Did you know that a research report can have a perfectly accurate quotation and still become misleading after the source table is revised? Websites sometimes correct classifications, add late records or change their reporting rules. A student who sees only the latest table may no longer be looking at the numbers used in their first draft.
A source revision log is a dated record of which version of a source supplied each claim, what changed, how it affects the calculation and which wording must be reviewed. The immediate answer is simple: keep the old figure as evidence of what you originally consulted, verify the newer figure, recalculate any affected result, and make the new status explicit. Do not silently rewrite your research history or claim that the earlier value never existed.
The quick method: five checks before changing a sentence
- Record the original source title, full address, author or publisher, access date and the precise figure used.
- Open the latest original page, not a search snippet. Check the table heading, units, denominator, date range and notes.
- Find whether the change is a correction, additional data, a redefinition, or only a rearrangement of presentation.
- Recalculate the sentence using the same specified method; check related graphs, captions and conclusions.
- Write a short revision note saying what changed, when you checked it and whether the claim still holds.
Why a moving source is a different problem from a wrong calculation
A calculation answers a mathematical question about stated inputs. Source revision asks an earlier question: were those inputs still the ones the author meant to publish? Even flawless division cannot repair a mismatch between a new numerator and an old denominator. Nor does the date at the top of a website necessarily tell you whether the underlying data has been refreshed. A date label can describe the article, the table, the reporting period or the last visible edit. Read it in context.
Some changes are factual corrections. Others add observations that arrived after a first release. Yet others change which records qualify. Those reasons matter because a correction of one erroneous entry has a different interpretation from legitimate growth over time. A student should not write ‘the class improved’ simply because a later publication includes one more qualifying record.
Worked example: a revised class-submission table
Entirely invented illustration, not a report of actual pupils: A class has 24 learners. The first version of an online class-project summary shows 18 submitted drafts by a stated cutoff. A student writes: ‘On the published cutoff definition, 18 of 24 learners, or 75%, had submitted.’ Two days later, the publisher corrects a data-entry classification: one on-time draft had been recorded in the wrong category. The revised count is 19 of the same 24 learners.
Step 1: keep the original calculation and its label
Version A: 18 ÷ 24 × 100 = 75%. That number was the result of the earlier table, not an invented error by the student. Keep its date and original screenshot or downloaded copy where permission permits. Do not use that 75% as a description of the revised figure.
Step 2: compute the corrected figure without changing the denominator
Version B: 19 ÷ 24 × 100 = 79.1666…%, approximately 79.2% to one decimal place. The absolute change in reported share is 79.1666… − 75 = 4.1666… percentage points, or approximately 4.2 percentage points. Do not call that a ‘4.2% improvement’: the data were reclassified, not necessarily the learners’ behaviour.
Step 3: revise the actual sentence
A safe updated sentence is: ‘The corrected class summary lists 19 of 24 drafts as submitted by the cutoff (about 79.2%); the earlier version listed 18 of 24 because one on-time draft had been misclassified.’ This distinguishes the current figure, historical version and publisher’s stated reason. If the reason for revision was not supplied, say that the reason is unknown instead of inventing one.
A small revision ledger that protects research integrity
- Source and claim ID — for example, ‘Class-summary table, claim C3’.
- Version checked — publication label or snapshot date, plus time zone if relevant.
- Original value and meaning — 18 of 24 on time; 75%.
- Current value and meaning — 19 of 24 on time; about 79.2%.
- Reason recorded by source — one corrected classification, if explicitly confirmed.
- Downstream changes — graph B, opening paragraph and conclusion 2 need checking.
- Disposition — revised and rechecked; retain original record for audit.
This can be done on paper or in a small spreadsheet. The ledger is not a licence to keep private pupil information unnecessarily. Use aggregated numbers and follow school privacy rules. A screenshot records what was visible, but it may omit important definitions below the table; preserve context when possible.
Learner diagnostic: what is actually wrong here?
Give a student two printed versions of the same invented table. Ask: ‘Which one is newer? Did the reporting period change? Did the class size change? Does the new number show growth or a corrected record?’ A learner who points only to the larger number has not yet identified the source change. A stronger answer locates the revised value, reads the note explaining it, verifies the unchanged denominator, and adjusts the conclusion.
