Imported from JiayiV/question_mixing (skills/public-02-claude-skills-main-finance-skills-stock-analysis-references-sectors-exchanges-payments-7e83d858/SKILL.md). Install upstream with npx skills add JiayiV/question_mixing --skill public-02-claude-skills-main-finance-skills-stock-analysis-references-sectors-exchanges-payments-7e83d858. Copyright stays with the author.

Exchanges, depositories, clearing houses, rating agencies and payment networks — sector playbook

Use this when: the company sits between two sets of counterparties and charges a toll on every transaction that passes through — stock, commodity and derivative exchanges; depositories and registrars; central counterparties and clearing corporations; credit rating agencies; index and market-data providers; card networks; and merchant acquirers, payment aggregators and processors.

These are volume × take rate businesses running on an almost entirely fixed cost base. Once the matching engine, the settlement rails or the scoring methodology exist, the marginal cost of the next transaction is close to zero, so nearly every incremental rupee of net revenue drops to EBIT — in both directions. That produces margins (50–70%) and returns on capital that look like accounting errors to a generic screener, alongside balance sheets stuffed with client margin money, settlement float and default funds that are not the company's assets and not the company's debts. Two further distortions dominate: payment companies can report the same economics as either gross or net revenue depending on a principal-vs-agent judgement, and almost every entity here operates under a licence regime that both bars entry and caps price. Analyse this sector as: volumes × net take rate, flow-through margin on a fixed cost base, and the two tail risks — a regulator that resets the price, and a clearing member that defaults.

All ranges below are indicative only. They vary by market, product, cycle and regulatory regime. The company's own 5–10 year history and its closest sub-sector peers override every absolute band in this file.

Contents

• Why the generic ratio set fails here
• The metrics that actually matter
• How to value companies in this sector
• Peer set construction
• Sector-specific red flags
• Cycle and structural context
• India vs global notes
• Checklist

Why the generic ratio set fails here

Before computing a single ratio, do two normalisations. Everything else depends on them.

Normalisation 1 — get to net revenue. Reported "revenue" in this sector routinely contains money the company never keeps:

• Merchant acquirers, payment aggregators and processors may report gross revenue including interchange paid to the card issuer and scheme fees paid to the network — economics they merely pass through. Whether they do depends on a principal-vs-agent judgement under Ind AS 115 / ASC 606. The same business reports ~15–25% EBITDA margin on gross revenue and ~40–55% on net revenue. Comparing a gross reporter to a net reporter on OPM, EV/Sales or revenue growth is meaningless; one of the two numbers is roughly four to six times the other.
• Card networks report gross revenue and then subtract very large client incentives and rebates paid to issuers and acquirers to win portfolios. Incentives commonly run in the 30%+ range of gross revenue and are the real pricing mechanism. Only net revenue is analysable.
• US equity exchanges report transaction revenue gross of liquidity rebates and routing costs under maker-taker pricing. Gross transaction revenue can be several times the net capture. Use "net revenue" (revenue less transaction-based expenses) as disclosed.
• Indian exchanges collect SEBI turnover fees, and (as agent) STT/CTT and stamp duty. Confirm what sits inside revenue and what is netted; regulatory fee pass-throughs are not earnings.

Normalisation 2 — strip client and clearing balances off the balance sheet. Exchanges, clearing corporations, depositories and acquirers carry client margin money, settlement obligations, core settlement guarantee funds, investor protection funds and in-transit merchant settlement balances as both an asset and an equal-and-opposite liability. These are custodial. Leave them in and every capital-based ratio is fiction.

With those two done, here is what the standard set does:

OPM / EBITDA margin — not wrong, but non-comparable and easily gamed. A 60% operating margin here is normal, not exceptional, and says nothing on its own. Its level is set almost entirely by the gross-vs-net reporting choice and by product mix (a market-data or index business runs at 60–75%; a merchant-acquiring business at a fraction of that). What matters is not the level but the incremental (flow-through) margin — the share of each additional rupee of net revenue that reaches EBIT.

