Methodology

How YieldIQ values a stock

Open methodology for the DCF, Prism scoring, and verdict bands behind every analysis.

01 — Valuation

The DCF

The core fair-value engine is a discounted-cash-flow model. Free cash flow is taken from the data pipeline — operating cash flow net of capex, cleaned for one-offs where disclosure permits. The discount rate is a sector-aware WACC: the Indian 10-year G-Sec serves as the risk-free rate, sector equity-risk premia and betas come from models/industry_wacc.py, and the cost of debt reflects the company’s own interest burden where reliable.

Terminal growth is sector-specific rather than a single blanket number. Mature FMCG and utilities are modelled at low single digits; IT services and select consumer names sit higher; cyclicals are held close to long-run nominal GDP. The intent is to avoid the single worst failure mode of generic DCFs — one terminal assumption papered across every industry.

Every valuation is published in three scenarios: bear, base, and bull. The scenarios flex growth, margin, and reinvestment jointly rather than one input at a time, so the spread reflects plausible end-states rather than sensitivity theatre.

Alongside the forward DCF, we publish a reverse DCF that solves for the growth rate implied by the current market price. When the implied number is higher than anything the business has ever delivered, the reverse DCF says so plainly.

All DCF outputs are model estimates. They are not price targets and nothing on this page should be read as a recommendation to transact.

02 — Routing

When DCF isn't enough — specialized models by sector

A single DCF cannot value every business shape. A bank does not have free cash flow in the normal sense; a REIT is a pass-through; a regulated utility earns a tariff-capped return that compounds in DCF math into something the regulator has never permitted. YieldIQ routes each ticker to a sector-appropriate engine, then anchors the answer against peer multiples and a reverse-DCF cross-check.

Multiples-based fair value

Peer-relative PE / PB / EV-EBITDA — runs on every ticker as a parallel signal to the DCF.

Routes: All tickers, including those already routed to a specialized engine.
Why DCF alone is inadequate: DCF outputs drift when terminal-growth and WACC assumptions move by a hundred basis points. Peer-relative multiples anchor the DCF against the price that comparable businesses actually trade at, surfacing disagreements rather than hiding them.
Key inputs: Sector-cohort PE / PB / EV-EBITDA percentiles
Bucketed peer set (large / mid / small-cap)
Cohort median + sigmoid-smoothed dispersion
Subject ticker's own three-year ratio history

Bank residual-income

P/BV times adjusted book value with a residual-income overlay.

Routes: Private and PSU banks (HDFCBANK, ICICIBANK, SBIN), large NBFCs (BAJFINANCE), life insurers (HDFCLIFE).
Why DCF alone is inadequate: Banks do not produce free cash flow in the sense an industrial does. Their economic engine is net interest margin earned on book equity, not operating cash flow net of capex. A DCF on a bank treats deposits as a financing item that funds operations, which inverts the actual business model. Book value plus the spread between return on equity and cost of equity is the textbook frame.
Key inputs: Return on equity, current and three-year trailing
Cost of equity from a bank-specific WACC sheet
Net interest margin and CASA mix
Provision coverage ratio and gross NPA trajectory
Adjusted book value (book equity net of intangibles + visible stress)

REIT net-asset-value with DPU yield gap

Underlying property NAV plus distribution-per-unit yield versus the 10-year G-Sec.

Routes: Listed REITs and InvITs (EMBASSY, MINDSPACE, BROOKFIELD, IRBINVIT).
Why DCF alone is inadequate: A REIT is a regulated pass-through that distributes 90 percent of its cash flow. A pure DCF on the distribution stream misses the property-level appreciation that drives unit value, and a pure NAV ignores the income premium. Both halves matter, weighted to the way the unit actually clears in the market.
Key inputs: Underlying property NAV from the sponsor's filings
Trailing-twelve-month DPU
10-year G-Sec yield as the risk-free anchor
Sector-cohort yield spread (REIT DPU yield minus G-Sec)
Occupancy and lease-roll schedule for the next three years

Regulated utility — RAB times allowed ROE

Regulated Asset Base multiplied by the regulator's allowed return on equity.