Another useful question is: ‘Which sentence in your report would still be true if you removed the numbers?’ The student may discover that a causal claim was unsupported in both versions. A source correction is an opportunity to inspect the whole argument, not just replace a percentage.
Common mistakes, with precise repairs
Mistake: using two versions as though they measured two different dates
A corrected April release and an original April release can describe the same period. Treating them as April-versus-May observations manufactures a trend. Repair it by writing ‘original release’ and ‘revised release’, then checking the underlying observation date.
Mistake: relying on the search-engine snippet
A snippet can lag behind the page or truncate a qualifying sentence. Follow the original source, identify the exact table and the surrounding notes. If an old page disappears, cite the accessible archived version accurately rather than pretending it is the current page.
Mistake: updating only the visible number
A graph, headline, comparison, threshold decision or recommendation may still depend on the earlier figure. Search the report for the original number and the claim it supported. Not every occurrence requires replacement, but every affected use requires review.
Mistake: claiming the new version is automatically perfect
A later version deserves checking too. It may be provisional or reflect another definition. ‘Updated’ means newer, not infallible. Retain the possibility that a later correction will arrive.
Progressive practice: from guided inspection to independent judgement
Guided practice: Compare a fictional school-library table listing 40 books returned on time out of 50, then a correction to 42 out of 50. Work out 80% and 84%, and state the change as 4 percentage points in the reported proportion. Explain why the reason for the change must be checked before saying anything about behaviour.
Shared practice: Two charts have the same title but one now excludes cancelled bookings. Mark the changed denominator and write a one-sentence note on why the values cannot be compared directly without harmonising definitions.
Independent practice: Present an earlier report of 30 of 60 students participating and a later corrected table of 33 of 60. The student should calculate 50% and 55%, identify a five-percentage-point difference and create a brief audit note. Then ask what happens if the later denominator is 66 instead. The answer becomes 33/66 = 50%; an apparently larger count no longer indicates a larger share.
What parents and teachers can observe
Ask the learner to locate the precise source line behind a claim, name the version and re-create the calculation without copying a previous answer. After a week, repeat with a different topic. Progress is visible when the learner identifies what changed and what remained constant, updates all affected claims, and labels uncertainty without prompting. Finishing the report is not by itself evidence of source literacy.
Teachers can model a version note in an assignment, especially where public data are updated during a project. Parents can help by asking whether an apparently different statistic uses the same definition. Neither should demand a confident numerical conclusion when the underlying source is inaccessible or contradictory.
Boundaries: do not exaggerate what a revision proves
A revision log does not determine whether the original source’s sampling method was good. It does not verify another institution’s collection process, and it cannot recover a lost source with certainty. If a key version is no longer available, document the gap and qualify the claim or choose a verifiable source. For public reports, data licensing and privacy restrictions may also limit which snapshots can be reproduced.
Frequently asked questions
Should I delete the first number after correcting a school report?
Correct the public claim where needed, but keep an appropriately labelled work history. A reader needs the current defensible statement; the researcher needs enough record to explain how it was reached.
What if the website changes a number but gives no explanation?
Do not assume why. Record the values and dates you observed, look for release notes, and write ‘reason for change not stated’ if that remains true. Avoid a causal story you cannot support.
Does a newer page always replace an earlier source?
No. A historical question may specifically ask what an institution reported at a past date. In that case the earlier version is the object of study. Identify that purpose clearly rather than representing the old figure as current.
Is a screen capture sufficient for a citation?
It can support a record of what was displayed, but a complete citation should still identify the creator, title, location, relevant version or date, and any important explanatory notes. Follow the citation rules for the assignment.
Should every tiny wording correction require a new calculation?
Only changes that affect the values, definitions, selection, timeframe or interpretation require recalculation. Other edits may need an editorial note but not a numerical one.
Related learning and practical next steps
For claim-level checking, see How to Check Whether a Citation Supports the Exact Sentence. The author, publisher and host guide helps identify whose update you are reading. Explore the Core Aim learning hub, Education Hub and Clementi teaching reference for approaches based on diagnosis, explanation and independent checking.
The best conclusion is not ‘the latest figure wins’. It is: ‘I can show which version supports each sentence, I can explain what changed, and I will revise my conclusion only as far as the evidence permits.’ That habit is useful whenever students encounter a moving world of online information.