On this page
INVESTRA is a data-driven research dashboard covering 50 listed Indian BFSI companies - banks, NBFCs, housing finance, microfinance and small finance banks. Every score and signal is a transparent, reproducible function of reported financials. Nothing here is a forecast of price, a call to act, or a substitute for your own judgement.
What it is
- A quantitative screening lens on quality (ASTRA) and valuation (Gordon Growth).
- A way to read where a company sits versus its sub-sector on the same banded scales.
- Fully transparent math - bands, weights and formulas are documented on this page.
- Honest about gaps: any unavailable metric renders N/A, never a guess.
What it is not
- Not investment advice, a buy/sell/hold call, or a solicitation of any kind.
- Not a price target or a prediction of future returns.
- Not a substitute for an issuer's filings, a qualified advisor, or your own due diligence.
- Not a trading feed - scores use the workbook's valuation-date price and each company's latest reported quarter; live prices are shown for context only.
The dashboard describes the mathematics; the interpretation and the decision are entirely yours. Valuation words such as Undervalued or Richly Valued describe the model's output relative to its own justified value - they are not recommendations.
Every number on this site travels one fixed, scripted path. Nothing is typed into a page by hand, and no page computes its own metrics - each figure is computed once, upstream, and every page merely displays it. That is why the same metric can never disagree with itself between two pages.
How many times its book value (net worth per share) the market is willing to pay for the company today. It is the share price on the valuation date (from the workbook) divided by book value per share - a market fact, no assumptions. The live price on company pages is for display only and never changes this number. A bank at 2.5× is priced at two-and-a-half times the equity on its balance sheet.
The multiple a company deserves to trade at given the actual quality of its business - specifically how profitably it uses shareholder money, how risky that profit stream is, and how fast it can grow. It is an estimate of fair value, built from the company's own fundamentals rather than from its share price.
Why we need the justified multiple. A share price on its own says nothing about whether a company is cheap or expensive - a "high" price can be a bargain for a superb business, and a "low" price can be dear for a weak one. The justified price-to-book gives us an independent yardstick built from the fundamentals, so we can ask the only question that matters in research: is the market paying more, or less, than this business is actually worth? Comparing the two is what turns a price into a judgement.
The justified multiple comes from the Gordon Growth Model, the standard way to value a financial company. It links three fundamentals - the sustainable return on equity (the profit earned on every rupee of shareholder capital), the cost of equity (the return investors demand for the risk they take), and the long-run growth rate - into the multiple the company should fairly command:
The intuition is simple: a company that earns far more on its equity than investors demand should trade well above book value; one that barely covers its cost of capital should trade near book. INVESTRA uses a two-stage version - a higher-growth phase that fades to a sustainable terminal rate - and reads valuation against that steady-state justified price-to-book. The valuation gap is then just the distance between the two multiples, measured in turns of book value (a difference, not a percentage):
positive → the market pays more than fundamentals justify · negative → the market pays less (undervalued)
A company at +1.3 turns trades about 1.3× of book above what its fundamentals justify; one at −0.9 trades nearly a turn of book below. When the justified multiple works out to zero or negative - which happens only for loss-making companies, where the model breaks down - no meaningful fair value exists and it is shown honestly as N/A.
The justified multiple is built from today's reported fundamentals. The market,
however, pays for the future - it will happily price a high-quality compounder at 4-5× book
because it expects years of rising profitability and growth that current numbers do not yet show.
So the best franchises often trade above their justified value and read "Overvalued" or
"Richly Valued." That is the model behaving correctly, not an error.
Read the valuation tier for what it is: "is the price ahead of, or behind, the current
fundamentals?" - not "is this a good company?". Business quality is answered
separately by the ASTRA score below. A name can be excellent on quality and still expensive on price;
the two readings are meant to be used together.
Because the fair P/B is sensitive to its inputs, every company page also shows a range: the fair P/B recalculated with the cost of equity one percentage point higher and growth one point lower, and the reverse. The calculation lives in the workbook's "Fair Value Range" sheet, which first re-computes the model's own fair P/B (a check that passes for all 50 companies) and then shifts the two inputs. A market price inside the range is within the model's margin of error.
Tiers are a direct read of the valuation gap, in turns of book. Boundaries: gap below −0.20 → Undervalued · below +0.10 → Fair Value · below +0.50 → Slight Premium · below +1.50 → Overvalued · +1.50 and above → Richly Valued. They describe the mathematics, not a recommendation.
ASTRA is a 0-100 quality score built from reported numbers. Each number is placed on a scale for its type of lender - for example, a 1.5% bad-loan ratio is excellent for a microfinance lender but only average for a large private bank - and turned into a 0-100 score. Those scores are grouped into 10 factors (asset quality, capital, funding, profitability, growth and so on), and the factors are added up with weights that depend on the type of lender into the headline score.
