TSM TW Equity☆

Every bar is a figure as filed. The blue bars are the reported subtotals (Revenue → Gross profit → Operating income → Pretax income → Net income), drawn from zero at exactly the number in the filing — never a sum computed here. The coloured steps between them are the filed component lines, applied as decreases (the taxonomy stores expenses as positive magnitudes) or, for “Other non-operating”, exactly as signed in the filing.

Residuals. Whatever part of a reported subtotal-to-subtotal gap the served lines do not name is drawn as its own hatched step and labelled a residual. Stretching a neighbouring step to make the arithmetic close — what a conventional waterfall does — would put a number on screen that nobody filed.

Nested lines left out: interest expense. This filer tags that line AND a broader line that already contains it — Apple, for instance, files a marketing figure that sits inside its own SG&A, and many filers report an interest-expense line that is also inside their net “other non-operating”. Applying both would double-count and break the reported subtotal, so the bridge keeps the decomposition that reconciles and names the dropped line here. Every figure is still shown, unchanged, in the statements grid below.

The ghost. Each dashed outline is the same step in the prior period (2023), anchored at this period’s running level so the reader sees whether the step grew or shrank. It contributes nothing to this period’s arithmetic.

Reconciliation.

Every figure is a filed cash-flow line, summed over the fiscal years listed below. The taxonomy stores payments as positive magnitudes, so each use is applied here as a decrease. Nothing is annualised, inflation-adjusted or imputed.

The window. Operating cash flow is the anchor, so the bridge covers exactly the 10 fiscal years that serve it (2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023, 2024). A year the store has no operating cash flow for is outside the window entirely — its capex and buybacks are not summed either.

Partial legs. Acquisitions (1/10 years), Stock issued (3/10 years) — the filer did not tag those lines in every window year, so the bar sums only the years that did and UNDERSTATES the full window. The unfiled years are absent, never estimated.

Debt is netted. Long-term debt issued less repaid is shown as one net step, because a refinancing that issues and repays the same amount is not two events of capital allocation. Short-term / commercial-paper movement is a different tag and is not included — it lands in the remainder.

The remainder is not free cash flow. It is what the named lines do not account for: net purchases and sales of investment securities (the largest piece for a cash-rich filer), short-term-debt movement, other financing and investing items, FX, and the change in the cash balance itself. Calling it “cash retained” would assert a reconciliation this bridge has not done.

The hatch inside operating cash flow. TWD 2.05B of the pool is the stock-based-compensation add-back (0% of it, filed in 6 of 10 years) — a real cost of employing people, settled in shares rather than cash, which is why it sits inside operating cash flow. The dilution it causes shows up in the share count, not here.

A blank cell means the filer did not tag that line for that year — never a zero.

No composite. The three scores are not averaged into one “quality” number. They point in different directions (F is a 0–9 count, higher-better; Z is a distress distance, higher-safer; M is a manipulation screen, lower-cleaner) and were fitted on different samples for different questions — a blended figure would be ours, not theirs, and would hide exactly the disagreement that is worth reading.

Altman Z′

the private-firm re-estimation: BOOK equity instead of market cap, with its own cut-offs (2.90 / 1.23). It sits lower than the market Z for a highly-valued company by construction — that is the model changing, not the company weakening.

Beneish M

earnings-manipulation SCREEN (Beneish 1999, eight variables): above −1.78 flags accounting behaviour that resembles historical manipulators. LOWER is cleaner. A flag is a question, never an accusation.

Sector context. The percentile columns on the trend rows are the ones this payload serves: “vs own” ranks the latest value inside this company’s own served history, “vs sector” inside its sector pool. A blank sector cell means the pool was too thin to rank honestly (or the metric is a currency level, where a cross-currency rank would be meaningless) — never a filled-in guess. The three SCORES carry no sector percentile: the payload’s percentile block covers the served ratios only, so “is this Z good for Manufacturing?” is a question we cannot answer from what is served, and we don’t pretend to.

No score history. A 12-year F-Score / Z-Score strip is not drawn. The point-in-time factor store carries valuation, returns, margin, momentum and growth factors — not these composites — and recomputing them here from the annual columns would produce a series that disagrees with the headline above (different vintage of facts, different restatement handling). A score that disagrees with itself is a wrong number, so the history is omitted rather than approximated.

The nine named F-Score checks, the Merton distance-to-default / default probability, ROIIC and the growth-durability CAGRs live on FS · Financial Strength; the full ratio grid — every served ratio, with its own trend and both percentiles — is the Ratios section below.