Skip to content

alirezarezvani/claude-skills / finance/skills/stock-analysis

stock-analysis

Unverified

Analyze, research or value a listed stock or company (US, global, NSE or BSE) with verified data, ratios, scores and valuation. Use whenever asked if a stock is a buy, what it is worth, or how a company looks.

FileKindSize
SKILL.mdskill md32.7 KB
evals/evals.jsonconfig13.2 KB
evals/fixtures/synthetic-ar-excerpt.mdmarkdown5.7 KB
evals/fixtures/synthetic-drhp-excerpt.mdmarkdown3.4 KB
examples/README.mdmarkdown1.9 KB
examples/forensic-analysis-example.mdmarkdown8.8 KB
examples/standard-analysis-example.mdmarkdown15.6 KB
references/01-data-sourcing.mdmarkdown38.5 KB
references/02-core-factors.mdmarkdown49.3 KB
references/03-earnings-quality.mdmarkdown40.1 KB
references/04-balance-sheet-and-cashflow.mdmarkdown56.6 KB
references/05-returns-and-dupont.mdmarkdown49.8 KB
references/06-valuation.mdmarkdown53.7 KB
references/07-forensic-red-flags.mdmarkdown57.5 KB
references/08-governance.mdmarkdown67.0 KB
references/09-risk-and-macro.mdmarkdown52.7 KB
references/10-peer-set.mdmarkdown45.5 KB
references/11-scoring-rubric.mdmarkdown36.8 KB
references/12-report-template.mdmarkdown26.6 KB
references/13-situations.mdmarkdown60.6 KB
references/14-accounting-comparability.mdmarkdown48.6 KB
references/15-document-diligence.mdmarkdown76.1 KB
references/16-market-mechanics-and-tax.mdmarkdown48.6 KB
references/17-process-and-epistemics.mdmarkdown59.6 KB
references/18-forensic-mode.mdmarkdown16.5 KB
references/19-ipo-mode.mdmarkdown15.6 KB
references/20-challenge-pass.mdmarkdown14.1 KB
references/21-data-integrity-tools.mdmarkdown7.2 KB
references/sectors/_index.mdmarkdown67.8 KB
references/sectors/auto.mdmarkdown47.5 KB
references/sectors/aviation-hotels.mdmarkdown49.3 KB
references/sectors/banks.mdmarkdown38.6 KB
references/sectors/biotech-clinical.mdmarkdown47.8 KB
references/sectors/chemicals-cement.mdmarkdown46.2 KB
references/sectors/exchanges-payments.mdmarkdown41.2 KB
references/sectors/fmcg-consumer.mdmarkdown35.7 KB
references/sectors/holdco-assetmgr.mdmarkdown41.4 KB
references/sectors/infra-capitalgoods.mdmarkdown43.0 KB
references/sectors/insurance-brokers-services.mdmarkdown49.9 KB
references/sectors/insurance.mdmarkdown38.0 KB
references/sectors/it-saas.mdmarkdown44.3 KB
references/sectors/metals-mining.mdmarkdown49.4 KB
references/sectors/mortgage-reit-specialty-finance.mdmarkdown43.5 KB
references/sectors/nbfc.mdmarkdown40.3 KB
references/sectors/oil-gas.mdmarkdown43.5 KB
references/sectors/people-businesses.mdmarkdown46.3 KB
references/sectors/pharma-healthcare.mdmarkdown40.6 KB
references/sectors/rail-freight.mdmarkdown66.2 KB
references/sectors/realestate-reit.mdmarkdown44.3 KB
references/sectors/retail-ecommerce.mdmarkdown48.3 KB
references/sectors/semiconductors.mdmarkdown43.2 KB
references/sectors/shipping-logistics.mdmarkdown47.2 KB
references/sectors/telecom-media.mdmarkdown43.1 KB
references/sectors/utilities-power.mdmarkdown44.9 KB
references/sectors/waste-environmental.mdmarkdown46.5 KB
scripts/benchmarks.jsonconfig200.5 KB
scripts/lint_report.pyscript47.8 KB
scripts/ratios.pyscript127.4 KB
scripts/score.pyscript96.8 KB
scripts/valuation.pyscript37.9 KB
scripts/verify_data.pyscript72.5 KB

references/sectors/it-saas.md

View on GitHub
# IT services, software, SaaS and internet platforms — sector playbook

Use this when: the company earns most of its revenue from software licences, subscriptions, cloud/platform services, IT outsourcing, systems integration, engineering/R&D services, BPM, or internet marketplace/ad monetisation — including Indian tier-1 and mid-cap IT, global SaaS, and product companies whose balance sheet is essentially cash plus goodwill.

This is the sector where the generic ratio set fails hardest. The assets that produce the earnings — code, brand, customer relationships, trained engineers — are expensed as incurred and never appear on the balance sheet, so every ratio with capital, book value or assets in the denominator is an accounting artefact rather than an economic measure. At the same time, the largest true cost at many software firms (stock-based compensation) is excluded from the numbers management asks you to use. Your job is to replace the balance-sheet lens with a contract-and-cohort lens: retention, contracted backlog, unit economics of customer acquisition, and — on the services side — utilisation, pyramid and pricing. Almost everything that predicts the next two years is disclosed outside the financial statements.