Missing data is never treated as zero. If a company does not report a number, that input is left out and the remaining weights are scaled up to fill its place - so nobody is punished (or rewarded) for a gap, and no value is invented. Two guards keep the score honest: a company in severe credit stress (credit-quality factor below 25) is capped at 60, so deep loan problems can never rate as top quality; and profit growth from a tiny or loss-making base year is not trusted (next paragraph).
A profit that jumps from ₹5 Cr to ₹50 Cr shows "900% growth" - which says more about how bad the starting year was than about the business. Following the CFA approach of normalising earnings before measuring growth, ASTRA only trusts a profit-growth figure when the starting year was profitable and at least a quarter of the period's average profit. Otherwise it uses the growth of book value (a steadier measure of how the business compounds) and leaves momentum out. Companies where the guard is active are tagged "Low-base profit growth" on their company page.
The factors are weighted differently by lending type because the drivers of quality differ. State-owned and microfinance lenders lean harder on credit quality and capital; non-bank and housing financiers carry more weight on growth and funding. These weight sets are read straight from the scoring engine - the table renders them, it does not invent them.
The fastest way to trust a methodology is to watch it run once, in full. Below, one company's headline score and valuation read are rebuilt step by step from the engine's own live numbers: its 10 factor scores, its lender type's weight row from the table above, the weighted average that produces the headline score, the grade band it lands in, and finally the valuation gap. Reload with new data and every figure here updates itself.
We take each company's reported numbers - ten years of yearly results plus the latest quarter's loan quality and capital - from its own filings and investor presentations, into one Excel workbook.
ASTRA turns those numbers into a 0-100 quality score: how clean the loan book is, how well capitalised and funded the lender is, how profitable it is and how steadily it grows - each judged against lenders of the same type.
From its returns, risk and growth we work out the price-to-book a company deserves (its fair value), and compare it with the price-to-book the market actually pays.
Every score, formula, weight and source is on this page. Missing data shows as N/A - it is never guessed. Nothing here tells you to buy or sell.
Every metric shown anywhere on this site is defined below. Coverage is counted live from the loaded dataset - how many of the covered companies have the input available. Where an input is missing, the site shows N/A and the scoring renormalises around the gap; a blank is never treated as a zero.
| Metric | What it tells you | How it is measured | Source | Coverage |
|---|
"Reported" = taken from issuer filings and investor presentations via the verified workbook. "Derived" = computed from reported inputs by the build or the scoring engine - never estimated. Coverage is counted from the live dataset at page load.
Every model rests on assumptions. These are INVESTRA's, stated plainly, with an honest note on how sensitive each output is to them:
- Gordon Growth inputs. The justified price-to-book is built from three inputs: the sustainable return on equity, the cost of equity, and the terminal growth rate. Because growth is subtracted from the cost of equity in the denominator, the multiple is most sensitive exactly when those two are close together - a small change in either input can move the justified multiple by a large fraction. That is why the valuation gap is read in broad tiers, never as a precise fair price.
- Expected return in the portfolio tool (VISHLESHAN). It is the return today's price implies if the business keeps earning its 5-year ROE: book growth plus the cash yield on the price, g + (ROE - g) / P/B - the same fair-value formula solved the other way round. No change in the multiple is assumed in the base case; a move toward fair value is shown only as a scenario (half the gap closing every three years). The outcome band uses portfolio beta times an assumed 15% market volatility and is illustrative.
- Momentum is an annual proxy. Momentum compares the latest year's profit growth with the three-year trend. It is an annual trajectory read, not a quarter-by-quarter momentum signal.
- Two quarters of asset quality. Loan-quality, capital and funding metrics are each company's latest reported quarter (mostly Q1FY27). For gross NPA, net NPA and coverage the workbook also keeps the previous reported quarter, so the site can show the direction of travel. A move smaller than 0.05 percentage points is shown as "stable"; coverage is compared on the same definition in both quarters (1 - net NPA / gross NPA).
An outlier is a lender whose earnings are negative or so close to zero that ratios built on them (return on equity, P/E, profit growth, fair value) stop meaning anything.
A lender is an outlier if any of these five tests is true. The tests read only the Excel export (never live prices), and they change no ASTRA score, grade or rank.
| Test | What it checks | Line |
|---|
Companies that fail at least one test today:
| Company | Tests hit | Why |
|---|
Closest to the line (not outliers)
These lenders pass all five tests but sit near a line, so a weak year could flag them: 5-year ROE under 6%, P/E above 50x, fair P/B under 0.10x, or a loss in the year before last. They stay in every chart and average.