Routes: Power transmission and generation utilities (POWERGRID, NTPC, transmission DISCOMs).
Why DCF alone is inadequate: A regulated utility earns a tariff-capped return on a defined asset base. The regulator (CERC, SERCs) sets the allowed return on equity in five-year tariff orders. A DCF that compounds historical FCF growth ignores the cap and produces a number the regulator has never permitted. RAB times allowed-ROE is the frame the company itself reports against.
Key inputs: Regulated Asset Base from the latest tariff order
Allowed return on equity in the current control period
Capex run-rate that flows into next-period RAB
Under-recoveries and regulatory assets on the balance sheet
Tariff-petition outcomes from prior cycles

Platform / recent-IPO sector relative

Sector-relative valuation for businesses with under 36 months of public data.

Routes: Recent IPOs and platform businesses (ZOMATO, NYKAA, PAYTM, POLICYBZR, MANKIND in its first year).
Why DCF alone is inadequate: A DCF needs at least three to five years of revenue, margin, and reinvestment history to calibrate baseline assumptions. A recently-listed platform has neither the history nor a stable cohort to lean on. Forcing a DCF here produces a spuriously precise number from noisy inputs. Sector-relative valuation flags the price as a band rather than a point and waits for the data to mature.
Key inputs: Closest comparable cohort (Indian + global platform peers)
Trailing GMV / contribution-margin trajectory
Cohort EV / sales and EV / contribution-margin multiples
Months of public reporting (gates the model out under 6 months)

Tier-2 cohort

Cohort-relative valuation for mid-caps where direct DCF inputs are noisy.

Routes: Hand-curated tier-2 list across cyclicals, capital goods, and specialty chemicals.
Why DCF alone is inadequate: Small and mid-cap DCFs are dominated by single-year cash-flow swings — one capex year or one inventory cycle can flip the answer. Cohort-relative valuation smooths these inputs by anchoring against five to ten name-level peers in the same sub-sector. The output is wider but more honest about its own uncertainty.
Key inputs: Sub-sector cohort of 5-10 hand-curated peers
Median PE, PB, EV-EBITDA across the cohort
Subject's three-year normalized margin (median-of-window)
Confidence haircut applied when cohort dispersion is wide

Reverse DCF

Back-solves the growth rate the current market price already implies.

Routes: All tickers — surfaces alongside the forward DCF as a transparency signal.
Why DCF alone is inadequate: The forward DCF asks 'what is this worth?' The reverse DCF asks 'what does buying at the current price assume?' When the implied growth rate is higher than anything the business has delivered, the reverse DCF says so plainly. It reframes valuation as a check on market-implied expectations rather than a single answer.
Key inputs: Current market price
Last reported FCF (or normalized FCF for cyclicals)
Discount rate (same sector WACC as the forward DCF)
Terminal growth assumption (same as the forward DCF)
Implied growth rate solved over the explicit forecast window

Story DCF / Day-89 backtest

Historical revenue and margin path plus what-if scenarios for the YIQ50 backtest set.

Routes: YIQ50 backtest universe — the harness that produced the public Day-89 results.
Why DCF alone is inadequate: A single headline DCF hides which assumption is doing the work. The story DCF lets a user flex revenue growth, margin trajectory, or reinvestment rate one at a time and watch the fair value move. The Day-89 backtest panel shows how each story would have played out historically, so users can challenge the central case with their own narrative.
Key inputs: Historical revenue and margin path (10-year window)
User-flexed growth, margin, and reinvestment scenarios
Sector-specific terminal growth ceiling
Backtest accuracy versus realized prices on the YIQ50 set

Composite intrinsic value

Weighted blend of DCF, multiples-based fair value, and Wall Street consensus.