ROCE / ROE — mechanically broken in both directions, and never in one. Inflate capital employed with clearing member margin and the SGF, and ROCE prints in low single digits for a business with no real capital needs. Strip those out and you get returns of 40–100%+, because the true invested capital is a data centre and some software. Meanwhile mature global networks and rating agencies have bought back so much stock that book equity is small or negative — ROE is then undefined, infinite or negative and carries zero information. Conversely, payment processors built by large acquisitions carry goodwill and acquired intangibles that swamp capital employed and depress ROCE toward the cost of capital, regardless of the underlying franchise. Use ROIC on tangible operating capital excluding client funds, and separately ask whether acquisition prices were ever earned back.

D/E, net debt and interest cover — the denominator problem. Client margin, settlement obligations and default funds are not debt; a clearing corporation that "owes" a member their margin is not levered. Equally, restricted regulatory capital, core SGF contributions and IPF balances are not cash available to the shareholder and must never be netted against borrowings. Where equity has been bought back to near zero, D/E is arithmetically meaningless. Assess leverage as net debt / EBITDA on genuinely unrestricted cash and genuine borrowings, and test regulatory net-worth headroom separately.

Current ratio and working capital — undefined in substance. Settlement receivables and settlement payables are of near-identical size and settle within a day or two, so the current ratio hovers around 1.0 whatever the health of the business. Debtor days for an exchange or a network are a measure of clearing-cycle mechanics, not of credit quality. (Rating agencies are the exception — they carry real issuer receivables and their debtor days do carry information.)

FCF — high quality, but the cash flow statement is contaminated by float. These businesses genuinely convert 85–105% of net income to free cash flow, with capex typically 4–10% of net revenue. But the working-capital line in the cash flow statement can swing wildly with settlement timing across a period end — an acquirer or clearing entity can show thousands of crore of "working capital release" that is purely a calendar artefact. Compute FCF excluding movements in client, settlement and margin balances, and check it against net income over a 3–5 year window rather than for one period.

EV/EBITDA — the EV is usually wrong before the multiple is even computed. Cash on an exchange's balance sheet is largely not its own; equity may be negative; restricted funds cannot be netted. Build EV as market cap + genuine borrowings − genuinely unrestricted own cash, and only then compute the multiple. Also note that EBITDA ignores the amortisation of acquired intangibles, which is precisely where the roll-up processors' economics are buried.

P/E — the least bad headline metric, but contaminated by treasury income. Exchanges, clearing corporations and depositories earn substantial investment income on their own corpus and on float. In a high-rate period this can be a large share of PBT and it is not operating performance; it also de-rates when rates fall. Always compute core P/E on operating PAT excluding other/treasury income, and strip the associated investable corpus out of market cap before comparing.

Sales/asset turnover, inventory turns, cash conversion cycle — undefined or trivially meaningless. Do not report them.

The metrics that actually matter

Ranges are indicative and strongly sub-sector specific. Judge within sub-sector and against the company's own history.

How to value companies in this sector

Default: core P/E and DCF, with a mix-aware sum-of-the-parts. Do not lead with EV/EBITDA and never with P/B.

1. Start from core earnings. Core PAT = operating profit excluding other/treasury income, taxed at the effective rate. Then value the investable corpus separately at (or slightly below) carrying value, excluding restricted funds — SGF, IPF and regulatory capital — entirely. For Indian MIIs in particular, a headline P/E computed on total PAT while a large unrestricted corpus sits inside market cap makes a business look far cheaper than it is.

2. DCF works genuinely well here — better than in most sectors — because volumes are forecastable over a cycle, margins are stable, capex is low and working capital is nil. Build it as volume × take rate × flow-through margin. The three assumptions that decide the answer are terminal take rate (assume compression unless there is a specific reason not to), the mix shift toward recurring revenue, and the probability-weighted impact of an adverse regulatory pricing decision. Run the regulatory scenario as an explicit branch, not as a discount-rate bump.

3. Sum-of-the-parts where the mix is genuinely different. Do not apply one multiple to an entity that contains a transaction business, an index/data business and a technology-services business. Indicative relative levels: index and market-data franchises command the highest multiples (recurring, pricing power, near-zero marginal cost), clearing and depository annuities next, cash-equity transaction revenue lowest. Rating agency groups with large analytics/research arms must be split the same way — the non-ratings segment often has different growth, different margins and different competitive dynamics from the ratings franchise.

