How to Keep Calendar-Year and Fiscal-Year Figures Separate in AI Research
Summary
- Calendar-year (CY) and fiscal-year (FY) numbers are not interchangeable; mixing them can silently break trend lines and comparisons.
- In AI-assisted research, keep CY and FY separate by recording the period definition, exact dates, and the unit of measure alongside every figure.
- Use a four-layer evidence record: original text, paraphrase, interpretation, and verification status.
- Never “convert” FY to CY (or vice versa) unless you have a defined method and the underlying quarterly/monthly data to support it.
- A reusable record format plus a worked example reduces confusion when you revisit notes, share findings, or ask an AI to summarize.
When you use AI to help with research, it can summarize quickly but it will not automatically protect you from a common accounting trap: mixing calendar-year (Jan–Dec) figures with fiscal-year figures (a 12-month period that can start in any month). The result can be subtle errors like comparing “FY2024 revenue” to “2024 revenue” as if they cover the same months, or building a chart where one point is CY and the next is FY.
This guide shows a practical evidence-handling workflow to keep CY and FY figures separate, with a worked example and a reusable record format that clearly distinguishes: (1) original text, (2) paraphrase, (3) interpretation, and (4) verification status. No invented quotes, no made-up sources, and no “magic conversions.”
What “calendar year” and “fiscal year” mean (and why AI can mix them up)
Calendar year (CY) is the period from January 1 to December 31 of a given year.
Fiscal year (FY) is a 12-month accounting period used by an organization. It may align with the calendar year, but it can also run, for example, from April 1 to March 31, or July 1 to June 30. Companies and governments may label fiscal years differently (for example, “FY2024” might mean the year ending in 2024, or the year starting in 2024). The label alone is not enough; you need the start/end dates or “year ended” date.
AI tools can mix CY and FY because:
- They may treat “2024” as a universal time bucket even when the source uses FY labels.
- They may compress multiple statements into one sentence and drop the period definition.
- They may infer a missing date range instead of flagging it as unknown.
The core rule: every figure must carry its “time-period metadata”
To keep CY and FY separate, treat the time period as part of the figure itself. A number without a period is incomplete.
For each figure you capture, record:
- Period type: CY, FY, trailing twelve months (TTM), quarter (Q1/Q2/Q3/Q4), half-year (H1/H2), month, or “as of” date (balance sheet items).
- Exact date range: start date and end date, or “year ended” date.
- Label as written: “FY2024,” “2024,” “year ended March 31, 2024,” etc.
- Unit and basis: currency, thousands/millions, nominal vs constant currency (if stated), and whether it is consolidated, segment-only, continuing operations, etc. (only if the source says so).
- Source pointer: where you got it (URL, document name, page, table name). If you cannot store a pointer, note that it is missing.
A four-layer evidence record (original text, paraphrase, interpretation, verification)
To prevent AI summaries from blurring CY and FY, separate what the source literally says from what you think it means.
Layer 1: Original text (verbatim)
Copy the exact sentence or table row text that contains the figure and the period definition. Do not “clean it up.” Preserve the fiscal-year wording.
Layer 2: Paraphrase (plain-language restatement)
Restate the original text in your own words without adding new facts. The paraphrase should still preserve the period type and date range.
Layer 3: Interpretation (your analytical meaning)
Write what you believe the figure implies for your analysis (trend, comparison, ratio, etc.). This is where mistakes happen, so keep it explicit and conditional.
Layer 4: Verification status (what you have checked)
Mark whether you have verified the period definition and figure against the original source and whether any ambiguity remains. Use a small, consistent set of statuses so you can filter later.
Example verification statuses you can use:
- Verified: period dates and figure confirmed in the source.
- Partially verified: figure confirmed, but period labeling is ambiguous (or vice versa).
- Unverified: captured from a secondary mention; original source not checked.
- Ambiguous: source uses FY label without dates and you cannot confirm what months it covers.
Worked example: keeping CY and FY separate (without inventing a conversion)
Below is a fictional example to demonstrate the method. The point is the record structure and the discipline, not the numbers themselves.
Scenario
You are researching an organization’s performance and you find two statements in different places:
- A press release says: “Revenue for FY2024 was $120 million (year ended March 31, 2024).”
- A separate dataset you already have is organized by calendar year and includes “2024 revenue.”
