Computed
A fixed engine reads the filing’s XBRL. Deterministic, cached on the accession, version-pinned to the engine that produced it. The same 10-K gives the same gross margin to everyone, forever.
46.9% gross margin
Ask anything about any public company.
Every answer comes from its SEC filings, with the source attached. FilingEye reads annual and quarterly reports, 8-Ks, proxies and insider filings, computes the financials itself, and answers your questions with the numbers and the passage they came from.
- From the filings themselves — not a third-party database, and not model recall.
- Figures are computed, not generated. A fixed engine reads the XBRL. The model summarizes and judges tone; it never produces a number.
- Every claim carries its quote. If the passage cannot be located in the document, the claim is downgraded and says so.
Free covers the whole analysis on three filings a month. No card.
What happened to Apple’s gross margin?
Optik Computed from filing data
Gross margin rose to 46.9% in FY2025, its fourth consecutive annual increase.
Gross margin percentage rose on a more favorable mix and higher Services revenue, partially offset by foreign exchange.
Follow up
Figures above are live from Apple’s FY2025 10-K.
Two kinds of information, never mixed
Everything the product shows you is either arithmetic or judgment, and you can always tell which from the layout alone. This is the whole competitive position, so it is a rule about the interface rather than a line in the footer.
AI generated
Summaries, sentiment, findings and materiality ratings. Cached per filing too, so two readers see the same answer — but these are judgments, and every one carries the verbatim passage it rests on.
“…partially offset by foreign exchange.”
When the quote is missing
If a claim’s quote cannot be located in the document, it is downgraded from “stated in filing” and loses its jump link. The badge and the proof are one decision, never two.
See within filing
And where a filing is silent, the product says so rather than filling the gap. The chat declines off-filing questions outright.
Check our work
These are not illustrations. Each figure below was computed from the company’s own XBRL as filed with the SEC, by the same method the product uses — revenue and gross profit matched to the same filing and the same annual period. Open the filing and check any of them.
AAPL
Computed from filing data Apple Inc.
46.9% gross margin, FY2025
- Revenue
- $416.2B
- Prior gross margin
- 46.2%
- Period ended
- 2025-09-27
- Net margin
- 26.9%
- Current ratio
- 0.89
MSFT
Computed from filing data Microsoft Corp.
67.9% gross margin, FY2026
- Revenue
- $331.8B
- Prior gross margin
- 68.8%
- Period ended
- 2026-06-30
- Net margin
- 40.3%
- Current ratio
- 1.23
NVDA
Computed from filing data NVIDIA Corp.
71.1% gross margin, FY2026
- Revenue
- $215.9B
- Prior gross margin
- 75.0%
- Period ended
- 2026-01-25
- Net margin
- 55.6%
- Current ratio
- 3.91
META
Computed from filing data Meta Platforms, Inc.
82.0% gross margin, FY2025
- Revenue
- $201.0B
- Prior gross margin
- 81.7%
- Period ended
- 2025-12-31
- Net margin
- 30.1%
- Current ratio
- 2.6
Gross profit derived: revenue less cost of revenue (META does not tag GrossProfit).
Accession numbers and links resolve to the filing index on sec.gov. Meta does not tag gross profit in its XBRL, so its figure is derived from revenue less cost of revenue — the kind of per-company variation a fixed tag list gets wrong.
Filing against filing
What changed since the last one
A 10-K is mostly last year’s 10-K. The interesting part is the handful of sentences that moved — a revenue policy reworded, a segment redefined, a contingency that appeared. FilingEye compares the notes against the prior filing and rates each change for materiality, so you read the differences instead of the document.
Notes, against the prior filing
10-K FY2025 compared with 10-K FY2024
Computed from filing data AI generated
The comparison is computed. The materiality rating on each row is the judgment.
- HIGH Revenue recognitionWas Extended warranty revenue is recognized ratably over the contract term.Now Extended warranty revenue is recognized ratably over the contract term, except where the arrangement includes a bundled service component, which is recognized at a point in time.
Changes when revenue lands, not just how it is described.
- HIGH Commitments and contingenciesNew The Company identified an understatement of expenses of $151 million, which management does not consider material to the consolidated financial statements.
Flagged regardless of the company calling it immaterial.