4. Indicative multiple bands (vary with rates, cycle and jurisdiction; use as orientation only, never as a target):

• Global vertically-integrated exchanges and index/data-heavy groups: high-teens to low-30s P/E, 12–22x EV/EBITDA.
• Card networks: premium multiples, typically 25–35x earnings, justified by ~50%+ net margins, near-zero capital intensity and duopoly economics — the question is never whether they are expensive but whether the take rate survives.
• Merchant acquirers and processors: structurally de-rated to low-to-mid teens P/E and single-digit-to-low-teens EV/EBITDA on disintermediation and pricing fears. Use net-revenue multiples only.
• Rating agencies: 25–35x, reflecting oligopoly and a regulatory mandate to be rated, discounted for issuance cyclicality.
• Indian MIIs: premium to global peers is common, driven by structural growth in demat accounts and derivative participation. The premium is only defensible if you have explicitly stress-tested regulatory intervention in derivatives.

5. Useful cross-checks. Market cap ÷ annual net revenue; EV ÷ annual contracts or ÷ annual payment volume (compare like-for-like products only); implied terminal take rate embedded in the current price — solve for it and ask whether a regulator would tolerate it.

What NOT to use: P/B (book value is either negative from buybacks or inflated by client funds); EV/Sales on any gross-revenue reporter; ROE for a company with a bought-back balance sheet; consolidated ROCE without excluding client funds and SGF; EV computed by netting restricted cash; EV/EBITDA compared across gross and net reporters; and peer-average multiples across sub-sectors with different recurring-revenue shares.

Peer set construction

Never put these in one table. They have different pricing power, different regulators and different failure modes.

1. Vertically integrated derivative exchanges with captive clearing. Own the matching engine and the CCP, so open interest and margin offsets are locked in. Near-monopoly share, stable take rates, highest multiples. Compare only with each other.

2. Cash-equity and competitive-venue exchanges. Where clearing is unbundled or interoperable and best-execution rules permit competing venues (US Reg NMS, EU MiFID/MiFIR), order flow is genuinely contestable and take rates fall every year. Off-exchange internalisation, dark venues and payment-for-order-flow arrangements take share of the addressable pool. Do not benchmark these against a vertically integrated derivatives franchise.

3. Depositories, registrars and custody-adjacent infrastructure. Annuity economics — annual issuer charges, per-account and per-debit fees, corporate-action and e-voting fees, KYC infrastructure. Far less volume-sensitive, more account-growth-sensitive, and priced by the regulator. Different beta from exchanges even in the same country.

4. Clearing corporations and CCPs. Where separately listed or separately disclosed, treat as a distinct sub-sector: earnings are margin-float-driven and rate-sensitive, and they carry tail default risk no other sub-sector has.

5. Index, market-data and analytics providers. Highest-quality revenue in the sector: recurring licence fees, AUM-linked index royalties, contractual escalators. Comparable to information-services businesses, not to trading venues.

6. Credit rating agencies. Issuance-cyclical, regulated, oligopolistic, issuer-pays. Compare only with other CRAs, and split out non-ratings analytics revenue before comparing margins.

7. Card networks. Four-party scheme operators that do not earn interchange (interchange goes to the issuer) and do not take credit risk. Revenue is service assessments, data-processing fees, cross-border fees and value-added services, less incentives.

8. Issuers, acquirers, PSPs, payment aggregators and gateways. These take credit risk (issuers), merchant/chargeback risk (acquirers), or neither but with thin economics (gateways). Their take rate, capital requirements and regulatory perimeter are entirely different from a network's. Never place a network and an acquirer in the same peer table, and never compare a gross-revenue reporter with a net-revenue reporter without restating.

Cross-cutting splits that must be respected:

• Price-regulated vs price-free markets. A market with capped interchange or a zero-MDR mandate is a structurally different business from an unregulated one, regardless of similar volumes.
• Monopoly/duopoly by licence vs contestable markets. The presence or absence of a realistic second venue determines whether take-rate compression is a certainty or an option.
• Gross vs net revenue reporting, restated before any comparison.
• Currency, tax and transaction-tax regimes (STT/CTT in India, Section 31 fees in the US, financial transaction taxes in parts of Europe) change both volumes and reported revenue.

