When Valuation Metrics Disagree:
Lessons From a Ten-Stock Sample

Investors often want valuation to produce one clean answer. A stock trades at a certain P/E, the number looks high or low, and the conclusion seems obvious.

The problem is that no single multiple captures the whole economic picture. P/E depends on net income. EV/EBIT looks at operating profit while also incorporating debt and cash through enterprise value. EV/FCF adds cash conversion and working-capital effects. All three can be valid, yet they can tell different stories about the same company.

That matters most when the business itself is cyclical. Temporarily high earnings can make a company look unusually cheap just before those earnings normalize. An experimental Qualtix research framework looked at ten companies using point-in-time historical observations and three valuation measures: P/E, EV/EBIT and EV/FCF. The purpose was not to calculate price targets. It was to ask a narrower question: where does each company's current multiple sit relative to its own history, and what should happen when the measures disagree?

Nucor shows why a low P/E can mislead

Nucor is the clearest cyclical example in the sample.

At the September 2022 observation, NUE traded at a P/E of 3.33x. On its own, that looks extremely low. But the denominator matters. Nucor was coming through an unusually strong part of the steel cycle, during an unusually strong earnings period.

By the September 2025 observation, the picture had changed. Trailing net income had fallen sharply and the P/E had risen to 23.74x. EV/EBIT was 16.41x, while EV/FCF was excluded because trailing free cash flow was nonpositive.

The point is not that NUE was definitely cheap in 2022 or expensive later. This framework cannot establish either conclusion. The point is that a very low P/E can be caused by an unusually high earnings denominator. For a cyclical business, that distinction matters.

The current September 2026 observation reinforces why several measures are useful. NUE's P/E is 19.68x, EV/EBIT is 14.77x and EV/FCF is 38.64x. All three are elevated relative to the company's eligible 2019-2025 history. Even so, according to the experimental Qualtix research framework, the independent metric estimates differ enough that the framework withholds a combined conclusion.

That is a useful result in itself. A valuation framework should not always be forced to produce one neat answer.

What "stability-adjusted historical valuation" means

The framework uses a stability-adjusted historical valuation approach. The current observation is September 20, 2026. The historical lookback uses one observation on or around September 20 in each year from 2019 through 2025.

Only financial statements filed on or before each observation date are eligible. Trailing net income, operating income and free cash flow are built from the latest four consecutive eligible quarters. Debt, cash and short-term investments come from the latest eligible balance sheet.

The three measures capture different parts of the company. P/E compares equity value with trailing net income. EV/EBIT compares enterprise value with trailing operating income. EV/FCF compares enterprise value with trailing free cash flow.

A method is excluded when its denominator is nonpositive, missing, nonfinite, or produces an extreme multiple of at least 500x. Historical quartiles and percentiles are calculated from the surviving observations using linear interpolation.

The framework then treats disagreement as information. According to the experimental Qualtix research framework, a combined conclusion is withheld when fewer than two methods survive or when the independent metric estimates differ by more than 40%. The purpose is to avoid forcing unlike signals into a single answer.

Three cases where the framework gives no combined conclusion

The clearest disagreement cases in the sample are Delta Air Lines, Nucor and Deere.

DAL is the most visible disagreement. Its current P/E of 13.24x sits at the 79.3rd percentile of its eligible history, and EV/EBIT at 12.26x sits at the 92.2nd percentile. EV/FCF tells a very different historical story: 11.15x is below the historical minimum and therefore sits at the 0th percentile.

That does not tell us automatically which measure is "right." It shows that the answer changes depending on whether the focus is net income, operating earnings or free cash flow. The approved dataset does not establish a single economic cause for the divergence, so the framework leaves the disagreement visible.

NUE is different. Its current P/E, EV/EBIT and EV/FCF are all elevated within their own histories, at the 78.3rd, 93.3rd and 100th percentiles respectively. The no-conclusion result therefore does not come from one visible multiple looking low and another looking high. It comes from the framework's independent internal metric estimates differing by more than 40%.

DE also has high current positions. P/E is 38.00x at the 100th percentile, EV/EBIT is 26.42x at the 99th percentile, and EV/FCF is 61.82x at the 80.1st percentile. Again, the framework withholds a combined conclusion rather than compressing materially different estimates into a single number.

Deere also illustrates an accounting-scope issue. Its consolidated debt includes John Deere Financial. Removing that debt while leaving consolidated finance earnings, cash flow and assets in the calculation would mix scopes, so the framework retains the consolidated treatment and flags the limitation.

When the metrics tell a more consistent story

Disagreement is not inevitable. Home Depot, Target and Union Pacific provide useful contrasts.

For HD, P/E, EV/EBIT and EV/FCF all sit below their historical medians, with current percentiles of 27.2, 28.5 and 26.0. The readings are not identical, but they are tightly grouped.

TGT shows a similar pattern. Its current P/E is at the 31.3rd percentile, EV/EBIT at the 16.5th percentile and EV/FCF at the 32.2nd percentile. All three sit in low-to-moderate parts of the company's eligible historical range.

UNP points the other way. Its P/E, EV/EBIT and EV/FCF sit at the 75.1st, 77.4th and 83.4th percentiles. All three are near or above the historical 75th percentile.

These observations do not establish what the shares should be worth. They simply show that, in these cases, the three measures tell a broadly consistent historical story.

Accounting scope matters

Historical comparisons are useful only if numerator and denominator remain economically compatible.

Baker Hughes provides a good example. For 2019 through 2021, the enterprise-value share basis includes Class A shares plus GE's paired Class B shares and BHH LLC units because those interests were economically exchangeable one-for-one. P/E, however, uses the diluted Class A share count because the relevant net income is attributable to Baker Hughes Company and Class A holders.

CAT and DE create a different issue through their captive-finance operations. Their consolidated debt includes financing debt. Simply removing that debt while retaining consolidated operating income, cash flow and assets would create an inconsistent comparison.

These details can look technical, but they matter when enterprise-value multiples are compared across time.

The important result may sometimes be no result

The main lesson from this ten-stock sample is not that one valuation multiple is superior to another. It is that valuation measures fail in different ways.

P/E can look unusually low when cyclical earnings are unusually high. EV/FCF can move sharply because working capital or cash conversion changes. Enterprise-value measures can become harder to interpret when financing operations are embedded in a consolidated company.

A historical range also does not establish intrinsic value. The lookback contains at most seven annual historical observations for each metric, and exclusions can reduce that further. Annual observations are not continuous history. Structural changes can make older periods less comparable, and a current reading can be unusual for valid company-specific reasons that this dataset does not establish.

The framework is therefore descriptive rather than predictive. It does not forecast future returns or issue an investment recommendation. It is also experimental editorial research and is not the valuation methodology deployed in the live Qualtix product.

Its purpose is narrower: show where current valuation measures sit relative to a company's own point-in-time history and whether several measures tell a coherent story.

Sometimes they do. Sometimes they do not.

When they do not, refusing to manufacture a precise answer can be more informative than pretending the disagreement is not there.

Source note: Experimental Qualtix research framework using a frozen dataset as of September 20, 2026, structured financial data from Financial Modeling Prep, and issuer filings available through SEC EDGAR. Historical observations use only filings available on or before each observation date. The framework is not deployed in the live Qualtix product. Historical multiple positions are descriptive, not price targets or investment advice.