## Contents

- [Why the generic ratio set fails here](#why-the-generic-ratio-set-fails-here)
- [The metrics that actually matter](#the-metrics-that-actually-matter)
- [How to value companies in this sector](#how-to-value-companies-in-this-sector)
- [Peer set construction](#peer-set-construction)
- [Sector-specific red flags](#sector-specific-red-flags)
- [Cycle and structural context](#cycle-and-structural-context)
- [India vs global notes](#india-vs-global-notes)
- [Checklist](#checklist)

## Why the generic ratio set fails here

**ROCE / ROIC / ROA / asset turnover — near-meaningless, and polluted in both directions.** The value-creating assets are expensed (R&D, S&M, salaries), so capital employed is a fictional denominator. Indian tier-1 IT posts 40-60%+ ROCE and US software firms post triple-digit ROIC not because they are ten times better operators than a good manufacturer, but because the denominator is structurally understated. That same denominator is then distorted the other way by (i) large net cash piles — Indian IT often holds 15-25% of market cap in cash, which drags reported ROCE *down* — and (ii) acquisition goodwill, which at software roll-ups swamps genuine operating capital. Comparing ROCE across two software firms compares accounting policies, not economics. If you use it at all, compute ROCE ex-cash and ex-goodwill, and only within the sector.

**D/E, interest cover, current ratio — non-diagnostic and often inverted.** Most quality names are net cash: D/E is ~0 and interest cover is infinite or undefined, so leverage screens produce no ranking information. Worse, the largest "liability" on a SaaS balance sheet — deferred revenue / contract liabilities — is cash already collected for services not yet delivered. It is a sign of *strength* that mechanically worsens the current ratio and D/E. A leverage screen will rank a healthy, fast-growing SaaS company as more indebted than a stagnant one. Ind AS 116 / ASC 842 lease liabilities further inflate "debt" for firms with large campus footprints without representing financial risk. Where leverage genuinely matters — PE-backed software, take-private LBOs, roll-ups, Indian mid-caps funding acquisitions — the right measure is net debt / EBITDA read against *recurring* revenue and contracted backlog, never D/E.

**P/B — unusable.** Book equity here is cash + goodwill + receivables. Sustained buybacks can drive book equity negative (a recurring feature of mature US software), making P/B negative or absurdly large. It carries zero information about earning power.

**P/E — misleading in both directions, but not uniformly.** For growth SaaS, GAAP EPS is depressed or negative *by design*: essentially 100% of customer-acquisition cost is expensed in year 0 while the revenue it buys is recognised rateably over 3-10 years. A company can be loss-making and creating enormous value; another can be profitable simply because it stopped growing. GAAP EPS is further crushed by non-cash SBC and acquired-intangible amortisation, while "non-GAAP EPS" adds back SBC, which is a real economic cost. GAAP P/E is therefore too high, non-GAAP P/E too low, and neither is comparable across companies with different growth rates. For Indian IT services the opposite holds: these are genuinely profitable, cash-generative annuity businesses and forward P/E *is* the primary tool — but read against currency, deal cycle and pyramid position, never as an absolute.

**OPM / EBITDA margin — a level without growth context tells you nothing.** A 5% margin at 40% growth is worth far more than a 25% margin at 4% growth; the growth company is *choosing* to spend the margin. Rule of 40 exists precisely to fix this. EBITDA is additionally a poor proxy for software cash economics because (i) it excludes SBC, often the single largest true cost (15-25% of revenue at many US SaaS names), (ii) capex is trivial so EBITDA ≈ EBIT and the metric adds nothing over EBIT, and (iii) firms capitalising internal software development (ASC 350-40 / IAS 38) shift opex into capex and inflate EBITDA versus peers who expense the identical activity. For Indian IT, reported OPM differences across tier-1 names reflect onsite-offshore mix, subcontractor usage, utilisation and hedging — not intrinsic efficiency.

**FCF — a billing-terms artefact, not a profitability measure (in SaaS).** A company that shifts customers to annual-upfront billing produces a step-up in FCF with zero change in economics; one that concedes monthly billing to win deals shows FCF deterioration while performing fine. Negative-working-capital models can produce FCF > net income for years. And "adjusted FCF" that ignores SBC-funded payroll overstates cash generation: the honest number is FCF less SBC, or FCF with the buyback spend needed to hold share count flat treated as an operating cost. Conversely, for Indian IT, FCF/PAT conversion is a valid and important quality test — one of the few generic metrics that survives intact.

**Inventory turns, working-capital cycle, fixed-asset turnover, capex/sales — not applicable.** No inventory; capex is typically 2-4% of revenue. Replace with DSO computed *including unbilled revenue / contract assets*, and for SaaS with DSO plus the deferred-revenue trend.

**Dividend yield / payout — ranks lifecycle, not quality.** Indian large-cap IT returns 80-100% of FCF via dividends and buybacks (yields 2-4%); US SaaS pays nothing and often has a *negative* net buyback yield after dilution. Comparing them on payout is comparing where they sit in the corporate lifecycle.

**Headline revenue growth — not comparable as reported.** Indian IT must be read in constant currency and ex-acquisitions (reported USD growth swings 200-400bps on cross-currency alone). Global software must be read organically, because roll-ups buy growth. Any "total revenue growth" that blends subscription, professional services and hardware/pass-through resale can hide subscription deceleration entirely.