The rule is deliberately one-sided to stay honest. Outliers are left out of every chart, average and model portfolio (so visuals and averages reflect the comparable core), but they stay present in every table, tagged Outlier with the reason, so nothing is hidden. A lender that passes all five tests at a later refresh returns to the main universe automatically.
Company pages carry risk flags. Each one is computed from reported metrics against a fixed threshold - never an editorial judgement. This is the complete catalogue; a flag whose input a company has not reported is simply not raised, in either direction.
| Flag | Raised when | Banks | SFBs | MFIs | HFCs | NBFCs | Severity |
|---|
Warning flags mark stress that is already measurable; positive flags mark cushions already in place. Each type of lender has its own thresholds (the columns above, read straight from the scoring engine): banks report a narrow early-warning bucket (SMA-2, 61-90 days overdue), NBFCs and HFCs report Stage 2 (30-90 days) and microfinance lenders portfolio-at-risk, so each is judged on its own scale. Secured lenders - housing and other collateral-backed NBFCs - recover much of a bad loan from the collateral and normally hold lower provision coverage, so their coverage floor is lower. The same scales are used inside the ASTRA credit-quality and safety factors. A stress-pool figure reported as an amount (rupees) instead of a ratio is left blank rather than misread.
A research framework is only as trustworthy as the process behind it. These four commitments govern everything on this site.
Deterministic scoring
The same inputs always produce the same scores. There is no analyst discretion anywhere in the pipeline - every band, weight and threshold is fixed, and every one of them is published on this page. Nothing is nudged, smoothed by hand, or overridden case by case.
Corrections flow one way
When a source figure is corrected upstream, the fix enters the workbook, the build regenerates the data file, and every page recomputes from it. There is no way to hand-edit a single displayed number - a correction either changes the underlying data for everyone, or it does not happen at all.
A fixed, objective outlier rule
Outliers are decided by a fixed, objective five-test rule (see Outliers above), re-run on every data refresh. They are excluded from every chart and aggregate so the visuals reflect the comparable core - but they stay present and clearly tagged in every table. The rule is the same everywhere on the site and is never applied selectively to flatter a result.
Independence
The platform holds no positions in the companies it covers, sells no advisory service, and receives no compensation from any issuer. Nothing on the site is a recommendation, a buy or sell call, or a solicitation of any kind - it is research, and the decision is always yours.
One source of truth. Every figure on this site is generated from a single Excel workbook. Inside that workbook a dedicated macro sheet carries the manually-maintained macro figures - the repo rate, system asset quality and capital, growth and inflation - each with its own as-of date and named source, so the macro backdrop is versioned exactly like the company data.
One command to refresh. npm run update recalculates the workbook in
Excel, exports the data file, checks it (50 companies, 7 segments, no broken values) and publishes it.
Every page reads that one file, so all pages change together - and an open page checks for a new
file every 15 minutes (only while it is in view) and reloads itself when the numbers have changed.
Live prices are display only. Company pages show a live quote, an intraday chart and a live 52-week bar from Yahoo Finance, fetched through the site's own Cloudflare function and labelled LIVE. These never touch the model: every score, fair value and ranking uses the valuation-date price in the workbook.
"As of" stamps. Where a macro figure carries its own as-of date and source, that stamp is shown beside it - for example under the repo-rate and system-health tiles on the Macro Context page - so you always know how current each number is.
Per-company financials come from issuer filings and investor presentations, compiled in the verified valuation workbook; each company's latest-quarter row in the Metrics sheet carries a link to its source document. The macro and system-health figures come from named, dated RBI and IMF documents kept in the workbook's MACRO sheet - each figure traces to a source:
A transparent framework names its own boundaries. The following limits apply to everything on this site:
- Two-quarter view, not a long series. Asset-quality metrics hold the latest and the previous reported quarter - enough to show direction, not a multi-year quarterly trend. Momentum in ASTRA still uses yearly profit growth.
- Different disclosures. Lenders do not all report the same way (for example SMA-2 for banks, Stage 2 or 30+ days overdue for NBFCs). Each figure carries its own label, but cross-type comparisons of these ratios need care.
- Gordon Growth sensitivity. Justified P/B is sensitive to its inputs - small changes in assumed sustainable ROE, cost of equity or terminal growth move the multiple. The gap is an anchor for discussion, not a precise fair price.
- N/A is honest. Any unavailable metric renders N/A and is excluded from its score via renormalisation. A blank is never silently treated as a zero, and no value is ever invented.
- Annual history is FY-end. Trend charts use FY2017-FY2026 annual data; the latest fiscal year may be partial or pending for some companies and will read N/A where not yet reported.