The risk: you might compare FY2024 ($120m covering Apr 1, 2023–Mar 31, 2024) to CY2024 (Jan 1, 2024–Dec 31, 2024) as if they represent the same 12 months. They overlap but are not the same period.
Evidence record (reusable format)
| Field | Example entry |
|---|---|
| Record ID | REV-ORG-001 |
| Metric | Revenue |
| Figure | 120,000,000 |
| Unit | USD |
| Period type | Fiscal year (FY) |
| Period label (as written) | FY2024 |
| Period start | 2023-04-01 |
| Period end | 2024-03-31 |
| Original text (verbatim) | “Revenue for FY2024 was $120 million (year ended March 31, 2024).” |
| Paraphrase | The organization reported revenue of $120m for its FY2024, which ended on March 31, 2024. |
| Interpretation (your analysis) | This figure covers Apr 2023–Mar 2024. It should not be compared directly to calendar-year 2024 revenue without aligning periods (e.g., using quarterly data). |
| Verification status | Verified |
| Source pointer | Press release URL + section/table reference (fill in) |
| Notes / caveats | FY label includes explicit “year ended” date; safe to treat as Apr–Mar fiscal year. |
How you use this record in AI-assisted research
- When asking an AI to summarize: provide the record fields (especially period start/end) and instruct it to keep FY and CY separate in outputs.
- When building a chart: only plot FY figures on an FY timeline (or label them clearly as FY ending dates). Do not place FY2024 at “2024” on a CY axis without a clear convention.
- When comparing to CY data: either (a) find CY figures from the same source, or (b) obtain quarterly/monthly data to construct a comparable 12-month window. If you cannot, state that the comparison is not directly comparable.
Common failure modes (and how to prevent them)
1) Treating “FY2024” as “the year 2024”
Prevention: require a “period end date” field. If it is missing, mark the record as ambiguous and do not merge it with CY data.
2) Mixing “year ended” with “as of”
Income statement metrics (revenue, profit) are for a period. Balance sheet metrics (cash, debt) are at a point in time.
Prevention: include a “period type” field that distinguishes “for the year ended” vs “as of.” If it is “as of,” do not treat it as a 12-month total.
3) Silent unit drift (thousands vs millions)
Prevention: store the unit exactly as stated (including scale). If the source says “$ in millions,” record that and normalize only in a separate calculation step you label as your own.
4) AI paraphrase that drops the date range
Prevention: in your record format, make the date range mandatory. If the AI output omits it, treat the output as incomplete and regenerate with the missing fields required.
5) “Conversion” without underlying data
Converting FY to CY requires knowing how the metric is distributed across months/quarters. Without that, any conversion is an assumption.
Prevention: if you must align periods, document the method as an interpretation step and label it clearly as an estimate. If you cannot justify it, do not do it.
A reusable record template you can copy into your notes
Use this as a consistent structure for every captured figure. The goal is that you (or a teammate) can understand it months later without re-reading everything.
| Field | What to write |
|---|---|
| Record ID | Unique ID (e.g., METRIC-ORG-###) |
| Metric | Exact metric name (Revenue, Operating income, Headcount, etc.) |
| Figure | Number only (store commas/formatting consistently) |
| Unit | Currency + scale (USD, USD millions, EUR thousands) or non-currency unit |
| Period type | CY / FY / TTM / Quarter / Month / As-of date |
| Period label (as written) | Exactly what the source calls it (FY2024, 2024, etc.) |
| Period start | YYYY-MM-DD (or blank if truly unknown) |
| Period end | YYYY-MM-DD (or “as of” date) |
| Original text (verbatim) | Exact sentence/table row including period wording |
| Paraphrase | Plain-language restatement with period preserved |
| Interpretation | Your analysis; include comparability notes and assumptions |
| Verification status | Verified / Partially verified / Unverified / Ambiguous |
| Source pointer | URL + page/table/section reference (or document name + page) |
| Notes | Anything that could trip you up later (restatements, scope, exclusions) |
How to ask an AI for help without losing the CY/FY distinction
AI can still be useful if you constrain the task. Instead of “Summarize these results,” ask for structured output that repeats the period metadata.