Routes: All tickers — surfaces as a parallel column next to the DCF-only fair value.
Why DCF alone is inadequate: A pure DCF can run high relative to peer multiples and analyst consensus on names where terminal-growth assumptions matter most — large private banks and high-quality compounders. The composite blends three independent signals so no single methodology can dominate the headline number. Each input remains visible so users can see where the disagreement sits.
Key inputs: Forward DCF fair value with three-scenario distribution
Multiples-based fair value (sector-cohort anchored)
Wall Street consensus target where coverage exists
Sector-specific weighting between the three inputs

Holdco sum-of-the-parts

SOTP for pure holding companies whose value is a basket of stakes — currently shipping as a DCF-only stop-gap.

Routes: Pure holdcos (BAJAJHLDNG and similar). Subsidiaries with operating businesses route to their own sector engine.
Why DCF alone is inadequate: A pure holding company's value is the sum of its stakes in listed and unlisted subsidiaries, not a discounted-cash-flow on its own thin parent-entity cash flows. DCF on the parent entity misses the subsidiary value entirely. A proper SOTP marks each stake to its underlying fair value and applies a holdco discount for liquidity, control, and tax frictions.
Key inputs: Market value of each listed-subsidiary stake
DCF fair value of each unlisted-subsidiary stake
Holdco discount (typically 20-40 percent in the Indian market)
Parent-entity net debt and treasury holdings

Honest caveats

Composite IV closes a real DCF-only gap. On HDFCBANK, the DCF-only fair value ran materially above the peer-multiples and AlphaSpread numbers. The composite engine blends DCF, multiples, and consensus so that bias collapses into a band the three methods actually agree on. We publish both columns side by side rather than hiding the DCF-only output.

Holdco SOTP is a stop-gap. The current holdco route is a DCF-only fallback that under-counts subsidiary value. A real sum-of-the-parts engine is in the roadmap; until it ships, BAJAJHLDNG and similar pure holdcos carry an under-review caveat on the analysis page.

Three sector engines are still on the roadmap. Pharma pipeline-adjusted DCF (probability-weighted molecule cash flows), Insurance embedded-value plus value-of-new-business, and Telecom ARPU-driven models are not yet shipped. Until they are, the relevant tickers route through the general DCF with a sector-cohort multiples cross-check, and any large gap between the two surfaces as an under-review band.

03 — Scoring

The 6-pillar Prism

The Prism is a decomposition of business quality and valuation into six independently scored pillars. Each pillar is scored 0–10, the six are composited to a /10, and the composite is rendered as an A–F grade on a /100 scale for quick scanning.

Pulse

Short-horizon signal from recent price action and sentiment — momentum, volatility regime, and revision direction. Informative, not decisive; it sits alongside the slower-moving pillars rather than overriding them.

Quality

Return on capital employed, return on equity, operating and net margins, and the stability of reported earnings across cycles. High scores require durability, not just a good last twelve months.

Moat

Persistence of gross margin, evidence of pricing power through input-cost shocks, and the durability of return on capital versus peers. A high Moat score means the excess returns show up year after year, not as a one-period spike.

Safety

Balance-sheet resilience — leverage ratios, interest coverage, and an Altman-Z-style composite adapted for Indian reporting. Financials use bank-appropriate substitutes (capital adequacy, NPA ratios) where the standard formula does not apply.

Growth

Revenue and earnings CAGR across both 3-year and 5-year windows, blended to reward consistency over one-off spikes. Growth is reported in isolation; a high Growth score does not imply a high Value score.

Value

The DCF margin of safety combined with sigmoid-smoothed relative multiples against sector peers. Smoothing prevents extreme multiples from collapsing the score, and the MoS weight dominates so that the label tracks the model rather than the screen.

04 — Labels

Verdict bands

Verdicts are descriptive, not imperative. They describe where the current price sits relative to the modelled fair-value distribution. They do not tell anyone to buy or sell.