Sector-specific red flags

Volume and revenue quality

• Notional turnover quoted instead of premium turnover for options. Notional grows explosively as strikes proliferate and expiries shorten while the economically relevant base (premium, or contracts) grows far less. Insist on the premium/contract-count series.
• Volume concentrated in ultra-short-dated, retail-driven products. High revenue per unit of underlying economic activity is precisely what attracts regulatory intervention. Treat it as high-quality cash flow with a low-quality life expectancy.
• Take rate rising for reasons management cannot decompose. A rise from a one-off fee-slab change, a tax pass-through or a mix shift is not pricing power.
• Revenue growth accompanied by falling take rate and rising incentives — the company is buying volume.
• Treasury income rising as a share of PBT while core operating profit is flat. Very common in Indian MIIs during high-rate periods and the most frequent way core stagnation is disguised.

Risk and capital

• Default fund or margin model calibrated to a stale volatility regime. Look for regulator-mandated top-ups, changes in stress scenarios, or an unusually low SGF relative to peak open interest.
• Thin or falling skin-in-the-game relative to the mutualised default fund — it misaligns the CCP's incentives on margin adequacy.
• Rising concentration of open interest or margin in a single member or client group. The historical CCP losses all began this way.
• Procyclical margin models that force large intra-crisis calls: they protect the CCP but generate member defaults, litigation and regulatory backlash.
• Restricted funds presented as cash in investor presentations, or netted in a "net cash" figure.

Accounting and disclosure

• Gross-vs-net revenue reclassification in a payment company, or a change in the principal-vs-agent judgement. It resets the entire revenue and margin history — rebuild a like-for-like series before quoting any growth rate.
• Aggressive capitalisation of internally developed software, with capitalised development rising as a share of technology spend. In a business whose only real investment is technology, this directly manufactures EBIT.
• Roll-up processors' "adjusted EBITDA" excluding perpetual integration, restructuring and share-based costs; organic-growth definitions that change year to year; acquisition accounting that parks costs in purchase-price allocation.
• Segment re-definition that moves data, connectivity or listing revenue between buckets, obscuring the transaction/recurring split.

Governance and regulatory

• Conflicts of interest at the infrastructure institution: a market-infrastructure entity competing with its own members, preferential access to data or infrastructure for some participants, or opaque colocation and tick-by-tick data allocation. This has been the source of the most damaging regulatory actions against exchanges globally and in India, with multi-year overhangs, disgorgement and blocked listings.
• Repeated technical glitches and the associated penalty framework — a leading indicator of underinvestment and of regulatory patience running out.
• CRA-specific: rating shopping, unusually fast fee growth in one structured product class, high analyst attrition, ratings that lag market-implied spreads by long periods, or a large share of revenue from a small group of frequent issuers.
• Regulatory consultation papers on fee caps, product design, expiry structure, position limits, interchange or MDR. These are public and typically precede the earnings impact by 3–12 months. A model that has not read the live consultation papers is stale.

Cycle and structural context

Cyclicality is volatility-linked, not GDP-linked — and the direction differs by sub-sector. Volatility spikes drive trading volumes, margin balances and clearing revenue up while simultaneously shutting the IPO window (listing fees), freezing bond issuance (rating agency revenue) and suppressing discretionary spending (card volumes). A single macro shock therefore moves the sub-sectors in opposite directions. Do not model "market infrastructure" as one cycle.

Rates are a first-order earnings driver, twice over. Higher rates raise float income on margin and settlement balances (a pure windfall to CCPs, depositories and acquirers) and simultaneously compress the multiple applied to fee streams. When rates fall, that float income disappears — check how much of the last three years' earnings growth was rate-driven before extrapolating.

Retail participation cycles. Derivative and cash-equity volumes in emerging markets are heavily retail-driven and mean-reverting after drawdowns, with a lag. A multi-year bull market that has doubled demat accounts and trebled option volumes is not a permanent base rate.