## The metrics that actually matter

All ranges below are **indicative only**. They shift with market (India vs US), sub-sector, cycle phase and reporting period. A company's own 3-5 year trend and its position versus a tight peer set override any absolute band in this table.

| Metric | Definition / how to compute | Indicative healthy range | Why it matters |
|---|---|---|---|
| Net Revenue Retention (NRR / NDR) | Current-period revenue from the cohort of customers present 12 months ago ÷ that cohort's revenue then. Captures churn, downgrades, price rises, seat expansion, cross-sell; excludes all new logos. | Enterprise SaaS 115-130% (best-in-class >130%); SMB/PLG 100-115%; usage-based models can exceed 130%. <100% means the installed base is shrinking. A fall from >120% to <105% over four quarters is a structural break, not noise. | The single highest-explanatory-power variable in SaaS multiple regressions. It measures whether the product is embedded in the customer's workflow. At 125% NRR a company grows a quarter annually without selling to a single new customer. It also sets LTV, and therefore how much CAC is rational. |
| Gross Revenue Retention (GRR) and logo churn | Retention *before* any upsell: fraction of last year's revenue base surviving churn and downgrades. Logo churn is the same in customer counts. | Enterprise/mission-critical GRR 90-95%+ (best 95-98%), annual logo churn <5-8%. SMB GRR 80-88%, monthly logo churn <1.5-2%. Enterprise GRR below 85% signals a nice-to-have product. | GRR is the ceiling on business quality and the truest read on switching costs. 125% NRR built on 85% GRR plus 40 points of upsell into a few whales is far more fragile than 115% NRR on 96% GRR — the first breaks the moment budget scrutiny arrives. Firms that report NRR but refuse GRR are telling you something. |
| Rule of 40 (and its composition) | YoY revenue growth % + FCF margin % (or EBIT margin — state which, and be consistent across the peer set). | >40 is the benchmark, >50 elite, 30-40 acceptable at scale, <30 sustained is a value trap absent a credible margin path. Indian IT equivalent bar ≈ 10-12% cc growth + 20-26% EBIT margin, i.e. 32-38. | Stops you penalising a firm for deliberately spending margin to compound revenue, and stops you rewarding one that manufactured margin by starving R&D and S&M. Composition matters more than the score: 40 growth + 5 margin is worth far more than 5 growth + 40 margin. A deteriorating Rule of 40 driven by growth deceleration is the most reliable precursor of multiple compression. |
| CAC payback / LTV:CAC / Magic Number | CAC payback = S&M spend ÷ (next-period net new ARR × subscription gross margin), in months. Magic Number = annualised change in quarterly revenue ÷ prior-quarter S&M. LTV:CAC = (ARPA × GM ÷ churn) ÷ CAC. | CAC payback <12m excellent, 12-18 healthy, 18-24 acceptable for enterprise with high GRR, >24-30 broken. Magic Number >0.75 = scale spend, <0.5 = stop. LTV:CAC >3x; <1x destroys value per customer won. | This is the actual capital-allocation decision inside the business. Because S&M is expensed but buys a multi-year asset, the income statement cannot tell you whether growth spend is investment or waste — these ratios can. Deteriorating CAC payback while growth still looks fine is the earliest sign of saturation or competitive pressure, typically 2-4 quarters ahead of a growth collapse. |
| Subscription gross margin (split from services GM) | GM on recurring software revenue only, after hosting/cloud cost, support, third-party licence fees and capitalised-software amortisation — reported separately from professional-services GM. Services equivalent: gross margin before SG&A. | Pure SaaS subscription GM 75-85%; infrastructure- or inference-heavy 60-72%; <65% needs explanation. Professional services GM 0-20%, often negative — fine if small. Blended >70%. Indian IT gross margin 30-36%; EBIT 20-26% tier-1, 14-19% mid-cap. | Sets the marginal economics and the LTV ceiling. Blended margin hides mix: a company whose growth is increasingly professional services and pass-through resale is degrading in quality while headline revenue looks fine. In the AI era this is the metric to watch hardest — GPU/inference cost is compressing subscription GM, and a 10-point GM decline destroys the unit economics that justified the multiple. |
| SBC % of revenue and net diluted share count growth | Total SBC ÷ revenue; YoY change in fully diluted shares after buybacks; cash spent on buybacks purely to offset dilution. | US SaaS: <10-12% of revenue at scale is good, 15-20% high, >25% is a transfer of ownership to employees. Net dilution <1-2% p.a. Indian IT: typically <1-2% of revenue — a genuine structural advantage. | The sector's defining accounting distortion. Non-GAAP EPS, "adjusted EBITDA" and "adjusted FCF" all add SBC back, yet it is real compensation that either dilutes you or forces cash buybacks. A firm with 20% SBC/revenue and a 25% adjusted FCF margin has roughly zero economic FCF margin. Always recompute FCF − SBC and treat dilution-offsetting buybacks as opex, not return of capital. |
| Billings, RPO and cRPO growth | Billings ≈ revenue + change in deferred revenue. RPO = total contracted revenue not yet recognised; cRPO = the portion due within 12 months. Track each growth rate against reported revenue growth. | cRPO growth at or slightly above revenue growth in a healthy company. Book-to-bill / TCV-to-revenue >1.0x, 1.1-1.3x indicating a strong services deal cycle. RPO ÷ annual revenue of 1.5-3x typical for enterprise SaaS. | Revenue is a lagging indicator — the rateable release of contracts signed earlier. cRPO and billings turn 2-3 quarters before revenue. Use cRPO, not total RPO: total RPO can be inflated by one 7-year mega-deal or a bulk renewal, masking a collapse in near-term demand. Decelerating cRPO with steady revenue is the classic pre-derating setup. |
