V-SBM: Viswanathan–Wasserstein Schrödinger-Bridge and Rough-Signature Valuation Model

Interactive Tool Illustrative Model CA V. Viswanathan

V-SBM is an illustrative scoring model developed by CA V. Viswanathan. It tests the valuation claimed in a pitch deck against the audited GST and bank record, and reduces it for the gap between the two, for drift in unit economics and for cash-flow volatility, with a small credit for strong retention. It draws on optimal transport, information geometry and rough-path theory. The page has a quick calculator driven by scores and a data engine that computes the measures from monthly figures. V-SBM is not a method prescribed by any statute or valuation standard; it supplements a valuation and does not replace one.

V-SBM Interactive Calculator

PAC-Bayes Verified Manifold Bridge
V-SBM value₹69.64 Cr-30.4% vs Pitch Claim
Ledger gap squared (W_2²)0.336Gap factor: 0.776×
Cohort drift (d_FR)0.420 radDrift factor: 0.846×
Indicative range±₹13.83 CrRetention factor: 1.060×
₹100 Cr

Model term: terminal distribution ν_T of the bridge

0.58

Model term: 2-Wasserstein distance W_2(μ_0, ν_T)

0.42 rad

Model term: Fisher–Rao distance d_FR

0.65

Model term: path signature of the data

0.25

Model term: entropic regularisation ε

Claimed path, supported path and indicative range

Path implied by the pitch deckPath supported by the recordIndicative range
Audited GST and bank record Supported path Supported value Time: today → five-year horizon
Indicative range, lower₹55.81 Cr
Indicative range, upper₹83.48 Cr

Move a slider or choose a scenario. The scenarios are illustrative settings, not valuations of any company.

Run V-SBM on Your Own Data

The quick calculator above works from scores. The engine below computes the same measures from monthly figures: it calculates the 2-Wasserstein distance between the forecast and the actual growth distributions, cross-checks it with an entropic (Sinkhorn) transport plan, measures the Fisher–Rao distance between the first and second halves of the record, computes the level-2 path signature of revenue and spend, and builds a split-conformal band from the forecast errors. It starts with illustrative sample data; replace it with your own.

Columns: month, actual revenue (GST or bank), the forecast made for the same month, and product or R&D spend (optional). Paste from a spreadsheet or type comma-separated values, one month per line, at least 8 months. Use plain numbers without thousands separators, all in the same unit. The figures are processed in your browser and are not sent to our server.

V-SBM value–
90% band–
W₂ distance, in units of volatility–
Fisher–Rao distance–
Actual revenueForecast90% conformal band around the forecast
MeasureResultDetail
The engine needs JavaScript to run.

What does V-SBM measure?

A pitch deck describes where a company says it will be. The GST returns and bank statements describe where it has been. V-SBM asks how far apart those two pictures are and reduces the claimed value accordingly. In mathematics, a Schrödinger bridge is the most likely path between two probability distributions; the model applies that idea, with related ones, in its three factors.

Startup inputLaw appliedWhat it stands for
Valuation claimed in the pitch deckTerminal distribution of the bridgeThe starting figure that is being tested
Gap between the audited GST and bank ledger and the forecast path2-Wasserstein distance of optimal transportHow much the forecast has to move away from the record
Drift in unit-economics cohorts: LTV, CAC and churnFisher–Rao distance of information geometryHow much the business model itself is shifting
Net dollar retention and product-led synergyPath signature from rough-path theoryWhether the order of events in the data supports compounding
Cash-flow volatility and compliance frictionEntropic regularisation in the Sinkhorn algorithmNoise that widens the range and weakens every other signal

What does the calculator compute?

The table gives the exact formula behind each factor and its value at the default settings. The constant √24 reflects an assumed 24 months of data.

FactorFormula usedDefault
1. Ledger gape−0.62 × gap² − 0.18 × volatility0.77603
2. Cohort drifte−1.15 × √((drift² + 0.12 × volatility) ÷ (2 × √24))0.84627
3. Retention1 + 0.22 × (retention − 0.35) × e−0.35 × volatility1.06047
Indicative range± value × (0.095 + 0.28 × volatility + 0.08 × drift)± ₹13.83 Cr
V-SBM value = claimed value × ledger gap × cohort drift × retention

Worked example: B2B AI-SaaS company in Chennai

Illustrative figures, not a real client. A B2B AI-SaaS company claims a pre-money valuation of ₹100 crore in its pitch deck. Reconciling 24 months of GST returns and bank statements gives a ledger gap of 0.58, cohort drift of 0.42, a retention score of 0.65 and volatility of 0.25.

StepWorkingResult
Ledger gape−0.62 × 0.58² − 0.18 × 0.25 = e−0.25360.77603
Cohort drift(0.42² + 0.12 × 0.25) ÷ (2 × √24) = 0.02107; e−1.15 × √0.021070.84627
Retention1 + 0.22 × (0.65 − 0.35) × e−0.35 × 0.251.06047
V-SBM value₹100 Cr × 0.77603 × 0.84627 × 1.06047₹69.64 Cr
Indicative range₹69.64 Cr × (0.095 + 0.28 × 0.25 + 0.08 × 0.42) = ± ₹13.83 Cr₹55.81 Cr to ₹83.48 Cr

The model supports about 70% of the claimed value. The ledger gap accounts for most of the reduction.

