Describe criteria before seeing the answer
Start with the comparison question. Are you valuing a mature business with stable earnings or a company still expanding production? Select features such as revenue source, geography, scale, cyclicality, cost structure and capital intensity. If an important feature differs, record it rather than hiding it to enlarge the group.
Write inclusion rules in advance. Otherwise it is tempting to add an expensive peer to justify a higher valuation or remove an inconvenient cheap one. A small group of carefully explained companies can be more informative than a long list sharing only an industry label.
Collect data from official disclosures and checked market sources. An annual report describes operations and finances but does not remove the need to check period and group composition.[14](#source-valuation-annual-report) If one peer has only old reporting, mark it separately or exclude it from that calculation.
Bring metrics to one basis
Use one price date and the same financial period. The last completed year, trailing twelve months and next-year forecast cannot be silently mixed. For P/E, match price to the corresponding EPS, or market capitalization to corresponding earnings.[15](#source-valuation-pe)
Currency and units also matter. “Millions” for one company and “thousands” for another can create a huge error before valuation. Check share classes and the portion of earnings attributable to the selected equity.
If earnings are analytically adjusted, explain the adjustment for every participant. Do not remove a one-off loss from a favourite company while leaving a one-off gain for others. Keep original and adjusted values side by side.
Example where ranking changes
All figures are fictional and in millions of roubles. Company A has capitalization 100 and earnings 10, so P/E is 10. B has capitalization 150 and earnings 10, so P/E is 15. C has capitalization 200 and earnings 20, so P/E is again 10.
A and C initially look equal relative to earnings. Suppose 8 million of C’s earnings came from a one-off event excluded from the teaching estimate of repeatable results. Adjusted earnings become 12 and P/E about 16.67. The equality disappears because the denominator was examined, not because price moved.
If B’s capitalization came from another date and was 135 million, its ratio would be 13.5. Mixed dates can change the picture without any business change. Reconcile figures from different services before computing a median.
What a difference from the group means
A low multiple may reflect weaker expectations, unstable earnings, debt or limited information. A high multiple may reflect growth expectations, stability or simply an expensive valuation of the whole group. The difference does not identify its cause and is not an automatic trade signal.
A median reduces the influence of extremes but does not repair poor peer selection. Show the range and participant characteristics, not just one summary cell. For some industries, the familiar metric may be unsuitable; no common denominator can be a valid result.
The memo should state selection rules, data date, formula, adjustments and possible explanations. Compare the result with an independent cash-flow analysis. If they differ, investigate which assumptions created the difference rather than choosing the more pleasing answer.
Modified multiples: show both sides of the adjustment
A modification changes a metric to investigate a specific company difference. NYU’s lecture separates this from subjective storytelling and statistical adjustment.[2](#source-gv-relative-lecture) A name does not make a ratio standard. Preserve original values, apply one rule to the whole group and show a calculation that reconstructs both versions.
In a teaching construction, remove excess cash and related financial income. A has E=150 million roubles, excess cash C=30, owner earnings NI=15 and after-tax interest income I=1.5. Original P/E=150/15=10. Define modified operating residual equity as (E−C)/(NI−I): 120/13.5≈8.89. This is not EV/EBIT: debt is not added and operating result remains after debt service.
For B, E=100, C=0, NI=10 and I=0, so both ratios are 10. Subtracting cash only from A would give 120/15=8, an artificial improvement because income from the excluded asset remains in the denominator. Check that C is excess and belongs to the same shareholders; operating liquidity, withdrawal restrictions and minorities can invalidate simple subtraction. With NI−I=0 there is no final ratio; with negative operating earnings the ordinary cheapness scale does not apply.
Another modification is PEG=(P/E)/g, where this guide expresses g as percentage points of expected annual EPS growth, not a decimal. At P/E=20 and growth 10%, PEG=20/10=2; at P/E=24 and growth 20%, it is 1.2. Using 0.10 instead of 10 gives 200, so declare the unit. At zero growth PEG is undefined; negative growth does not create a meaningful ranking, and values near zero are unstable. State forecast horizon and source. PEG does not remove risk, payout differences or growth duration.
Ratio families answer different questions
Before ranking, record question, numerator, denominator, date and units. Read reports and notes for actual data: Investor.gov’s educational bulletin explains why balance sheet, results and cash flows should be read together rather than selecting a convenient line.[1](#source-rm-read-10k) Liquidity, debt coverage, activity, profitability, return on capital and market valuation answer different questions and have different limitations. For example, EBIT=30 and interest=10 gives coverage 3, but principal payment 40 and cash flow before service 35 leave a 15 shortfall against total service of 50. Coverage 3 does not certify payment capacity on the date.
A lender asks about timely cash service and claim seniority; an equity holder asks about residual flows after obligations. Unknown is not zero, and correlated ratios may count one risk twice. A critical constraint such as no confirmed funding for a required payment cannot be offset by an attractive margin through arbitrary weights. Show several transparent rankings, not a universal shopping list or professional credit rating.