| Constant-currency organic growth (with the full bridge) | Reported growth = organic cc growth + inorganic + FX. Strip FX translation and any revenue acquired in the last 12 months. | Indian tier-1: 8-12% cc in an upcycle, 4-7% mid-cycle, 0-3% a downturn; mid-caps need 12-18% cc to justify a premium. Enterprise SaaS at $1bn+ scale: 15-25% strong; <10% means value it as a mature software company. | Reported growth swings hundreds of basis points on cross-currency alone, and roll-ups routinely mask organic deceleration with M&A. The bridge is where managements hide the story. Note the INR asymmetry: rupee depreciation flatters Indian IT margins and INR EPS even when the underlying dollar business is shrinking — never judge Indian IT on INR revenue growth. |
| Utilisation (ex-trainees) and offshore mix | Billable ÷ available hours for delivery staff, disclosed both including and excluding trainees; share of effort and of revenue delivered offshore vs onsite. | Utilisation ex-trainees 82-86% is the effective ceiling; above ~87% signals no bench and future delivery/attrition risk. Offshore effort mix 72-78%, offshore revenue mix 50-55%. Roughly 100bps of utilisation ≈ 25-30bps of EBIT margin. | In a people business, margin is manufactured from utilisation, pyramid (fresher share), offshore mix, pricing and subcontractor cost — nothing else. Knowing which lever produced a beat tells you whether it repeats: a beat from utilisation at 87% is borrowed from next year; one from offshore mix or pricing is durable. Utilisation also falls *before* revenue when the pipeline dries up. |
| Attrition (LTM voluntary) and headcount vs revenue | LTM voluntary attrition in delivery staff; quarterly net adds and fresher hiring, compared against cc revenue growth. | Indian IT normalised attrition 12-15%; >18-20% is a wage-cost and delivery-quality crisis; <11% often signals a weak external job market and therefore weak demand. Headcount growth ≈ cc revenue growth minus 2-4% productivity. | Attrition is simultaneously a cost signal (backfill at market wages, more subcontracting), a delivery-risk signal (escalations, fixed-price overruns) and a demand signal. Divergence is equally diagnostic: revenue up on flat headcount means utilisation/pricing gains that eventually exhaust; headcount outrunning revenue means pyramid deterioration and margin pressure ahead. Most exposed metric to GenAI disruption of the linear people-to-revenue model. |
| Revenue per employee (and its trend) | Annual USD revenue ÷ average headcount, tracked over 3-5 years. | Indian tier-1 $50,000-62,000, with 2-4% annual improvement a good outcome. Accenture and Western peers $95,000-120,000 (different onsite mix). Product SaaS at scale $250,000-450,000+; best-in-class PLG >$500,000. | The cleanest cross-check on both the AI/non-linearity thesis and pricing power. If a services firm claims automation, platform IP and outcome-based pricing, revenue per employee must rise; flat for five years means the "non-linear revenue" story is marketing. For software it is the operating-leverage proxy that ROCE cannot provide in an asset-light model. |
| Deal TCV, book-to-bill, client concentration, account ladder | Quarterly TCV split net-new vs renewal; book-to-bill; top-5/top-10 client revenue share; count of clients >$1m / $50m / $100m. SaaS equivalents: customers >$100k ARR and >$1m ARR and their growth. | Book-to-bill >1.0x, ideally 1.1-1.3x. Top-10 concentration <22-28% for large IT services; any single client >10% of revenue is a material risk. Growth in $1m+ accounts should outpace total customer-count growth. | TCV is the forward order book; the net-new vs renewal split separates real share gain from defensive re-signing. The account ladder shows whether the company is moving upmarket into stickier, higher-NRR relationships or drifting into churn-prone SMB. Concentration risk is acute here: one client's insourcing decision, bankruptcy or vendor consolidation can remove several points of growth overnight. |
| DSO including unbilled revenue / contract assets | DSO computed on receivables **plus** unbilled revenue (contract assets), not receivables alone. Track the unbilled component separately as a % of revenue. | Indian IT 65-78 days including unbilled is normal; >85 days, or a 10-day YoY rise, warrants investigation. SaaS 45-70 days depending on billing seasonality (Q4-heavy billers spike). Unbilled should be stable as a % of revenue. | In percentage-of-completion fixed-price contracts — a growing share of IT services revenue — unbilled revenue is where aggressive recognition lives: revenue booked for work not yet invoiced or client-accepted. Unbilled growing materially faster than revenue is the single most common accounting warning in Indian IT. It also flags client financial stress and pricing concessions dressed as payment-term extensions. |
| R&D % of revenue (software) / subcontractor cost % (services), plus FCF/PAT conversion | Software: R&D ÷ revenue and the capitalised share of it. Services: subcontracting/contractor cost ÷ revenue. Both: FCF ÷ net income. | SaaS R&D 15-25% of revenue at scale; a drop below 12% while claiming product leadership is harvesting. Capitalised software should be a small, stable share of R&D. Services subcontractor cost 6-9%; >11-12% compresses margin. FCF/PAT >85-90% for Indian IT; >100% for SaaS with upfront billing (recompute after SBC). | R&D intensity is the reinvestment rate that sustains the moat, and the expensed-vs-capitalised split is where reported margins are manufactured. Subcontractor cost makes an Indian IT capacity constraint visible — it spikes when demand outruns hiring or niche skills (cloud, AI, SAP) are unavailable in-house. Cash conversion is the honest cross-check that reported profits are real. |
| Pricing and duration signals (secondary but decisive) | Realised price per seat / per unit of consumption; average contract duration; discount depth at renewal; share of multi-year and of usage-based contracts. | Flat-to-rising realised pricing is healthy; persistent discounting to hold logos is not. Lengthening duration with deepening discounts is a demand tell, not a strength. | Retention can be bought. NRR that holds only because renewals are discounted 15-20% shows up here long before it shows up in NRR. In IT services, rate-card compression and vendor consolidation rebids are the same phenomenon. |