What does the data engine compute?

MethodWhat is computedHow it enters the value
2-Wasserstein distanceExact optimal-transport distance between the empirical distributions of forecast and actual monthly log growth, divided by the volatility of actual growthW₂ in the ledger gap factor
Schrödinger bridge (entropic optimal transport)Sinkhorn iterations in the log domain between the same two distributions, with regularisation 0.1; reported as the debiased Sinkhorn distanceCross-check on W₂; shown, not multiplied
Fisher–Rao distanceRao distance between two normal distributions fitted to the first and second halves of actual monthly growthd in the cohort drift factor
Path signature, level 2Iterated integrals of the path of revenue and spend, each scaled to a unit range; the Lévy area gives a lead score between −1 and 1 (positive when spend leads revenue)Retention input = (1 + lead score) ÷ 2
Split-conformal predictionThe ⌈0.9 × (n + 1)⌉-th smallest relative error between forecast and actual90% band applied to the value
VolatilityStandard deviation of actual monthly log growthε in all three factors

Inputs to the factors are kept within the range of the quick calculator (W₂ 0.10 to 1.40, Fisher–Rao 0.04 to 1.20, ε 0.05 to 0.80); the table flags any value that was limited. The engine was checked on the sample data against two open-source libraries, POT (Python Optimal Transport) and FilterPy, and against independent numerical calculations.

Worked example on the sample data

Illustrative figures, not a real client. The sample has 24 months of actual revenue, the forecast made for those months and product spend, against a claimed valuation of ₹100 crore.

MeasureResult
Volatility of actual monthly growth6.54%
2-Wasserstein distance0.0655, or 1.0013 in units of volatility
Sinkhorn distance1.0043
Fisher–Rao distance0.8617
Lévy area and lead score-0.0652; lead score 0.1303; retention input 0.5652
Split-conformal quantile63.26%
Factorsledger gap 0.5308, cohort drift 0.7249, retention 1.0463
V-SBM value₹100 Cr × 0.5308 × 0.7249 × 1.0463 = ₹40.26 Cr
90% band₹14.79 Cr to ₹65.73 Cr

The forecast grows faster and far more smoothly than the record, so the two growth distributions sit about one standard deviation apart and the ledger gap factor falls to 0.5308. Growth also became more volatile in the second half, which lowers the drift factor. The record supports about 40% of the claim.

How should the result be read?

  • The constants are judgement. The numbers inside each formula were set by the author from practice. They are not fitted to a data set and have not been independently validated. Changing them changes the result.
  • What the range means. In the quick calculator it is a rule-of-thumb width. In the data engine it is a 90% split-conformal band of the forecast's relative error; applying that band to the value is the model's convention.
  • Short series are noisy. With 24 months of data the measured statistics can move noticeably from one month to the next. Treat the result as a reading to be explained, not as a precise figure.
  • Avoid double counting. If the starting value already reflects these risks through its discount rate or probability-weighted cash flows, applying the model again counts the same risk twice.
  • It is not a statutory method. V-SBM is a supporting analysis. A valuation under section 247 of the Companies Act, 2013, Rule 21 of the FEMA (Non-debt Instruments) Rules, 2019, Rule 57 of the Income-tax Rules, 2026 or Regulation 35 of the IBBI (Insolvency Resolution Process for Corporate Persons) Regulations, 2016 is made by the methods and the signatory that the law requires; see which valuation report each law requires.

Frequently Asked Questions

What is the V-SBM valuation model?

V-SBM is an illustrative scoring model by CA V. Viswanathan that reduces a valuation claimed in a pitch deck for the gap between the forecast and the audited GST and bank record, for drift in unit economics and for cash-flow volatility, with a credit for strong retention.

Is V-SBM a recognised statutory valuation method?

No. It is a supporting analysis. Valuations under the Companies Act, FEMA, income-tax law or the Insolvency and Bankruptcy Code are made by the methods and the signatory those laws require.

What inputs does V-SBM need?

The quick calculator takes five scores. The data engine takes monthly actual revenue from GST returns or bank statements, the forecast made for the same months and, optionally, product spend, for at least 8 months, with the valuation claimed.

Published 14 November 2026. Figures on this page are illustrative and are not a valuation of any company.

Other Viswanathan Valuation Models

These proprietary models are illustrative supporting analyses, not regulator-approved or prescribed statutory valuation methods. Use purpose-appropriate valuation evidence and avoid double-counting risk across models.

Need a Signed Valuation Report?

V-SBM is a supporting analysis. For a valuation report under the Companies Act, FEMA or income-tax law, signed by CA V. Viswanathan, IBBI Registered Valuer (IBBI/RV/03/2019/12333), contact us. We serve clients across India from Chennai.