a shape, a pattern — not "increase the saturation" and not "switch to greyscale."

Choosing a palette follows the data's structure:

• Sequential — one hue, light to dark — for magnitude that runs in one direction.
• Diverging — two hues meeting at a neutral midpoint — when there is a meaningful centre such as zero and deviations both ways matter. Profit margin is the canonical case.
• Qualitative — distinct unrelated hues — for unordered categories, chosen precisely because it implies no ranking.

Using a diverging scale on strictly positive magnitudes, or a sequential scale on unordered categories, invents structure the data does not have.

Dashboards

Three principles make dashboard decisions more defensible.

A dashboard is defined by its audience's decisions. An executive view scanned for thirty seconds should surface a few high-level indicators above the fold, with detail on drill-down. Twenty-five charts across four scrolling screens is a failure of prioritisation, not thoroughness.

One metric, one definition. Two tiles both labelled "active users" showing different numbers destroys trust in everything else on the page. The fix is a certified, centrally maintained definition that every visual references — not a disclaimer, and not deleting one tile while the divergent logic survives elsewhere.

Drill-down goes from summary toward detail; roll-up goes the other way.

When a dashboard is slow because every tile scans raw detail, the answer is pre-aggregated summary tables or materialised views on a refresh schedule. That makes freshness an explicit decision instead of an accident.

Delivering the result

Executive audiences get the conclusion and its business implication first, supported by two or three clear visuals, with methodology available as backup. Leading with methodology spends the whole meeting before reaching the point, and handing over raw data transfers your job to the audience.

Two delivery details matter: when a metric's definition changed mid-year, annotate the break so readers do not mistake it for a real trend; and when recipients should not see certain rows or columns, enforce that with row- and column-level security rather than a shared file and a request not to look.

What to drill

Read the question the chart must answer, then pick from the table. Check the axis baseline on every bar chart, check for a rate rather than a count on every map, and check whether the plotted window is complete. Know the three palette families and the redundant-cue rule. Apply the pattern to a scenario instead of selecting a chart by habit.