Structural forces to price explicitly:

• Regulatory price-setting is the dominant risk in this sector. Interchange caps (EU IFR, US debit-interchange regulation and routing mandates), mandated zero-MDR regimes, exchange transaction-fee caps and "true-to-label" pricing rules have each permanently reset take rates in their markets. A licence that bars entry usually comes with a regulator that sets price; the moat and the risk are the same object.
• Account-to-account real-time payments (UPI, Pix, FedNow, European instant payments) bypass card rails entirely and, where mandated free, remove the toll rather than shrink it. This is the single largest secular threat to card-network and acquirer economics, and it is furthest advanced in India.
• Disintermediation of the venue. Off-exchange internalisation, dark pools, systematic internalisers, payment for order flow, all-to-all bond platforms and direct listings each take a slice of the addressable pool without competing on the exchange's own terms.
• Settlement compression (T+1, moves toward T+0 and same-day settlement, and tokenised/DLT settlement experiments) reduces the margin and float balances that generate income for CCPs, and reduces the risk that justifies their existence. Positive for systemic safety, negative for float earnings.
• Passive investing growth is a durable tailwind for index franchises (AUM-linked royalties) and for ETF-linked trading, and a headwind for high-turnover active flow.
• Stablecoins and crypto rails are a genuine medium-term threat to cross-border payment economics — the highest-margin line in the network P&L — and a source of new listed/traded products for exchanges.
• Alternative credit assessment (private credit not requiring public ratings, bank internal models, data-driven scoring) slowly erodes the CRA mandate at the margin, while regulatory requirements to be rated keep the floor in place.

Where the moat actually is. Not the technology — matching engines are commodities. It is (a) the licence, (b) pooled liquidity and open interest with their margin offsets, (c) the network's two-sided installed base of issuers and acceptance points, and (d) contractual, embedded recurring revenue in data and index licensing. Rank any company here by which of those four it actually owns.

India vs global notes

Regulatory architecture (India). SEBI regulates exchanges, clearing corporations and depositories as Market Infrastructure Institutions under the SECC and Depositories regulations; RBI regulates payment systems, payment aggregators and card networks' domestic operations; NPCI operates UPI, RuPay, IMPS and NACH as a not-for-profit. Distinctive features to check for any Indian MII:

• Ownership caps and governance: shareholding in an MII is capped for most holders (a low single-digit percentage for the general category, with a higher ceiling for specified financial institutions), so there is no promoter in the conventional sense. Do not run promoter-holding or pledge analysis; instead read the board composition, the Public Interest Directors, and the SEBI-approved appointment of key management. Governance risk here is regulatory-relationship risk, not promoter risk.
• Mandated profit transfers: exchanges and clearing corporations must contribute a prescribed share of profits to the clearing corporation's Core Settlement Guarantee Fund, and exchanges maintain Investor Protection Funds. Verify the current prescribed percentages from the latest SEBI circular — they change — and treat those transfers as a permanent reduction in distributable earnings and the balances as restricted.
• Price regulation: SEBI's "true-to-label" charges framework requires uniform, non-slab transaction charges, which removed the volume-rebate structure exchanges previously used. SEBI's regulatory turnover fee on options is levied on premium rather than notional. Both are examples of regulatory pricing decisions with immediate revenue impact.
• Product intervention: SEBI has periodically tightened index-derivative rules — larger contract sizes, fewer weekly expiries per exchange, upfront premium collection, removal of calendar-spread benefits on expiry day, higher expiry-day margins. These measures reduced retail derivative turnover materially and quickly. Any Indian exchange model must carry an explicit product-intervention scenario.
• Technical glitch framework: SEBI prescribes uptime, disaster-recovery switchover timelines and financial disincentives for outages at MIIs. There is no direct global equivalent of comparable specificity.

Payments (India). MDR on UPI person-to-merchant and on RuPay debit is effectively zero by statute for most merchant categories, so the domestic debit and A2A rails are not monetisable at the network layer — value accrues to banks, to the merchant, and to whoever monetises the customer relationship, not to a toll collector. Card economics in India are therefore essentially a credit-card story. Payment aggregators require RBI authorisation with prescribed net-worth thresholds; card-on-file tokenisation is mandated; cross-border PA activity has its own licence. Contrast this with the US and EU, where interchange (capped in the EU, partially regulated for debit in the US) remains a large monetised pool. Never apply developed-market card take rates to India.

