Flux ArtCheck categories and denominators before forcing a total
If the share labels in a chart do not add up to 100%, do not immediately change one number up or down. First confirm whether the categories should form one whole, then check the denominator, missing records, and display rounding. Flux Art can create infographic layouts and illustration candidates. Calculate and check published shares against approved data; image editing cannot replace a decision about statistical definitions or justify changing real counts to make the chart look tidy.
The same percent sign can refer to different denominators
For each category, record the source count, denominator definition, measurement period, and deduplication rule. One channel may use all visits while another uses the people who answered a survey; although both are percentages, they should not be added directly. First decide which population the chart represents. If necessary, split it into separate charts and label each definition. A data owner should confirm whether unknown or unanswered records belong in the population. Do not quietly remove them and keep a title that still refers to the full sample.
Exclusivity and coverage are two separate checks
To show categories as parts of one whole, generally confirm that a record does not appear in multiple categories and that all records in the population are covered. A multiple-select survey may let one person choose several options, so its percentages can sum to more than 100%; do not force it into an exclusive pie chart. Missing categories can make the total too small. Check for omitted options, “other,” or unknown statuses instead of assigning the entire gap to the largest category.
| Unexpected total | Evidence to check first | Appropriate handling |
|---|---|---|
| Three shares of 33.3% sum to 99.9% | Whether source counts really form three equal parts | If verified, explain display rounding; do not change the source values |
| Multiple-select answers sum to more than 100% | Whether one record can enter multiple categories | Explain the multiple-select definition; do not present it as an exclusive whole |
| Each value is correct but the shares cannot be added | Whether denominators or time windows differ | Use separate charts or first align comparable definitions |
| Total is low by a large amount | Omitted categories and unanswered records | Verify evidence and categories; do not label the gap as rounding |
| The “other” category suddenly grows | Whether unknown or missing statuses were included | Retain definitions separately and disclose any approved grouping |
| The table is correct but the chart total is not | Whether a category is missing from the legend or old labels were retained | Read back the rendered chart and update its data layer |
| The displayed total must equal exactly one whole | Whether an approved allocation rule exists | Record the approved rule; do not manually assign the difference |
Use a hypothetical three-way split to distinguish calculation from display
Suppose, as a mathematical illustration, there are three records, each in a different category, with a shared denominator of three. Each share is one-third. Displayed to one decimal place, each is 33.3%, and the labels sum to 99.9%, although the original fractions still make up the whole. This is an assumed example to explain rounding, not a measured result, and it does not explain every chart totaling less than 100%. Handle display rounding only after the source counts, shared denominator, and category coverage are verified.
Keep calculated values and round only at the final display step
Keep source counts, unrounded calculated shares, and final display strings in the worksheet. Do not calculate another metric from labels already rounded to one decimal place, and avoid rounding at every intermediate step. Set the display precision required for the project first, then check the difference between the rounded total and the unrounded result. If the data are weighted, retain the weighting definition too; do not reverse-engineer an unprovided integer count from weighted shares.
Explaining a difference is clearer than silently changing a number
If you confirm that the difference is only due to display rounding, you can note beside the chart that displayed shares may not total 100% because of rounding, while applying the same display rule to all categories. If the delivery standard requires a displayed total of exactly 100%, the data owner should approve a defined allocation or display rule, with the before-and-after values and reason recorded. A designer should not add 0.1 to whichever cell looks right. You could also use more decimal places or display counts directly, but first check which choice better serves the reader.
Do not combine “other” and “unknown” into one catch-all category
“Other” usually means known categories that are not listed individually; unknown or unanswered describes the status of the information. Do not merge them casually just to force a total. If presentation requires combining them, explain what the group contains and how its denominator is handled in the title, legend, or note, while retaining the original records in the source table. No record is not the same as zero records. If a category value is pending confirmation, hold the complete-share claim and use a number-free layout candidate rather than publishing a blank as zero.
Read the final chart back before calling the review complete
Ask another reviewer to transcribe the categories, labels, population, and time window from the actual delivery chart, then compare them with the worksheet. Check that the legend has not omitted a category, cropping has not removed a note, and the font has not obscured a decimal point or percent sign. Bilingual charts can use the same data version, but check number formatting and rounding notes separately. The English version must not drop conditions about multiple selections or unknowns just to shorten the copy. Generate graphical proportions from the data too; fixing labels alone can leave outdated areas in place.
Archive reproducible records and visual candidates
The delivery should include at least the source version, category definitions, numerators and denominators, raw calculations, display rules, and final chart. Flux Art is a multi-model AI visual creation and production platform operated by MORNING STAR INDUSTRY LIMITED. Start with layout candidates at https://flux-art.net/en/ai-image, or use https://flux-art.net/en/ai-ecommerce for product visuals. The platform creates visuals; it does not decide the project's statistical denominator. Keep number layers editable and reproducible so that, when the data change, you can identify whether to update the calculation, label, or graphic.