figran

A rounded percentage is not an exact count: a chart-reading lab

By Figran Team. We are associated with Figran, a commercial AI-assisted visual creation service. This original teaching activity was prepared with AI assistance and checked for internal arithmetic consistency. Every count below is invented for the exercise. There are no measurements, classroom trials, or claims about improved learning or product performance. No software account is needed.

A chart label can display a percentage without identifying an exact count. A label can also be derived correctly even when independently rounded category labels do not add to 100.0%. The useful question is which source values and rounding rule could have produced the display.

This lab separates three things: an assigned integer count, its exact share of a declared total, and a rounded percentage label. It does not treat extra decimal places as evidence of measurement precision. For real research, check the actual source and reporting rules; the fictional contract here cannot establish them.

The display contract

Use a positive integer total n and an integer category count m with 0 <= m <= n. The exact percentage is 100 × m / n. A category count is not rounded before this calculation.

For this activity, round each percentage independently to the nearest tenth of a percentage point. At an exact halfway value, round upward. Thus 9.95% displays as 10.0%, while 10.05% displays as 10.1%. This is the declared half-up rule for nonnegative percentages, not a claim about every spreadsheet or charting tool.

Keep exactly one digit after the decimal point in a displayed label. Do not adjust a label merely to make the displayed total 100.0%. A different reporting policy would be a different contract and would need to be stated.

Two complete invented cards

Each card has three mutually exclusive categories, and those categories exhaust that card's total. The category names do not identify real people, species, or measurements.

Category Card A assigned count Card B assigned count
Larch 1 1
Pine 1 1
Spruce 1 4
Total 3 6

Use each card's own total. Card B's total is not a denominator for Card A. The source table supplies exact assigned counts, so there is no missing category in either card.

Six review questions

  1. Calculate the three displayed percentages for each card. What do their displayed labels add to? Does either sum prove that an item or a category is missing?
  2. A separate fictional card has total n = 10,000 and one displayed category label of 10.0%. The original category count is not supplied. List every integer count consistent with the contract. Then repeat with total n = 1,000.
  3. With total n = 10,000, what counts are consistent with a displayed 0.0%? What counts are consistent with a displayed 100.0%? State whether either label determines the exact endpoint count.
  4. A reviewer changes Card A's Spruce label to 33.4% so the three labels add to 100.0%. Does that revised label follow this card's declared independent half-up rule? Explain what would have to be disclosed if a different label-allocation policy were used.
  5. A caption supplies only “10.0%” and the rounding rule, with no total n. Can a reader recover the category count by multiplying 10.0 by an assumed total? Identify the missing source information.
  6. An additional card has total n = 10,000 and exactly two exhaustive categories, each labelled 50.0%. Find the possible ordered count pairs. Explain why choosing any permitted count for each label independently can violate the known total.

Try the questions before reading the answers. Record the total, the rounding rule, and the source constraints with each answer. You do not need to count 10,000 objects; these are arithmetic assignments.

Worked answers

1. Correct labels can sum to 99.9% or 100.1%. Each Card A category has exact share 100/3%, which displays as 33.3%. Three such labels add to 99.9%. Card B displays 16.7%, 16.7%, and 66.7%, which add to 100.1%. Its exact shares still add to 100%. Both source counts add to their declared totals. The discrepancy arises from adding independently rounded displays; it does not establish missing or extra source items.

2. Reverse the rounding interval before choosing a count. A 10.0% label under this contract means 9.95% <= exact percentage < 10.05%. With n = 10,000, multiply both limits by n/100 to obtain 995 <= m < 1005. The permitted integers are 995 through 1004: ten different counts. The lower endpoint is included because a half-up tie at 9.95% rounds to 10.0%; the upper endpoint is excluded because 10.05% rounds to 10.1%.

With n = 1,000, the interval becomes 99.5 <= m < 100.5. The only integer is 100. The same displayed label can therefore identify one count or several counts, depending on the declared total. Multiplying the displayed 10.0% by the total gives a candidate, not a general proof that it is the original count.

3. Clip the intervals to the permitted source range. With n = 10,000, a 0.0% display can come from counts 0 through 4. Their exact percentages range from 0.00% to 0.04%; count 5 reaches 0.05% and displays as 0.1%. Thus this display does not prove a count of zero.

A 100.0% display can come from counts 9995 through 10000. Their exact percentages range from 99.95% to 100.00%. Larger counts are not allowed by 0 <= m <= n. Thus this display does not prove that every item belongs to the category. These ranges describe display ambiguity, not measurement uncertainty or a confidence interval.

4. The proposed 33.4% violates the stated rule for this source. Spruce has count 1 out of 3, so its exact share is the same as the other two categories and independently rounds to 33.3%. Changing only its label to 33.4% would no longer follow the declared contract. A policy that allocates rounding differences across labels can be defined, but its labels must be identified as adjusted under that policy. This activity does not silently substitute such a policy for independent rounding.

5. The denominator is missing. A rounded percentage and a rounding rule do not supply the total. An assumed denominator creates a new assignment rather than recovering the source. Request the actual total and, if exact reconstruction is required, the original category count. Retain the displayed label separately from those source fields.

6. Use the total constraint as well as each label interval. Each 50.0% label separately permits counts 4995 through 5004. However, the two counts must sum to 10,000. The permitted first-category counts are therefore 4996 through 5004, with second count 10000 minus the first. There are nine ordered pairs, from (4996, 5004) to (5004, 4996), including (5000, 5000).

The choice (4995, 5005) fails because 5005 displays as 50.1%, and (4995, 4995) fails because its total is 9990. Independent label intervals are necessary but are not sufficient when the source supplies an additional total constraint.

Change just the tie rule

For an optional extension, keep n = 10,000 and the displayed label 10.0%, but replace half-up rounding with nearest-tenth, ties-to-even rounding. At a tie, choose the neighbouring tenth whose tenths integer is even. The value 10.0 has tenths integer 100.

Now both 9.95% and 10.05% round to 10.0%. The permitted counts are 995 through 1005: eleven counts. This changes one endpoint of the earlier result. Do not carry a half-up interval into a ties-to-even source merely because its displayed label looks identical. If the source does not declare its tie convention, retain that as unresolved information rather than inventing one.

A reconstruction log

For each proposed reconstruction, retain these fields:

Source card and category | Known total n | Original count, if supplied | Displayed percentage | Decimal places and tie rule | Permitted integer counts | Additional source constraints | What remains unknown.

Example: Separate card / unnamed category | 10,000 | Not supplied | 10.0% | One decimal, half-up | 995–1004 | 0 <= m <= n | Exact original count.

A range of compatible assignments is often a more faithful answer than a guessed single count. Keep it distinct from an observed count and from a claim about statistical uncertainty. Discuss any confusion as feedback on this proposed worksheet, not evidence that it has improved a class's performance.

Scope, credit, and reuse

This is an arithmetic and source-reading exercise, not an independent product review or guidance for validating research measurements. A polished label still needs a source, a denominator, and a declared display rule. Figran's website is figran.net; public contact is support@figran.net.

The original teaching text and invented cards are offered under Creative Commons Attribution 4.0 International. Credit Figran Team and this article title, link the licence, and indicate changes. This licence does not cover Figran software, brand assets, Mataroa software, or the platform interface.