Depositories and demat (India). A duopoly with volume driven by new demat account additions, transaction debits, corporate actions, e-voting and KYC infrastructure — an annuity profile that behaves quite differently from exchange turnover. Monthly account and transaction data are public.

Rating agencies (India). SEBI-registered CRAs; a large share of revenue comes from bank-loan ratings driven by Basel capital rules, in addition to bond issuance — a revenue pool with no direct US analogue. Several Indian CRAs earn a majority of consolidated revenue from non-ratings research, analytics and global delivery centres; those segments must be valued separately, since they are outsourced-services businesses with different margins, currency exposure and competitive dynamics.

Accounting and filings. India: Ind AS, standalone and consolidated; crore/lakh — normalise before cross-border comparison. Clearing corporations and depositories are usually subsidiaries, so read the subsidiary financials in the annual report, not just the consolidated statements, to see margin balances, SGF and restricted funds. CARO reporting, the related-party note and the contingent-liability note (regulatory penalties, disgorgement, litigation with members) carry most of the risk signal. Concalls and monthly turnover releases are the highest-frequency data source. Global: 10-K/10-Q on EDGAR under US GAAP or IFRS; exchanges disclose net revenue after transaction-based expenses (use that line, not gross); networks disclose gross revenue, incentives and net revenue plus PV, cross-border volume and processed transactions in the same tables; CCPs publish CPMI-IOSCO PFMI quantitative disclosures quarterly — margin, default fund, stress-test coverage and member concentration — which is the single best public source for clearing risk and has no equally structured Indian equivalent.

Market data and colocation. In the US and EU, market-data and connectivity fees are a large, high-margin and politically contested revenue line, subject to regulatory fee reviews, consolidated-tape initiatives and litigation by users. In India, data and colocation revenue is a smaller share and colocation has itself been the subject of significant regulatory action. Do not assume the developed-market data-revenue mix is available to an Indian exchange.

Checklist

• Restate revenue to a net basis (after interchange, scheme fees, incentives, liquidity rebates and regulatory pass-throughs) before computing any growth rate or margin.
• Strip client margin, settlement balances, core SGF, IPF and restricted regulatory capital out of assets, liabilities, cash, EV and capital employed.
• Decompose net revenue growth into volume, take rate and mix, for each major product, over 8+ quarters.
• Use premium/contract-count volumes for options, never notional turnover.
• Compute the recurring (data, listing, index, annual issuer/custody, surveillance, VAS) share of net revenue and its trend.
• Compute incremental flow-through margin (ΔEBIT/Δnet revenue) both in the up-cycle and in the last downturn.
• Compare core opex growth with volume growth over 3–5 years to test whether the cost base is genuinely fixed.
• Separate treasury/float income from core operating profit; compute core P/E ex-treasury and exclude the unrestricted corpus from market cap.
• Quantify rate sensitivity of float income and re-run earnings at a normalised rate.
• Measure product concentration and model an explicit regulatory-intervention scenario for the top product.
• For CCPs: read the PFMI quantitative disclosures — default fund vs Cover-1/Cover-2, skin-in-the-game, assessment powers, largest-member concentration.
• Check regulatory net-worth headroom and any mandated profit transfers to the SGF or protection funds.
• For networks/acquirers: track client incentives as % of gross revenue and cross-border share of net revenue.
• Check member, client and issuer concentration on both the revenue and the risk side.
• Review 5 years of outages, glitch penalties, regulatory orders, disgorgement and litigation with members.
• Read live regulatory consultation papers on fees, MDR/interchange, product design and expiry structure before finalising any forecast.
• Test capitalised software as a share of technology spend, and adjusted-EBITDA add-backs for roll-up processors.
• Compute FCF excluding movements in client/settlement/margin balances; check conversion against PAT over 3–5 years.
• Value by SOTP where transaction, data/index and technology-services revenue coexist; never one blended multiple.
• Confirm the peer set matches on vertical integration, price regulation, contestability and gross-vs-net reporting — not on market cap or index membership.
• State what the current price implies for terminal take rate, and ask whether a regulator would allow it.