When Valuation Models Disagree: Follow the Cash, the Units and the Assumptions

A valuation model should not turn into a recommendation until its methods, units and assumptions tell a coherent story. A workbook for the fictional Teaching Operating Company illustrates the problem. Its discounted-cash-flow output was ₹160 per share against a ₹400 reference price. A separate comparable-valuation table produced ₹1,200, ₹800 and ₹600 and labelled all three outputs “Undervalued.” A copied heading refers to Teaching Template D, a different fictional business. Those contradictions are central evidence, not cosmetic defects.

The workbook is a constructed teaching example, and every business identity, peer and numerical case input is synthetic. It preserves the conflicting methods, the unreconciled “Undervalued” labels and the model-control failures without carrying a valuation opinion on any real company. The source did not establish a publication date; none is inferred here, and the synthetic figures are not market observations.

Four valuations, two conflicting stories

The teaching workbook contains a cash-flow approach and three relative-valuation approaches. The first estimates value from projected cash generation and discounting. The others apply valuation multiples drawn from comparable businesses to selected operating or earnings measures.

Synthetic valuation outputs disagree with the same reference price

The percentage comparisons are arithmetic illustrations computed from the synthetic outputs. They are not price targets, expected returns or forecasts of convergence. A valuation is conditional on its assumptions; the market is under no obligation to move towards an unreconciled spreadsheet number.

The workbook’s “Undervalued” labels appear in all three rows of its comparable table. They do not reconcile the much lower discounted-cash-flow result. The labels remain an unsupported workbook conclusion: replacing the case inputs has not repaired the methodological contradiction or made that conclusion more defensible.

The first control: identify the object being valued

Enterprise value and equity value are different objects. Enterprise value is associated with the operating business before allocation among relevant capital claims. Equity value is the residual value attributable to shareholders after the appropriate adjustments. Mixing those objects can create a large error even when every spreadsheet formula is mechanically valid.

The synthetic comparable table shows the following bridge:

The printed synthetic bridge from initial value to equity value — Method column: Revenue multiple

The printed synthetic bridge from initial value to equity value — Method column: Operating-profit multiple

The printed synthetic bridge from initial value to equity value — Method column: Earnings multiple

The table then divides the equity values by 25 crore synthetic shares. That bridge exactly reproduces the synthetic per-share values. Reproducibility is useful, but it does not establish conceptual correctness.

Revenue and operating-profit multiples can be enterprise-value multiples if defined that way. A price-to-earnings multiple, by contrast, normally generates equity value directly when multiplied by earnings attributable to equity holders. Subtracting net debt again can then double-count the financing adjustment. The workbook’s third column therefore requires a definition check before its number is interpretable.

This is a good example of why a model review must examine labels and economic meanings, not just whether a cell equals its stated arithmetic expression.

A common unit discipline

The synthetic share count is in crores and the corresponding aggregate values are also in crores of rupees. Dividing one by the other produces rupees per share:

$$V_{\mathrm{share}}=\frac{V_{\text{equity, crore rupees}}}{N_{\text{shares, crore}}}.$$

In words: Divide the value belonging to shareholders, in crore rupees, by the share count in crores to obtain rupees per share.

Here $V_{\mathrm{share}}$ is value per share, $V_{\mathrm{equity}}$ total equity value and $N_{\mathrm{shares}}$ the relevant number of shares. The use of the same scale in numerator and denominator is essential. Substituting an unscaled share count into that calculation can create a difference of ten million times while leaving the formula superficially plausible.

The same discipline applies to percentages. A growth rate stored as 0.08 is 8%, not 0.08%. Discount rates, margins and working-capital ratios should have explicit units. Forecast dates stored as spreadsheet serial numbers also need deliberate conversion before they become chart labels or financial-year references.

A reliable workbook therefore includes a small assumptions dictionary: currency, scale, reporting period, share basis, nominal versus real treatment and percentage convention. This is not administrative decoration. It defines the meaning of the outputs.

Reconstructing the cash-flow method

A simplified enterprise discounted-cash-flow model can be written as

$$EV=\sum_{t=1}^{T}\frac{FCFF_t}{(1+r)^t}+\frac{TV_T}{(1+r)^T},$$

In words: Business value combines today’s values of projected cash flows and what remains after the forecast period.

where $FCFF_t$ is free cash flow to the firm in period $t$, $r$ the discount rate, $T$ the explicit forecast horizon and $TV_T$ terminal value at the end of that horizon. Equity value then requires a consistent bridge from enterprise value to shareholder claims.

