Editorially revised on 9 October 2026.
Separate the record from the explanation
Analytical work starts with a question that the available evidence can answer. Define what a record means, check which observations are present and distinguish a calculation from an explanation. A correct percentage can still support an unjustified conclusion when the groups, time windows or review conditions differ.
This guide uses an original fictional work log to make those distinctions concrete. It does not assess a person's intelligence, certify software proficiency or predict a hiring result. Statistics Canada's educational discussion explains the narrower distinction between association and causation: an observed relationship does not itself establish why it occurred. Its examples and teaching tables are not reproduced here. Statistics Canada: correlation and causality.
Define the unit and the question
Suppose a fictional volunteer, Anil, helps a coordinator review event information cards. Each card is one draft record. A card counts as checked only when its required source fields have been compared with the supplied coordinator record. Checking a card does not mean its information is approved for publication.
The coordinator's question is: What can the supplied log establish about the two review periods? It is not a request to prove that a new instruction improved productivity. No real employer, event, participant information or performance outcome appears in this exercise.
The log contains these explicitly supplied observations:
| Supplied period | What the fictional log records |
|---|---|
| Period A | Twenty draft cards were assigned; twelve were checked during a stated thirty-minute review slot. The previous instruction was used. |
| Period B | Thirty draft cards were assigned; eighteen were marked checked. A revised instruction was supplied. The duration field is blank. |
| Source condition | Period A included cards with missing source fields. The log does not state how many. Period B's source completeness is not recorded. |
| Publication condition | Neither period includes a release decision or a subsequent correction record. |
Before calculating anything, write down the limits. The unit is a draft card, the outcome recorded is checking, and the available periods are these two entries. The record does not describe an entire working week, all volunteers or a company's operations.
Calculate only what the denominators support
Period A records twelve checked cards out of twenty assigned: 12 divided by 20 equals 0.60, or 60%. Period B records eighteen out of thirty: 18 divided by 30 also equals 0.60, or 60%.
The checked count rose by six cards. Relative to the first checked count, six divided by twelve equals 50%. That arithmetic is correct for the change in the recorded count. It does not show a fifty-percent increase in a time-based checking rate. The second review duration is missing, and the assigned batch also increased by fifty percent, from twenty to thirty.
Keep the measures named separately. The fraction of assigned cards recorded as checked is sixty percent in both periods. The absolute checked count differs. A checking-per-minute comparison cannot be completed from the supplied log. The reader should be able to trace each statement back to its numerator, denominator and unit.
Do not fill the missing duration with thirty minutes because the earlier period used that slot. The coordinator may confirm it later, but a convenient assumption is not a supplied observation.
Sort facts, calculations and explanations
A useful analysis note separates several types of statement:
| Statement | How it should be treated |
|---|---|
| Period B records eighteen checked cards. | Directly supplied observation, subject to the meaning and reliability of the log. |
| Both recorded completion fractions are sixty percent. | Derived arithmetic using the supplied counts and assigned totals. |
| Period B also lasted thirty minutes. | Unsupported assumption; its duration field is blank. |
| The revised instruction caused the higher checked count. | Proposed explanation that the supplied comparison does not establish. |
| Period B's checked cards were accurate and ready for release. | Unsupported outcome; checking, accuracy review and release are different records. |
This classification avoids two common drafting errors: presenting an assumed value as a fact and presenting an explanation as the result of a calculation. Neither error is repaired by putting the statement into a chart or adding decimal places.
Identify competing explanations without selecting a winner
The revised instruction might matter. A longer second review period might also matter. Different source completeness or different card difficulty could affect what was checked. These are possible explanations to investigate, rather than findings about the fictional event.
The log does not let Anil decide which explanation is correct. It also does not establish whether the cards were comparable, whether another helper contributed or whether “checked” was applied consistently. Naming a possibility should lead to a bounded information request, rather than a confident story about another person's effort or motivation.
Keep an alternative in the note when it could change the interpretation. Avoid an unlimited list of speculative causes that no one can investigate. For this task, the missing duration and checking definition are immediately relevant to the coordinator's stated question.
Request the information you are authorised to use
Anil can ask the coordinator to confirm the recorded duration for Period B, the checking definition applied in each period and whether the assigned cards had comparable source requirements. He cannot obtain private participant data or silently change the original log.
An original request could read:
The log shows twelve of twenty cards checked in Period A and eighteen of thirty in Period B, so both recorded fractions are sixty percent. Period B's duration is blank. Please confirm that duration and whether the same checking definition applied in both periods. I have not recorded a time-based improvement or a cause for the difference in counts.
This request makes the arithmetic and the unresolved fields visible. It does not announce a conclusion first and ask someone to endorse it later. The actual coordinator's instructions and permission boundaries would govern a real task.
Update the note without erasing the earlier uncertainty
For a separate, explicitly supplied continuation of the fictional exercise, suppose the coordinator confirms that Period B lasted forty-five minutes. The log still lacks card-comparability and accuracy-review information.
Twelve cards in thirty minutes and eighteen in forty-five minutes each produce a recorded rate of 0.4 checked cards per minute. That result follows from the newly supplied durations. It does not establish that the revised instruction had no effect, that the cards were equally difficult or that either period's checked outputs were accurate.
Record the new duration, who supplied it and which question it answers. The original blank should remain traceable as an earlier missing field rather than being described as something Anil knew from the beginning. A revised conclusion can be narrower than the first proposed explanation and still be useful.
Write a complete bounded analysis note
Here is an original final note for that supplied continuation:
Scope: two fictional card-checking periods. Period A records twelve of twenty assigned cards checked in thirty minutes. Period B records eighteen of thirty; the coordinator later confirmed forty-five minutes. Both recorded completion fractions are sixty percent, and both recorded time-based rates are 0.4 checked cards per minute. These comparisons do not establish the revised instruction's effect because card comparability and consistent accuracy review remain unverified. No release decision is recorded. Preserve the original log and the duration confirmation with this note.
The note contains the question's scope, the relevant figures, the added information and the remaining limitations. It gives the next reader enough detail to check the calculation without pretending to settle a causal question.
Practise with a genuinely separate variation
For a separate author-created variation, imagine ten items assigned and seven checked in one period, followed by twenty assigned and twelve checked in another. The recorded fractions are seventy percent and sixty percent respectively. The checked count is higher in the second period while its completion fraction is lower.
That variation demonstrates why the measure must be named before comparing periods. It is not a later observation from Anil's event, a validated skills test or evidence about a learning programme. No duration is supplied, so a time-based comparison remains unavailable.
Keep your own practice record clear about whether figures are invented, classroom data or authorised work records. An independently created exercise can show the reasoning you practised; it does not become client analysis or verified business impact.
Explain your contribution accurately
If you describe this exercise in an application or interview, identify the setting and the work you actually performed: defining the unit, checking denominators, calculating the stated measures and recording missing information. Do not claim that you improved a real organisation's productivity or used software you did not use.
For broader option comparison, the separate decision-making guide addresses a different task. Use only a relevant, working resource and keep the evidence boundaries in your own explanation. Analytical work is useful when another reader can distinguish the observation, the calculation and the unresolved question.