Deep ValuePrice materially below the bear-case fair value. The market is pricing in an outcome worse than our most pessimistic scenario.
Below Fair ValuePrice below the base-case fair value but above the bear. A meaningful margin of safety on the central estimate.
Fair Value RegionPrice within the normal dispersion of the base case. No pricing edge either way on the modelled assumptions.
Above Fair ValuePrice above the base case but below the bull. The market is implying a better outcome than our central estimate.
Well Above Fair ValuePrice above the bull-case fair value. The market is pricing in an outcome better than our most optimistic scenario.
Under ReviewInsufficient data to assign a band. We apply this explicitly rather than guess — thin IPO history, unit-change ambiguity in filings, or a failed validator will all land here.

05 — Inputs

Data sources

Live quotes come from a supplementary global market data API for real-time and delayed prices, with a secondary feed for analyst estimates and corporate-event metadata. Quotes are cross-validated against NSE bhavcopy daily close.

yfinance is still used for parts of the fundamentals pipeline. It is a pragmatic dependency, not an ideal one. We mitigate the risk with an aggressive own cache, a process-wide circuit breaker that trips on rate-limit or error bursts, and validators that reject unit-jump corruption before it reaches the model.

Managed Postgres is the canonical store for cleaned financials, computed fair values, and Prism scores. Everything on the site reads through this layer.

In-process analytical engine on Parquet backs the ten-year history surfaces — price panels and the aggregated fundamental history used for CAGR and stability calculations. It is fast enough for ad-hoc analytical queries and immutable enough to rely on.

NSE/BSE XBRL filings are progressively replacing the yfinance fallback for fields that are reliably tagged. The rollout is line-item-by-line-item rather than a cutover, because any given filing’s quality varies by filer.

06 — Honesty

Known limitations

Recent IPOs with fewer than three years of post-listing financials are too thin for the Growth and Moat pillars to be trustworthy. These names are surfaced under Under Review rather than scored.

Unit-change events in filings (lakhs versus crores, thousands versus millions) are handled on a best-effort basis. The validator suite catches the common cases; the residual risk is real and we disclose it.

Peer selection uses a three-band market-cap bucketing — Large-cap above ₹50,000 Cr, Mid-cap between ₹10,000 Cr and ₹50,000 Cr, Small-cap below ₹10,000 Cr. Bucket boundaries are deliberate and infrequently moved, which means a stock right at a threshold can flip buckets on valuation days without a real change in its business.

Sector models are shared across their sector, not bespoke to each ticker. A bank is modelled as a bank; an IT services company as IT services. The approach is intentionally generic: bespoke per-ticker tuning is what produces post-hoc-justified valuations, which is exactly what we want the methodology to resist.

07 — AI

AI summaries — full disclosure

The single-paragraph summary at the top of each analysis page is generated by a large language model. The numbers below the summary are authoritative. The AI is summarising, not deciding. If they ever disagree, the numbers are right.

Model: Groq llama-3.3-70b-versatile
Temperature: 0.3 (low — deterministic-leaning)
Inputs: ticker, fair_value, current_price, MoS%, verdict, score, Piotroski/9, moat grade, revenue CAGR, key red flags
System prompt: “You are a SEBI-compliant analyst summarizing model output. Use ₹ for all values. No buy/sell/hold language. State the verdict band, the gap to fair value, and 2–3 driver metrics. Max 3 sentences.”
Regenerated: on each cache miss (typically every 24h or after a data change)

08 — Regulatory

SEBI posture

YieldIQ is not registered with the Securities and Exchange Board of India as an Investment Adviser or Research Analyst. Nothing on the site is investment advice, a recommendation, or a solicitation.

Verdicts are descriptive rather than imperative. Where data quality is insufficient, we apply an explicit Under Review label instead of forcing a call. Fair-value outputs are model estimates derived from publicly available inputs and disclosed assumptions; actual outcomes may differ materially.

Do your own research. Consult a SEBI-registered adviser before making investment decisions.

See pricing →About YieldIQ Terms Privacy