### If the company is an internet platform or marketplace

Platform/marketplace names sit in this sector but need three additional metrics, because "revenue" for them is a net take on someone else's gross flow and is not comparable to subscription revenue.

| Metric | Definition / how to compute | Indicative healthy range | Why it matters |
|---|---|---|---|
| GMV / gross bookings and take rate | Gross transaction value flowing through the platform; take rate = net revenue ÷ GMV. Check whether revenue is reported gross (principal) or net (agent) under Ind AS 115 / ASC 606. | Take rate varies enormously by category — marketplaces 8-20%, payments 1-3%, food/mobility 15-25%. What matters is the *trend* and whether it is rising through pricing or through mix into higher-take ancillary services (ads, logistics, credit). | Gross-vs-net presentation changes reported revenue by an order of magnitude with zero economic difference, which makes cross-company EV/revenue meaningless unless you normalise to GMV or gross profit. Take-rate expansion without merchant churn is real pricing power; take-rate expansion with rising incentive spend is not. |
| Contribution margin per transaction / per cohort, after incentives | Net revenue less directly attributable variable cost *and* all customer/merchant incentives, discounts and cashbacks — not the version that classifies incentives as marketing. | Positive and improving with scale. Negative contribution margin at scale means growth destroys value. | Incentives are frequently booked as a revenue reduction in one company and as S&M in another, which flips both revenue and margin optics. Recompute on a consistent basis before comparing. This is the marketplace analogue of subscription gross margin. |
| Engagement and cohort repeat behaviour | MAU/DAU and the DAU/MAU ratio; transactions per active user per year; cohort retention curves by signup month; share of GMV from cohorts older than 24 months. | Flat-to-rising transactions per user and cohort curves that flatten rather than decay to zero. Older cohorts should contribute a rising absolute GMV share. | User counts can be bought with incentives; cohort behaviour cannot. A platform whose newer cohorts monetise worse than older ones is buying growth from a shrinking-quality funnel, and reported user growth will mask it for 4-6 quarters. |

### When a metric is not disclosed

Do not silently drop it. Either (a) proxy it and label the proxy — deferred-revenue growth as a billings proxy, S&M ÷ net new ARR as a rough CAC payback, headcount disclosures from annual reports where quarterly data is absent — or (b) record the absence as a finding. Non-disclosure of NRR, GRR or net-new TCV in a company whose peers all disclose them is itself evidence, and should reduce your confidence in the thesis rather than being treated as neutral.

## How to value companies in this sector

There is no single sector multiple. The right tool depends on where the company sits on the growth/profitability curve, and the sector splits into two genuinely different valuation regimes.

**1. High-growth SaaS (growth >20%, GAAP-unprofitable or thin-margin): EV/NTM revenue, growth-adjusted.** Earnings multiples are undefined or absurd, so the market prices forward revenue. Use EV (market cap + debt − cash; note that convertible-funded cash balances are common) over *next-twelve-month* revenue, not trailing. The multiple is not a number to compare naively — it is the output of a cross-sectional regression whose dominant explanatory variables are forward growth, NRR and Rule of 40. Practical method: fit EV/NTM revenue against Rule of 40 (or against growth, with NRR and gross margin as secondary terms) across the peer set and ask whether the stock sits above or below the line. A crude but useful normaliser is EV/NTM revenue ÷ NTM growth %. Absolute levels are regime-dependent — public SaaS median EV/NTM revenue has ranged from roughly 4x in the 2022-23 trough to above 12x at the 2021 peak, with high-growth/high-NRR names at 2-3x the median. Only *relative* positioning is stable across regimes; never anchor on a multiple level from a different regime.

**2. EV/gross profit — the better cross-sectional comparator.** Because subscription gross margins range from ~60% (infrastructure- or inference-heavy) to ~85% (pure application software), EV/revenue systematically overvalues low-margin businesses. EV/gross profit normalises this and is the right multiple when comparing an AI-native company carrying GPU cost of goods against a classic 80%-margin vendor. It is becoming the preferred metric as AI COGS reshapes the sector.

**3. EV/ARR and EV/cRPO — for the recurring core.** Where reported revenue blends subscription, professional services and pass-through hardware, value the recurring base on EV/ARR and treat services at a low multiple (1-2x revenue) or as a cost centre. EV/cRPO is the sharpest tool late in a cycle because contracted backlog is harder to manipulate than recognised revenue. EV/ARR is also the standard currency in private and M&A comps, which makes it the bridge for take-private scenarios.

**4. Mature software and Indian IT services: forward P/E, FCF yield, DCF.** Once growth falls below ~15% and margins normalise, the sector reverts to conventional earnings valuation. For Indian IT this is the primary framework from the start — these are structurally profitable, capital-light, negative-working-capital annuity businesses with 80-100% payout. Indian convention is forward P/E on 1-year-forward EPS: tier-1 has historically traded around 20-28x forward with the highest-quality names at a premium and laggards at a discount; quality mid-caps have commanded 30-45x forward on faster growth. Cross-check with PEG and with FCF yield plus buyback yield (Indian IT total shareholder yield typically 3-5%). Important mechanical point: EV/EBITDA looks artificially cheap for Indian IT because of the large net cash pile — always state whether you are on an EV or equity basis, and prefer valuing operations ex-cash and adding cash back at (or below) book, since idle cash earning treasury yields deserves no operating multiple.

**5. DCF and cohort valuation — unusually credible here.** Because revenue is contracted and retention is measurable, SaaS cash flows are more forecastable than in most sectors. Decompose: value the existing customer base as an annuity decaying at the GRR rate and growing at the expansion rate, then value future customer acquisition separately using CAC-payback economics. This tells you how much of the market cap is embedded in customers already won versus customers yet to be acquired — often 40-60% of a high-growth SaaS valuation is the latter, which is precisely where the risk sits. In any DCF or FCF-yield exercise, treat SBC as a cash cost (deduct it, or model share-count growth explicitly). The most common valuation error in this sector is discounting SBC-adjusted cash flows to an unadjusted share count.

