Corporate Credit Rating Scorecard
Six criteria, 34 factors, three years of financials. Both the weights and the scoring bands move with the revenue bucket — because 3x debt to EBITDA is a comfortable number at $2bn of revenue and a refinancing problem at $5m. Rolls up to a letter grade, subject to override caps that no amount of good ratios can outvote.
Revenue bucket
Latest-year revenue picks the bucket. The bucket then decides every weight and every band on this page.
Financials
All figures in $ millions. The latest year drives the ratios; the two prior years exist so the trend criterion has something to measure.
Operational & structure
The part of the rating that survives a bad year. Weighted hardest in the smallest bucket, where the accounts tell you least.
Criteria roll-up
Each criterion scores 0–5 as the weighted mean of its factors. Open one to see the factors underneath.
Same company, four buckets
Identical financials, re-scored as if the company sat in each revenue bucket. The gap is the weights and the bands, nothing else.
| Bucket | Score | Grade | Biggest shift |
|---|
Why the bands move with the revenue bucket
Most scorecards vary the weights by segment and leave the bands alone. That is half a job. A weight decides how much a factor counts; a band decides what counts as good. Both are size-dependent, and the second one is where the errors hide.
1. The same ratio is a different fact at a different size
Take debt to EBITDA. S&P's indicative benchmarks put under 1.5x at minimal financial risk and over 5x at highly leveraged, with the intermediate band at 2–3x. Those bands were built around companies that can refinance. A $2bn issuer at 3x has a bond market, a revolver and a relationship bank; a $6m distributor at 3x has one lender and a personal guarantee. The ratio is identical and the risk is not, so this scorecard tightens the leverage bands at the small end and relaxes them at the large end.
Profitability runs the other way. A 2% net margin is a warning sign in a niche $8m business and unremarkable for a $3bn grocery chain, so the margin bands loosen as revenue rises. Liquidity behaves like leverage: small companies are scored against a higher current ratio requirement, not a lower one, because they have no committed facility to draw on when a large customer pays late.
2. What the weights do on top
Operational and structural factors carry 30% in the smallest bucket and 10% in the largest. That is not a judgement about small companies; it is a judgement about their accounts. Below $10m of revenue the statements are often unaudited, the owner's affairs and the company's are entangled, and a single contract moves every ratio. Age, filing discipline, litigation and audit quality are the more reliable signals at that size.
By the time revenue passes $1bn the position inverts. Leverage and coverage alone carry 32%, because the numbers are audited, the trend is real, and capital structure is the thing management actually chose.
3. Three years is the entry price
The trend criterion is the reason this scorecard asks for three years rather than one. A single year tells you the level. Three years tell you the direction and the volatility, and direction is what separates a company recovering from a bad year from one entering a bad decade. Revenue CAGR, margin drift in basis points, the change in gearing, net-worth growth and the standard deviation of the net margin are all read across the three-year window.
S&P makes the same point from the other side: its benchmarks may be relaxed where a company's financial measures show very little volatility. Stability is itself a credit factor, and you cannot observe it in one year of accounts.
4. Overrides, because some facts do not average out
A weighted scorecard has a structural weakness: strong factors dilute fatal ones. Negative net worth scored as a zero on one factor out of 34 gets buried. So certain conditions cap the grade outright, regardless of the weighted score:
| Condition | Cap | Why |
|---|
The audit-quality cap is the one people argue with. It says that management accounts, however good, cannot support an investment-grade rating. That is deliberate. The rating is a statement about information as much as about performance, and unverified numbers are a different kind of object from audited ones.
5. The cliff at the bucket edge, measured
Bucketing has one honest cost and it is worth putting a number on rather than hiding. Because every weight and every band switches at once, two companies with identical financials can be graded differently when one sits just below a bucket edge and the other just above it.
Stepping the same balance sheet across each boundary, the discontinuity is 0.7 points at $10m, 1.0 at $100m and 5.2 at $1bn. Sweeping across a range of leverage and margin profiles, the worst case found at the $1bn edge is 8.2 points — comfortably more than the width of a grade band. A company at $999m and the same company at $1,001m can differ by a full notch.
That is a property of the design the brief asked for, not a bug in the arithmetic, and the largest jump sits at the boundary where the weights move most (operational 15 → 10, leverage 27 → 32). The production fix is to interpolate: blend the weights and band cut-offs linearly across a zone either side of each boundary, so size shifts the rating continuously instead of in steps. Rating agencies do the equivalent by treating scale as one continuous input among many rather than a switch. If you use this framework on a real book, put the blend in before you use it near an edge.
6. What this deliberately does not do
No industry adjustment. A 60-day receivable cycle is normal in capital goods and alarming in food retail, and a serious deployment would carry sector-specific band overlays — CRISIL notes it uses different parameters entirely for traders, real estate and education. No peer comparison, no management-quality assessment, no parent or group support notching, which in practice is often worth several notches on its own. No projections: this reads filed history, where a rating committee would also model the next two years.
And no default calibration. The bands here are anchored on published rating-agency benchmarks and conventional lending floors, not fitted to an observed default outcome. That makes this a structured expert-judgement scorecard, not a statistical model — useful for consistency and for making a credit committee argue about the right things, but it does not produce a probability of default. Fitting it to outcomes on your own book is the next step, and it is the step that turns a framework into a model.