Nine of Our Valuation Methods Now Refuse to Answer Without One Balance-Sheet Line

Nine of the valuation methods in our engine will not return a number unless one line from the balance sheet is present. If it is missing, they abstain. They do not estimate it, and they do not treat its absence as zero.

That behaviour replaced something worse. Until it changed, a missing figure was quietly read as 0.00 - and because of what that figure does in the arithmetic, the error only ever ran in one direction: every levered company came out looking cheaper than the filings supported.

Enterprise value is not equity value

Multiples split into two families, and the split is easy to lose track of because the ratios look alike on the page.

Price-based multiples - price-to-earnings, price-to-book - are already about the shares. Price is in the numerator, so whatever comes out the other end is a value per share.

Enterprise-value multiples are not. EV/EBITDA, EV/EBIT, EV/Sales, EV/Gross Profit, EV/FCF and earnings power value all measure the whole firm: the claim of the shareholders plus the claim of the lenders, taken together. That is a deliberate property. EBITDA sits above interest expense on the income statement, so it belongs to everyone who financed the business, and pairing it with a share price alone would compare a whole-firm cash flow against a part-firm claim.

To get from one to the other you subtract what the lenders are owed, net of the cash already on hand:

Equity value per share = enterprise value per share - net debt per share

Net debt is total debt minus cash and equivalents. When a company holds more cash than debt, the figure is negative and the subtraction adds to the equity value. We wrote up the underlying accounting separately in net debt explained and enterprise value explained.

That one subtraction is the bridge. Skip it and you have not made a small error - you have reported the wrong quantity.

What defaulting the bridge to zero does

Take an industrial company with EBITDA of $8.00 per share, trading at a 10x sector EV/EBITDA multiple, with net debt of $25.00 per share. These are the figures our shipped engine returns:

The firm is worth $80.00 per share. The shareholders' portion is $55.00. Reading a missing bridge as zero publishes the first number in the place of the second: a 45.5% overstatement of the equity estimate, on a company whose leverage is entirely ordinary.

Note the third row. A negative bridge is not an error case - net cash genuinely belongs to the shareholders and genuinely adds to the estimate. A method that cannot tell a missing number from a zero also cannot tell a debt-free balance sheet from a cash-rich one.

The direction is what makes this the kind of defect worth building a gate around. Missing debt data is not randomly distributed. It goes missing most often on the companies with the most complicated capital structures, which are disproportionately the levered ones - so the error concentrates exactly where it is largest, and it never once runs the other way. Errors that cancel are noise. Errors with a sign are bias.

Why we abstain rather than estimate

Our composite estimate blends up to nineteen methods, weighted by sector. Something has to happen when one of them cannot run, and there are only three real options: publish a wrong number, substitute a guess, or return nothing from that method and let the others speak.

Here is what each does to the blend, using the same company at a $100.00 share price with price-to-earnings and price-to-book contributing $120.00 and $110.00:

Look at the third row honestly. Abstaining did not land closer to the truth - it landed further from it than the contaminated number did, in this particular configuration. That is not an argument against abstention, and it is worth saying plainly rather than dressing up.

The case for abstaining is not that it is more accurate on any single name. It is that the alternative is a number with a known, unfixable, one-sided error, and a composite that reports three methods agreeing when only two of them had their inputs. The count is a claim in its own right. A method that contributed a fabricated figure has not confirmed anything, and letting it vote inflates the apparent agreement of the whole model at precisely the moment the underlying data was weakest.

So the abstaining method is dropped from the blend and the remaining weights are renormalised. It does not enter as a zero, which would drag the composite toward nothing, and it does not enter as a guess. It simply is not counted, and the methods count falls to two so that the thinner evidence is visible rather than hidden.

Zero is a real answer; missing is not

The distinction the engine now enforces is between an observed zero and an absent value.

An explicit 0.00 computes normally. A company with no debt and no meaningful cash has a bridge of zero, that is a fact about its balance sheet, and the enterprise value and the equity value coincide. Refusing to answer there would be its own failure.

None - the absence of an observation - abstains, with a typed reason code rather than a generic failure. On the stock page the method row reads "Net debt unavailable" instead of showing an estimate. You can see which methods ran, which did not, and why.

Reason codes also have a precedence order, which matters more than it sounds. If EBITDA itself is missing, the method reports the missing EBITDA, not the missing bridge. If a margin arrives at an ambiguous scale, that is what it reports. The first thing that went wrong is what gets named, because a reason code that always blames the last check in the chain would be useless for finding the actual gap.

The general rule

This is the same discipline we applied to reported earnings that are not earnings: the engine's job is to be right or to say nothing, and a default value is a way of saying something while pretending not to have decided anything.

Defaults are seductive in financial code because they keep the pipeline running. Nothing throws, no dashboard turns red, and every row comes back populated. The cost is that a populated row is indistinguishable from a correct one, and the failure surfaces only as a slightly optimistic estimate that nobody has reason to question.

If you are running these multiples yourself, the check is quick: confirm you subtracted net debt before comparing an EV-based estimate to a share price, and confirm the figure you subtracted came from the balance sheet rather than from a blank field. Our EV/EBITDA calculator asks for the bridge explicitly for that reason, and the full weighting scheme is documented on our methodology page.

An estimate that declines to answer is less satisfying than one that always produces a figure. It is also the only one of the two whose silence tells you something.