- MEDIUM Segment reportingWas The Company operates in four reportable segments.Now The Company operates in three reportable segments.
Prior-year comparatives are restated on a different basis.
Passages illustrate the interface. In the product every row links to the note it came from in both filings.
Period over period
A score with nothing to compare it to is half a score
A sentiment of +34 means little on its own. Against the same company’s last filing it means the tone lifted. Every score FilingEye produces — sentiment, management confidence, Beneish and Sloan — is shown with the same score from the previous comparable filing directly beneath it.
Every score carries its predecessor
10-K FY2025, with 10-K FY2024 beneath each
Filing sentiment AI generated
+34
Prior filing +21 13
Tone across the narrative sections
−100 to +100 Management confidence AI generated
66
Prior filing 71 5
How firmly management commits to its own outlook
0 to 100 Beneish M-Score Computed from filing data
−2.14
Prior filing −1.86 0.28
Lower is better, so this fell away from the cut-off
Cut-off −1.78 Sloan ratio Computed from filing data
9.8%
Prior filing 4.2% 5.6%
Moved away from zero, so accruals carry more of earnings
Distance from zeroWhat it reads, and what it tells you
Everything here runs today. The badge on each one says whether it is arithmetic or judgment.
18 ratios, each against its prior period
Profitability, liquidity, leverage, efficiency and growth. Computed from the filing’s XBRL, cached on the accession, identical for every reader.
Forensic scoring that shows its working
Beneish M-Score with all eight factors attributed individually, each colored against its own threshold, plus the Sloan accrual ratio.
Footnote diffs, rated for materiality
What changed in the notes since the prior filing. A revision is flagged however the company characterizes it — including when it calls it immaterial.
Segments and geography
ASC 280 reportable segments, revenue split by segment and by region, read straight out of the filing’s tagged data.
Sentiment, with the previous filing beside it
A −100 to +100 score and a management-confidence reading, shown inline with the same scores from the last comparable filing.
Key findings, every one quoted
Positive developments and red flags, ranked high to low, each carrying the verbatim passage it came from and a link into the document.
A forensic read that shows why
Most tools give you a forensic score. FilingEye gives you the eight factors behind it, each measured against its own threshold, so you can see which one moved and by how much.
Alongside it, the Sloan accrual ratio flags earnings quality by its distance from zero. Both are computed, both are cached on the filing, and both deep-link to a guide explaining how to read them.
Forensic risk
Beneish M-Score, all eight factors
-1.64 Above threshold
Cut-off −1.78. Higher means more consistent with earnings manipulation.
- DSRI Days sales in receivables 1.420
Receivables growing faster than sales. Threshold 1.465.
- GMI Gross margin index 1.190
Margin deteriorating year on year. Threshold 1.193.
- AQI Asset quality index 1.610
Non-current assets other than PP&E rising. Threshold 1.254.
- SGI Sales growth index 1.310
Growth itself is not manipulation, but it raises pressure. Threshold 1.607.
- DEPI Depreciation index 0.880
Depreciation rate broadly stable. Threshold 1.077.
- SGAI SG&A index 1.040
Overheads tracking sales. Threshold 1.041.
- LVGI Leverage index 1.120
Leverage up slightly on the prior year. Threshold 1.111.
- TATA Total accruals to assets 0.034
Accruals above the benchmark. Threshold 0.018.
AI generated
The score is driven mainly by asset quality: non-current assets other than property and equipment rose faster than the balance sheet as a whole, while 3 of the eight factors sit above their own thresholds.
Figures shown illustrate the interface. In the product every factor is computed from the filing's XBRL and linked to the statement it came from.
Already reading a filing on EDGAR?
The extension puts the same analysis beside it, on sec.gov, without making you leave the document. It shares one account and one allowance with the site — a question asked in the panel and a question asked here draw on the same counter.
Coming to the Chrome Web Store See a filing analyzed
Every score has a guide behind it
A computed figure only helps if you know how to read it. One page per metric: what it measures, how to read its thresholds, and what it does not tell you.
Start with the whole analysis, free
The free tier is not a teaser. All 18 ratios with prior-period comparison, the executive summary, key findings, every forensic score with its eight factors, and every footnote diff — nothing blurred. What is limited is three filings a month, 10-K and 10-Q only, and no chat. Chat starts at Base.