Example prompt pattern (structure-first)
- “Extract each figure into a table with columns: Metric, Figure, Unit, Period type (CY/FY), Period start, Period end, Period label as written, and a verbatim quote.”
- “When you paraphrase, keep the period end date in the sentence.”
- “If the period dates are missing, write ‘unknown’ and flag it as ambiguous rather than guessing.”
Example prompt pattern (comparison guardrails)
- “Compare only figures that share the same period type and the same date range. If they do not match, explain why they are not directly comparable.”
- “Do not convert FY to CY unless quarterly data is provided; if not provided, state that conversion is not supported.”
Decision table: what to do when you encounter FY vs CY in the wild
| Situation | What you should record | What you should avoid |
|---|---|---|
| Source says “FY2024” and provides “year ended” date | FY + explicit end date (and start date if inferable from a 12-month year) | Relabeling it as “2024” without the FY context |
| Source says “FY2024” but gives no dates | FY label + mark dates unknown + verification status “Ambiguous” | Guessing the fiscal year months |
| Dataset is CY but your source is FY | Keep two separate series; document non-comparability | Merging into one timeline without a stated convention |
| You have quarterly data for the FY period | Method note for constructing a comparable CY window (if you do it) | Hiding the method inside a “cleaned” number |
| Balance sheet “as of” numbers | As-of date + label as point-in-time | Treating it as a yearly total |
Frequently Asked Questions
FAQ 1: Why is mixing calendar-year and fiscal-year figures such a big problem?
Answer: CY and FY can cover different months, so a “year-over-year” comparison can become a comparison of partially overlapping periods. That can distort trends, seasonality, and any conclusions you draw from changes between years.
Takeaway: Treat the time period as part of the number, not a label you can swap.
FAQ 2: If a source says “FY2024,” how do I know which months it covers?
Answer: Look for an explicit “year ended” date (or a stated fiscal year start/end). If the source does not provide dates, record the FY label as written and mark the period dates as unknown until you can confirm them from a primary document (for example, the organization’s annual report).
Takeaway: FY labels without dates are ambiguous; don’t guess the months.
FAQ 3: Can I convert FY figures to calendar-year figures for comparison?
Answer: Only if you have enough underlying detail (quarterly or monthly breakdowns) to construct a comparable calendar-year window using a stated method. Without that detail, a conversion is an assumption. If you do estimate, label it clearly as your calculation and keep the original FY figure unchanged in your records.
Takeaway: Conversions require data and a documented method; otherwise, keep series separate.
FAQ 4: What is the minimum set of fields I should store with every figure?
Answer: At minimum: metric name, figure, unit (including scale), period type (CY/FY/etc.), period start and end (or “unknown”), the original text verbatim, and a source pointer. Add verification status so you can separate confirmed items from ambiguous ones later.
Takeaway: If you can’t answer “what months does this cover?” from your notes, the record is incomplete.
FAQ 5: How do I label a chart when I have both FY and CY data?
Answer: Use separate series (and a legend) or separate charts. Label FY points by their period end date (for example, “FY ended 2024-03-31”) rather than placing them under a plain “2024” tick that implies a calendar year. If you must place them on one axis, state the convention in a caption.
Takeaway: Make the period definition visible in the chart labels, not hidden in footnotes.
FAQ 6: How should I handle a figure mentioned in a secondary source without the original table?
Answer: Record it as unverified (or partially verified) and capture the exact wording used by the secondary source. Do not “upgrade” it into a clean CY or FY record unless you can trace it back to the primary document that defines the period and unit.
Takeaway: Secondary mentions are useful leads, but they are not the same as verified primary figures.
FAQ 7: What is the difference between paraphrase and interpretation in my research notes?
Answer: A paraphrase restates what the source says in simpler words without adding new facts. An interpretation adds your meaning: what the figure implies, how it compares to another period, or what you think caused a change. Keeping them separate helps you spot where assumptions enter the workflow.
Takeaway: Paraphrase preserves the source; interpretation is where you must label assumptions.
FAQ 8: How can I use AI to summarize my notes without losing the CY/FY distinction?
Answer: Give the AI structured records that include period type and exact dates, and require it to repeat those fields in its output. Ask it to flag missing dates as unknown rather than inferring them, and instruct it not to compare or merge figures unless the periods match.
Takeaway: Constrain the output format so the AI cannot drop the time-period metadata.