Methodology
Documentation
The math behind every score.
Every score in a Sentinellis report is derived from public financial data using established academic models — not opaque proprietary blends. Here are the formulas, inputs, and how to read each output.
Where “about two hours” comes from
The work one brief represents
A brief is roughly two hours of company research. That is a claim about the work, so here is the work — every step a person would take to end up holding what one brief holds, with an honest range on each. The estimates assume someone who knows what they are doing and is not learning the formulas as they go.
| Step | Low | High |
|---|---|---|
| Find and open the latest annual report | 3 | 8 |
| Transcribe three statements, three to four years | 15 | 40 |
| Margins, ROE, ROA | 5 | 12 |
| ROIC and WACC (beta, risk-free rate, ERP, cost of debt, tax, structure) | 15 | 35 |
| Altman Z-Score (five ratios) | 6 | 15 |
| Piotroski F-Score (nine tests, needs the prior year too) | 12 | 30 |
| CAGRs: revenue 3/5/10y, EPS 3/5y, EBITDA 3y | 8 | 25 |
| Cash conversion cycle (DSO, DIO, DPO) | 6 | 15 |
| Enterprise value and eight multiples | 8 | 20 |
| Liquidity and leverage ratios | 5 | 12 |
| Scan a week of news, deduplicate it, judge what is actually relevant | 15 | 45 |
| Executive team and reported pay | 10 | 25 |
| Six months of price against a benchmark, beta, unusual days | 10 | 30 |
| Write it up in plain English with the sources attached | 15 | 40 |
| Total, in minutes | 133 | 352 |
That is 2h13m to 5h52mworking from the companies’ own reports. If instead you lean on a free screener that already computes the statements, the multiples, the margins, the liquidity ratios and the price chart, the work that remains is still 99 minutes — a floor of about 1h39m even with good free tools.
We quote about two hours because it sits below the low end of the first estimate and just above the floor of the second. It is the conservative number, not the flattering one.
What this is not. It is not a claim that you personally spend two hours per company — most people, reasonably, spend far less, which is rather the point. It is not multiplied into a weekly or monthly figure. And we do not do every step above from primary documents: we read no filings ourselves, and executive pay is reported by our sources rather than derived from a proxy statement. Where a figure was computed by us rather than read from a source, the brief marks it as estimated.
Sentinellis Health Score (0–10)
Composite score across 6 categories
Thirteen checks across six fundamental categories, each scored against a stated threshold and normalized to 0–10. A check whose input is missing is excluded from both sides of the ratio rather than counted as a failure, so the weights below re-normalize across whatever data a given company has. Every report shows the six category scores with the measured value and the threshold behind each check, so the headline number can be audited rather than taken on trust.
Strong (≥7)
Multiple positive signals across categories
Mixed (4–7)
Some strengths, some concerns
Weak (<4)
Multiple red flags or fundamentals missing
Piotroski F-Score (0–9)
Joseph Piotroski, 2000 — fundamental strength check
A nine-point binary scorecard testing profitability, leverage, liquidity, and efficiency. Each test passed earns one point. Originally developed by Stanford accounting professor Joseph Piotroski to identify financially strong companies in the value segment of the market.
Reading the score
- 8–9: Strong fundamentals across categories
- 4–7: Mixed — investigate the categories failing
- 0–3: Weak fundamentals; deeper due diligence required
Sentinellis awards partial credit when a Piotroski input is missing rather than scoring it as failed. Stated explicitly in each report so you can audit the score.
Altman Z-Score
Edward Altman, 1968 — bankruptcy risk predictor
A weighted formula combining five financial ratios to estimate bankruptcy probability over the next two years. Developed by NYU Stern professor Edward Altman in 1968 and widely used in academic and lending contexts.
Safe (Z > 2.99)
Low bankruptcy risk
Grey (1.81–2.99)
Investigate further
Distress (Z < 1.81)
Elevated bankruptcy risk
ROIC vs WACC (value creation)
The single most important question in fundamentals
ROIC — Return on Invested Capital
How much profit the business earns per dollar of capital invested in operations.
Where NOPLAT = EBIT × (1 − tax rate); Invested Capital = total debt + equity − cash & equivalents
WACC — Weighted Average Cost of Capital
The blended cost of the capital the business raised — debt plus equity, weighted by the proportion of each.
Where E = equity, D = debt, V = E + D, Re = cost of equity (CAPM), Rd = cost of debt, t = tax rate. Sentinellis sets debt_weight = 1 − equity_weight to keep them consistent.
The spread (ROIC − WACC)
When ROIC > WACC, the business creates value with every dollar reinvested. When ROIC < WACC, growth destroys value — the firm would be better off returning cash to shareholders. This single ratio is one of the most predictive signals of long-term equity returns.
Confidence Score (1–10)
How much to trust this specific report
Every report carries a confidence score reflecting how much source data was available. The score is set during AI synthesis based on:
- ·35% financial data completeness — were all key statements available?