In a constant-growth terminal model,

$$TV_T=\frac{FCFF_{T+1}}{r-g},$$

where $g$ is the perpetual growth assumption and $r>g$ is required for this expression to have its usual interpretation. A small change in the difference between discount and growth rates can produce a large value change. Terminal assumptions should therefore be visible rather than buried in a formula.

The synthetic ₹160 output cannot be accepted or rejected merely because it disagrees with a market price. The reviewer needs to inspect revenue growth, margins, taxes, capital expenditure, working capital, terminal value and the discount rate as a connected set. A projection that grows sales aggressively while requiring little reinvestment may overstate cash generation. A model that repeatedly subtracts the same capital need may understate it.

The output discrepancy is a reason to investigate these drivers, not a licence to choose whichever method gives the preferred conclusion.

Comparable selection is an economic judgment

Relative valuation depends on the businesses selected as peers and on the metric used to compare them. Companies can operate in adjacent sectors but have different margins, brand strength, working-capital requirements, capital intensity or earnings cyclicality. Applying their multiples without adjustment can import those differences into the target valuation.

A revenue multiple is particularly sensitive to margin comparability. Two businesses with the same sales can produce very different cash flows. An operating-profit multiple reduces some of that problem but still requires consistent definitions of depreciation, leases, exceptional items and financing structure. A price-to-earnings multiple is affected by taxes, leverage, non-operating income and the earnings period chosen.

The synthetic peers have deliberately different business models

These are invented businesses, not aliases for real issuers. Their differing business models deliberately retain the problem of establishing economic comparability before importing multiples.

The teaching workbook’s copied heading names Teaching Template D, which is neither the target business nor one of these peers. This raises an additional control issue. It does not prove that every formula is wrong, but it means the reviewer cannot assume that the peer set, metric mapping or narrative was fully updated. Template reuse should trigger targeted checks of named ranges, hidden sheets, cached formulas and hard-coded references.

The correct response is to establish why each peer is economically comparable, not merely to remove the inconsistent heading and leave the valuation untouched.

Why averaging does not solve the disagreement

Taking the average of ₹160 and the three higher outputs would create a new number without resolving the underlying differences. The methods may share inputs, contain different errors or measure different claims. Equal weighting would imply an evidential symmetry that has not been established.

A better reconciliation starts with a common basis. Use the same valuation date, share count, debt and cash definitions, reporting currency and earnings period. Then isolate the drivers that differ: growth, profitability, discounting, terminal assumptions and peer multiples.

An explicit bridge can show how much of the gap comes from each assumption. For example, the reviewer might hold forecast cash flows constant while changing only the discount rate, then hold the rate constant while changing terminal growth. These are sensitivity exercises, not repairs that should be hidden from readers.

The final narrative should explain which method is most appropriate for the business and why. If the evidence is insufficient to choose, the honest result is a range of unresolved conditional valuations rather than a confident investment conclusion.

A review sequence that catches the important failures

Start with the output bridge: total value, financing adjustments, share count and per-share value. This catches unit and claim-definition errors quickly. Then trace major assumptions back to the financial statements and operating forecast. Finally, inspect model behaviour under deliberately simple scenarios.

Useful checks include whether additional debt changes enterprise and equity values in the expected places, whether a larger share count reduces per-share value, and whether a higher discount rate reduces a conventional cash-flow valuation. These checks do not prove that the business forecast is correct, but they can expose mechanical inconsistencies before the model is interpreted.

The reviewer should also separate formula errors from judgment disagreements. A double subtraction of net debt is a mechanics issue. Whether a peer’s multiple is appropriate is a judgment issue. Both matter, but they require different evidence and should not be described with the same level of certainty.

The publication conclusion

The teaching workbook expresses undervaluation through its comparable table while producing a contradictory discounted-cash-flow result. The synthetic reconstruction retains that methodological failure and shows why model reconciliation must precede any investment call. It does not repair the erroneous labels by supplying a revised buy or sell recommendation.

The comprehensive output from a valuation process is not just a price. It is a traceable explanation of the object being valued, the assumptions that drive it, the uncertainty around them and the reasons competing methods differ. When that explanation is missing, a precise number can make weak analysis look stronger than it is.