Tableau / Basic: fields, models and views / Lesson 6

Charts, sorting and missing data

Choose encodings and ordering that preserve meaning when categories, periods or values are absent.

About 4 min readContent reviewed 2026-10-28

On this page

Lesson goals

Match chart form to the question.
Separate absence, null and zero.
Review scales, sorting and subset labels.

Start with the question

Bars support category comparison, lines emphasize ordered progression and scatter views can reveal relationships between marks. Choose from the reading task, not the chart gallery’s size. A precise small table can be better when exact amounts matter. Color, size and position need defined units and populations. A second axis can introduce incompatible scales without careful design. A valid view can still mislead if readers cannot identify metric meaning, exclusions, date grain or selection. Review those decisions before decoration, and retain a simple expected-value table as an independent reference.

Make ordering explicit

Alphabetical month names do not establish chronological order across years. Use a date or explicit sort key and distinguish category ranking from a temporal trend. The first record describes three months with no February row. It is a paper design, not a Tableau setting. Label a Top N subset and do not confuse its total with the entire population. When selection changes ranking, verify the intended scope and secondary tie rule remain clear. A moving bar can reflect changed context, not changed underlying business activity, so explain the interaction as well as the sort.

text Synthetic time-axis contract

order:2026-01,2026-02,2026-03
observed:January10,March20
February:absent source row
contract:explain missing coverage before interpreting a trend

Source absence is distinguished from observed zero. No missing-value display is configured.

Separate absence, null and zero

An absent row means no record for a domain member; null means a field is missing or undefined within a row; zero is a numeric observation. Showing missing dates or using a scaffold can expose the domain without proving inactivity. ZN or another null replacement changes meaning unless zero is justified. The second program places None between observed months. It neither implements Tableau densification nor proves connector support for display options. Decide whether the gap represents unavailable extraction, unknown activity or an approved no-event case before interpolation or replacement.

python Synthetic missing-period oracle

observed = {"2026-01": 10, "2026-03": 20}
months = ["2026-01", "2026-02", "2026-03"]
print([observed.get(month) for month in months])
print(sum(observed.values()))

Expected output

[10, None, 20]
30

[10,None,20] retains unknown February;30 covers observed months only. No visual interpolation occurs.

Use honest scales and labels

Bar length implies baseline comparison while a line emphasizes change; axes should support those readings transparently. State units and avoid an unexplained scale mixing percentages and amounts. Labels and tooltips should clarify population and missing coverage, not repeat a number without context. Color should not carry status alone. Tea 42 and Coffee 18 can be accurate yet misleading when called order counts. Validate metric name, aggregation and encoding together. A percentage needs its denominator, and a blank should not appear as verified zero simply because the layout prefers complete cards.

Review incomplete cases

Test default view, one category, several periods, no matching data and a deliberately missing period. Compare expected totals independently of appearance and explain whether a gap, zero or omitted mark is intended. This course creates no worksheet or axis setting. Its dictionary model verifies source coverage only. Actual completion, sort and connector behavior need an authorized Tableau check. A visual contract should expose uncertainty so a smooth line or empty cell cannot become a false business conclusion. Keep any imputation rule documented with the data owner rather than embedded silently in a formatting decision.

Common errors and fixes

Missing February is called zero sales

Why it happens: Source absence was confused with inactivity.
How to fix it: Expose missing-data policy and justify any zero replacement.
Prevent it next time: Keep absent, null and zero cases separate.

Month labels create a false trend order

Why it happens: Labels replaced chronological keys.
How to fix it: Sort using the documented date/month key across years.
Prevent it next time: Test a year boundary.

A subset is labeled as the full population

Why it happens: Selection or Top N scope was omitted.
How to fix it: State included population and denominator deliberately.
Prevent it next time: Keep scope labels aligned with interactions.

Check your understanding

1. Why is filling February with 0 a business decision?

Show worked solution

No February observation could mean incomplete extraction or genuinely no events. Zero asserts an observed lack of activity and changes averages and trends. Confirm the source contract and expose uncertainty first. A continuous-looking chart does not establish that an inferred value is true.

2. How should a top-product chart and overall total be explained together?