**6. What not to use.** P/B — book value is cash plus goodwill and can be negative after buybacks. EV/EBITDA — capex is negligible so EBITDA ≈ EBIT and it adds nothing, while rewarding firms that capitalise software development. Replacement cost, NAV and asset-based methods do not apply. Dividend discount models apply only to mature Indian IT and legacy licence software. Trailing P/E on GAAP earnings for a growth SaaS name is not a conservative choice — it is a meaningless one.

**7. Cross-checks before you commit to a fair value.** Compare the implied terminal EBIT margin against a genuinely mature comparable (the 25-30% zone that scaled software and Indian tier-1 IT actually achieve) rather than an aspirational one. Sanity-check implied revenue against a realistic TAM and market share. Stress-test NRR: a 10-point NRR decline compounds through the model and typically justifies a 25-40% cut to fair value on its own — run it before you write the target. And check that the FX assumption embedded in an Indian IT model is not doing the work that operations should be doing.

## Peer set construction

A valid comparable here shares **revenue model, growth stage, gross-margin structure and end-market**, not merely the "technology" label. Mixing these produces conclusions that are artefacts of the peer set.

Splits that must not be mixed:

- **Product software vs IT services.** Different gross margins (75-85% vs 30-36%), different scaling laws (marginal cost near zero vs linear headcount), different valuation regimes. Never put a SaaS vendor and an outsourcing firm in the same multiple table.
- **Recurring subscription vs perpetual-licence-plus-maintenance vs consumption/usage-based.** Usage-based revenue is genuinely more volatile and re-rates faster in a downturn; licence revenue is lumpy and front-loaded. Their NRR figures are not comparable definitions.
- **Enterprise vs SMB vs PLG/self-serve.** GRR bands differ by 10+ points, CAC payback and sales-motion economics differ entirely, and SMB churn is macro-sensitive in a way enterprise churn is not.
- **Indian tier-1 vs Indian mid-cap vs global SI (Accenture-type) vs captive/GCC-exposed ER&D.** Different growth, margin, client-size and revenue-per-employee bands, and different currency exposure.
- **Vertical/domain concentration.** BFSI-heavy, healthcare-heavy, retail-heavy and hi-tech-heavy books behave differently across a cycle; a hi-tech-concentrated services firm in a client capex freeze is not comparable to a BFSI-led peer in a rate-driven spending cycle.
- **Growth stage.** A 35%-growth name and a 6%-growth name are in different regimes even with identical products. Bucket by growth first, then compare.
- **Organic vs roll-up.** Serial acquirers carry goodwill, purchase-accounting effects and perpetual "one-off" charges. Compare them to each other, and always on organic cc growth.
- **AI-native / inference-heavy vs classic application software.** Different COGS structure; use EV/gross profit, not EV/revenue, when they must sit in one table.

Also normalise before comparing: put every peer on the same currency basis (cc where possible), the same margin definition (EBIT or FCF, stated), the same Rule of 40 formulation, and the same fiscal-year alignment — Indian companies report April-March, many US software companies use January or June year-ends, and a raw calendar comparison in a fast-moving cycle mis-ranks them.

Aim for 5-8 tight comparables. If you cannot find them, say so and widen to a global set explicitly rather than silently comparing across sub-sectors.