- ·30% legal & jurisdictional clarity — is the entity well-defined and tracked?
- ·35% market sentiment / news coverage quantity & quality
A confidence of 8–10 means we had rich data and high-tier sources. A score of 5–7 means a known gap (e.g., missing financials). Below 5 means treat the report as exploratory only.
Where each bar comes from
And the one rule that decides which source is allowed
A check is only as good as the bar it is measured against, so every bar on a Sentinellis report says where it came from. Two kinds are allowed, and which one applies is not a matter of convenience.
- ·Leverage and coverage come from published credit criteria, never from our own library. Net debt/EBITDA is scored against S&P’s corporate bands, and which band table applies depends on the kind of business: a regulated utility, a midstream operator, an equity REIT and a generic industrial tolerate very different debt, and one bar for all four was wrong by two to three turns. A bar derived from our own sample would certify whatever the sample happens to do — and leverage is partly a policy choice, so the circularity is not harmless.
- ·Margins and growth come from the company’s own peers, because they are structural. A fuel-and-grocery retailer does not have a software company’s gross margin and never will. The bar is the lower third of the company’s own industry in our library, or of a broader peer group when the industry is too thin to carry one, or of its sector. The report says which, in the row itself.
- ·The discount rate follows the currency of the cash flows, not the domicile. The risk-free rate is the one for the currency a company actually earns in, and the equity risk premium is a total by country — not a mature premium with a country spread stacked on top, which double-counts the same risk. Each report carries the table’s source and as-of date, and names the currency the rate came from. Where we do not hold a sovereign yield for that currency the report says so, and the calculation falls back to a default rate rather than inventing one — today that is true of the euro, sterling and a handful of smaller currencies, about an eighth of the companies covered.
- ·Scoring is graded, not pass/fail at a hair. On a single hard bar, 2.49x passes and 2.51x fails at a precision the threshold does not have. A check can score full, half or none.
Everything with no published bar and no meaningful peer set keeps an absolute one, stated on the row.
Versioned, and every report says which version scored it
Why a score can move with no new filing
Two of the three tables behind the score are not constants. The peer bars are rebuilt from a library that grows every week, so a company can cross a bar that moved underneath it without reporting anything new. That is a real problem for a tool whose whole claim is that its numbers can be checked, and the answer is to name the table rather than to pretend it never changes.
Every report produced from 18 September 2026 carries the version of each table that scored it — the peer margins, the credit bands and the cost-of-capital table — so two reports weeks apart can be compared, and a change in either direction can be attributed to the company or to us. We do not print those version numbers on this page on purpose: the report is the place that knows, and a page repeating them would eventually be wrong.
When we publish no score at all
And what goes in its place
A bank loses six of the thirteen checks outright — gross margin, free cash flow and the current ratio mean nothing for a lender — and none of the seven that survive is capital adequacy, asset quality or funding, which for a bank is the entire company. A composite built from what is left would be a profitability score wearing the word “health”, so we publish no headline for lenders and say which metrics are missing instead of leaving the gap silent.
For banks incorporated in the European Economic Area the report now carries the real figures in that gap — the CET1 ratio, the liquidity coverage ratio and the net stable funding ratio, read out of the institution’s own Pillar 3 disclosure, each shown beside the bar it is actually measured against. The liquidity minimums are set in law; the capital bar is the bank’s own overall requirement, carried two rows away in the same filing. Neither is ours, and there is no verdict on top of them.
Analyst consensus
Recommendations we reproduce, and do not make
Some reports carry a consensus rating and 12-month price targets. Those are not ours. They are produced by the analysts who cover the company, and we reproduce the aggregate of what they have published. Sentinellis issues no rating and no price target of its own, on any company, ever.
- ·What we hold. The number of analysts contributing, the averaged rating, and the mean, median, highest and lowest target. We do not hold their individual notes, the dates they were written, the identities of their authors, or the assumptions behind them.
- ·Why there is no “upside” figure. We publish the targets and the market price side by side and stop there. Subtracting one from the other would turn somebody else’s estimate into a directional claim of ours, which is exactly what we do not do.
- ·Why the numbers carry no colour. A green band on a price target would be our verdict on someone else’s number. The same rule removed the colour bands from every valuation multiple in the report.
- ·Staleness. An aggregate updates as analysts revise, and a target set months ago still counts toward it. Each report states the date and time it was produced; the consensus is as of that moment and is not refreshed afterwards.
- ·Coverage. Most Bucharest-listed companies have few analysts or none, and the section is simply absent when nobody covers the ticker. A rating averaged across six analysts is a thinner thing than one averaged across forty.
Sentinellis holds no positions in the companies it covers, takes no payment from issuers, and has no broker referral arrangements. Revenue comes from subscriptions only. Nothing in a report is investment advice.