## Sector-specific red flags

- **SBC excluded from every headline number.** Non-GAAP EPS, "adjusted EBITDA" and "adjusted FCF" all add back stock comp while the company spends most of its FCF on buybacks that merely hold share count flat. Test: compute FCF − SBC, and check net diluted share count over 3-5 years. If shares are up 3-5% p.a. despite large buybacks, the reported "cash generation" is largely being paid to employees, not owners.
- **Unbilled revenue / contract assets growing materially faster than revenue.** The classic fixed-price and Indian IT warning: percentage-of-completion revenue for work not yet invoiced or client-accepted. Watch DSO-including-unbilled creeping up 8-15 days, especially alongside a margin beat — it usually means revenue was pulled forward or a client is in distress.
- **cRPO or billings decelerating while reported revenue holds up.** Revenue is backlog burning off; cRPO is the truth. Related tell: management stops guiding on billings, shifts emphasis from cRPO to total RPO, or headlines an RPO number inflated by one long-duration mega-deal or bulk renewal.
- **Metric disappearance or redefinition.** A company that disclosed NRR, GRR, customers above $100k ARR, CAC payback or attrition every quarter for years and then quietly stops, changes the definition, or moves to annual disclosure "for consistency". This is the most reliable qualitative red flag in the sector and it almost always precedes a bad print. Equally: ARR redefined to include non-recurring services, usage estimates, "committed ARR", or partner/reseller bookings.
- **Rising capitalisation of internal software development.** Check capitalised software on the balance sheet and the capitalised share of total R&D. Higher capitalisation shifts cash costs out of opex, flatters margins and EBITDA, and defers the hit into amortisation. Margins improving while capitalised software jumps is not improvement.
- **Acquired growth presented as organic.** Serial roll-ups obscure a decaying core — demand the organic cc bridge. Corroborating signals: ballooning goodwill and acquired-intangible amortisation excluded from adjusted profit, purchase-accounting deferred-revenue haircuts that flatter subsequent growth, and integration/restructuring charges recurring every single year (i.e. they are operating costs).
- **NRR propped up by a few accounts while GRR erodes.** If enterprise GRR drifts toward the mid-80s while NRR still looks fine, the base is churning and being papered over with upsell into whales. Cross-check net customer adds: adding revenue while losing logos means the runway is shortening.
- **Margin beats from non-repeatable levers.** In IT services: utilisation pushed above ~87% (no bench, delivery risk, future attrition), a sharp cut in fresher hiring, a one-off drop in subcontractor cost, or FX/hedging gains presented as operating performance. In SaaS: a large cut to R&D or S&M that props up the Rule of 40 while growth quietly decelerates — margin harvested from the future.
- **Cash flow flattered by working-capital and billing manoeuvres.** Aggressive discounting to convert customers to multi-year upfront prepay (borrowing FCF from future years), receivables factoring or supply-chain finance, stretched payables, lengthening the amortisation period on capitalised sales commissions (ASC 606 contract cost assets) to lift reported margin, and channel/reseller stuffing at quarter-end. One spectacular FCF quarter with flat billings deserves suspicion.
- **Concentration risk being waved away.** A top client above 10% of revenue, top-10 above 30%, or heavy exposure to one stressed vertical. In services, also watch a mega-deal won on rebadging/transition terms that is margin-dilutive for 2-3 years but announced as record TCV.
- **Headcount and revenue diverging in the wrong direction.** Revenue flat-to-up while headcount falls sharply can be genuine automation, but is more often a one-time utilisation-and-pyramid squeeze that cannot repeat. The reverse — headcount growing well ahead of revenue — signals pyramid deterioration, bench build-up and margin compression 2-3 quarters out. Treat unquantified "AI-led productivity" claims sceptically unless revenue per employee is actually rising.
- **TCV and bookings disclosure games.** TCV bundling renewals, extensions and rebadged headcount without a net-new split; TCV including optional or contingent scope; a changed duration convention. A record TCV quarter with no book-to-bill and no net-new split is a marketing number.
- **Governance and personnel tells, unusually predictive here.** CFO or auditor turnover (especially Indian mid-cap IT), whistleblower complaints on revenue recognition, restated segment reporting, promoter pledging, related-party transactions with promoter-linked entities, revenue routed through associates or opaque subsidiaries, option repricing after a share-price fall, and heavy insider selling into a rally.
- **AI-era specific.** Subscription gross margin compression from inference/GPU cost attributed vaguely to "mix"; "AI ARR" or "AI pipeline" figures disclosed without definition or without appearing in reported revenue; and, on the services side, silence about the deflationary risk GenAI poses to time-and-materials, headcount-linked pricing. Also check the contractual-obligations note for large committed cloud/GPU purchase commitments — these are debt-like fixed costs sitting on an otherwise asset-light balance sheet.

## Cycle and structural context

**The demand cycle is a client-budget cycle, and it is observable early.** Enterprise IT spending tracks corporate profit expectations and rate levels with a 2-3 quarter lag. The order in which it shows up is consistent: discretionary/transformation spend is cut first (project ramp-downs, deferred starts), then pricing pressure and vendor consolidation rebids, then headcount and utilisation, and only then reported revenue. Cost-takeout and vendor-consolidation deals actually *increase* in a downturn — large multi-year outsourcing TCV can hit records while organic growth is collapsing, because the deals are long-duration, margin-dilutive in the transition years, and often involve rebadging client staff. Read a record-TCV headline against net-new TCV and near-term revenue, never in isolation.

**Where in the cycle matters differently by sub-sector.** Consumption/usage-based software de-rates fastest into a downturn (customers throttle usage within the quarter) and recovers fastest. Seat-based enterprise subscription lags in both directions because contracts are annual. SMB-exposed SaaS carries credit-like churn risk in a recession. IT services sits between: revenue is annuity-like, but discretionary project work is the swing factor and margin is levered to utilisation.

**Structural threats to underwrite explicitly.** (i) GenAI as a deflationary force on headcount-linked pricing — if a services firm's price per unit of output falls faster than its cost per unit, revenue shrinks even as it wins share; revenue per employee is the check. (ii) GenAI as a moat threat in software — thin application layers over a general-purpose model, seat-based pricing under pressure as agents replace seats, and vendors' own COGS rising with inference. (iii) Captive/GCC insourcing by large clients, which removes both revenue and the best delivery talent. (iv) Cloud-vendor platform encroachment on point-solution SaaS categories. (v) Consolidation of vendor rosters, which is good for scale players and lethal for tier-3.

**Currency and rates.** Indian IT earns in USD/EUR/GBP and pays largely in INR: rupee depreciation flatters INR margins and EPS while cross-currency (EUR/GBP/AUD vs USD) moves reported USD growth by hundreds of basis points. Hedge books smooth this with a 2-4 quarter lag, so an FX-driven margin beat is a timing effect, not performance. For unprofitable global SaaS, the discount rate is the dominant valuation variable — a duration asset whose value sits mostly in terminal years re-rates violently with long rates, independent of operations.

**Regulation and policy.** Data localisation and cross-border transfer rules (GDPR, India's DPDP Act, sectoral rules in BFSI and healthcare) raise delivery cost and can force in-country capacity. Visa and immigration policy in the US/UK affects onsite cost and mix for Indian IT. Antitrust and platform regulation bear on internet/marketplace names (app-store rules, self-preferencing, ad-tech). AI-specific regulation (EU AI Act and analogues) creates compliance cost and, for some vendors, a moat. Public-sector and BFSI clients add procurement-cycle and audit risk. Cyber-incident materiality disclosure rules mean a breach is now a reportable, quantifiable event — check for prior incidents and the associated contractual liabilities.

## India vs global notes

**Reporting and filings.** India: Ind AS, quarterly results filed with NSE/BSE within 45 days, figures in ₹ crore/lakh, mandatory quarterly earnings calls and investor presentations at large caps (the concall is where utilisation, attrition, offshore mix, client-count buckets and TCV are actually disclosed — the financial statements alone will not give you the operating KPIs). Annual report includes MD&A, CARO 2020 auditor reporting, and the auditor's report on internal financial controls. US/global: 10-K/10-Q on EDGAR, GAAP with a reconciled non-GAAP section, 8-K for material events, and the shareholder letter / supplemental metrics deck where NRR, RPO, cRPO and customer cohorts live.

**Disclosure asymmetries to expect.** Indian IT discloses operating KPIs generously (utilisation, attrition, offshore/onsite, client buckets, TCV) but rarely NRR/GRR in SaaS form. US SaaS discloses retention and RPO but almost never utilisation or attrition. Do not conclude a metric is absent because it is absent from the financials — check the concall transcript (India) or the supplemental metrics deck / prepared remarks (US) first.

**India-specific items to check.** Promoter holding and any pledging; promoter-group related-party transactions; CARO qualifications; RPT approvals and the audit committee's composition; ESOP pools at mid-caps (small vs US peers but rising); dividend and buyback policy (many large caps run explicit capital-return policies returning 80-100% of FCF); SEBI LODR disclosures on material events; and — for mid-caps — auditor changes, which have historically been an early warning.

**Accounting and convention differences that change the numbers.** Ind AS 115 and ASC 606 are converged on revenue recognition, so percentage-of-completion and contract-asset treatment are broadly comparable — but the *disclosure granularity* differs, and Indian firms often report unbilled revenue within trade receivables notes rather than as a headline. Ind AS 116 and ASC 842 both put leases on the balance sheet; strip them before any leverage comparison. Hedge accounting under Ind AS 109 pushes forward-contract gains through OCI and the P&L on different lines across companies — read the hedge note before attributing a margin move to operations. SBC is expensed under both frameworks, but the *magnitude* differs by an order of magnitude between Indian IT and US SaaS, which is why non-GAAP and GAAP converge in India and diverge sharply in the US.

**Valuation convention.** India quotes forward P/E on 1-year-forward EPS as the default and discusses growth in cc USD terms; the US quotes EV/NTM revenue for growth software and non-GAAP EPS for mature software. When comparing, convert explicitly and state the basis. Note also that Indian IT's large net cash makes EV-based multiples look optically cheap versus equity-based ones — pick one and be consistent across the peer table.

**Ownership and index effects.** Indian IT has large FII ownership and is index-heavy, so it carries flow-driven volatility unrelated to fundamentals; US SaaS is exposed to index/ETF and factor rotation between growth and value. Neither changes intrinsic value, but both explain multiple moves you should not attribute to operations.

## Checklist

- [ ] Classify the company first: product software / SaaS, IT services, or platform — and its growth bucket. Everything downstream depends on this.
- [ ] Discard ROCE, ROA, P/B, D/E, current ratio and EV/EBITDA as primary metrics; if used at all, state the adjustment (ex-cash, ex-goodwill, ex-lease) explicitly.
- [ ] Pull NRR and GRR together. NRR without GRR is unreadable. Note if GRR is undisclosed.
- [ ] Compute Rule of 40 and, more importantly, its composition (growth vs margin). Track its 8-quarter trend.
- [ ] Check cRPO and billings growth against revenue growth. Divergence is a leading indicator; use cRPO, not total RPO.
- [ ] Build the growth bridge: reported = organic cc + inorganic + FX. Never quote reported growth without it.
- [ ] Recompute FCF − SBC and net diluted share count over 3-5 years. Treat dilution-offsetting buybacks as an operating cost.
- [ ] Split subscription gross margin from services gross margin; check the trend for inference/GPU cost creep.
- [ ] Compute CAC payback and Magic Number; a deteriorating trend leads a growth collapse by 2-4 quarters.
- [ ] For services: utilisation ex-trainees, offshore mix, attrition, subcontractor cost, headcount vs revenue, revenue per employee. Identify which lever drove any margin change and whether it repeats.
- [ ] Compute DSO *including unbilled revenue* and track the unbilled share of revenue. Rising unbilled alongside a margin beat is the top accounting warning here.
- [ ] Check client concentration (top-1 >10%, top-10 >30%) and vertical concentration against the current macro stress point.
- [ ] For platforms/marketplaces: normalise gross-vs-net revenue presentation, compute take rate and post-incentive contribution margin, and check cohort monetisation by vintage.
- [ ] Check capitalised software as a share of R&D and whether it is rising; check capitalised commission amortisation periods.
- [ ] Compare against 5-8 true comparables matched on revenue model, growth stage, gross-margin structure and end-market; state the peer set and why.
- [ ] Value with the regime-appropriate tool: EV/NTM revenue (growth-adjusted) or EV/gross profit for growth SaaS; EV/ARR or EV/cRPO for mixed-revenue firms; forward P/E, FCF+buyback yield and DCF for mature software and Indian IT.
- [ ] In any DCF, treat SBC as a cash cost and model share count explicitly; decompose value into existing-cohort annuity vs future acquisition.
- [ ] Stress-test a 10-point NRR decline and an adverse FX move before writing a fair value.
- [ ] Scan the KPI history for metrics that were discontinued, redefined, or moved to annual disclosure.
- [ ] India: check promoter pledging, RPTs, CARO qualifications, auditor/CFO changes, and read the concall for operating KPIs absent from the financials.
- [ ] State explicitly which conclusions rest on management-defined, unaudited metrics (ARR, TCV, pipeline, "AI ARR") and how much of the thesis breaks if those definitions change.
add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks.Stars: 115kUpdated Claude CodeCodexVerifiedInstall