Abstract
This dashboard implements a coupled jump-diffusion and stage-Markov forecasting layer over a venture portfolio. Continuous between-round revaluation is modeled by a log-diffusion; discrete priced-round events are Merton-style jumps with stage-conditional mark-up distributions (Merton, 1976); company outcome trajectories follow an 8-state Markov chain over {Seed, A, B, C, D+, IPO, M&A, Dead} calibrated on Crunchbase 2010–2022 outcomes. The choice over the simpler Cochrane (2005) cross-sectional CAPM framework is deliberate: stage-conditional metrics (runway, dilution waterfall, follow-on probability) require an explicitly time-stepped Markov layer that the cross-sectional approach cannot produce (Cochrane, 2005; Kaplan & Schoar, 2005). Power-law tail behavior emerges naturally from the convolution of log-normal jumps with stage-attrition.
Stochastic model
Venture-backed companies are not best described by continuous-state diffusions because their valuations re-price discretely at priced funding events and because outcome distributions are heavily power-law in the upper tail. The well-known stylised facts are: (i) median follow-on multiples are modest (typically 1.5–2.5× per stage) but the upper-decile multiples are extreme (10–100×+); (ii) attrition is high and front-loaded (~25–35% of seed-funded companies fail before Series A); (iii) outcome correlation across vintage years is non-zero but weak. We model the post-money valuation V_t as a Merton jump-diffusion coupled to a discrete-state stage Markov chain:
d log V_t = μ dt + σ dW_t + sum_k J_k dN_k
The diffusion component represents continuous between-round revaluation (revenue progress reflected in the next round's negotiation but not yet a priced event). The jump component J_k dN_k fires only at priced round events and is the dominant source of variance for an early-stage company. The drift μ is stage-dependent, and reflects the Cochrane (2005) finding that the venture-asset CAPM beta is high (≈1.5–2.0) but its arithmetic-mean log-return is only mildly positive.
Stage transitions follow a discrete Markov chain over states S = {Seed, A, B, C, D+, IPO, M&A, Dead}. The 12-month transition matrix is calibrated from the Crunchbase 2010–2022 outcomes data and from the Correlation Ventures benchmark study. The transition matrix is not row-stochastic in the strict sense because we allow simultaneous probabilities of graduation, stagnation, M&A and shutdown in a single annual epoch; we re-normalize numerically.
Conditional on graduation to stage k+1, the priced-round jump J_k is drawn from a log-normal with mean equal to the empirical median step-up (Seed→A 2.4×, A→B 2.1×, B→C 1.7×, C→D 1.5×, D→IPO 1.3×) and standard deviation consistent with Pitchbook quartile spreads. Power-law tail behavior is captured because the convolution of log-normal jumps with a graduation probability that itself contracts at later stages produces a distribution with Pareto-like tails on the multi-stage outcome aggregate.
Liquidation preferences are modeled with a 1× non-participating preferred waterfall in stack order. The runway panel (extra) integrates cash balance minus monthly burn plus monthly MRR through the next-round date; the dilution overlay applies a 22% pool grant + new-money dilution at each priced round, calibrated to median Carta data.
Calibration
Per-stage drift μ is set so that the simulated stage-graduation conditional return matches the Correlation Ventures empirical mean priced-round multiple. Per-stage diffusion σ is calibrated to the Pitchbook quartile spread of follow-on multiples: Series A σ ≈ 0.55, Series B σ ≈ 0.45, Series C σ ≈ 0.38, Series D+ σ ≈ 0.32. These are large by public-equity standards but standard for the venture asset class.
The transition matrix is fit by maximum likelihood on a 12-month grid over the Crunchbase outcome dataset (filtered to U.S. companies with disclosed funding). Key entries: Seed→A 30%, Seed→Dead 20%, A→B 32%, B→C 30%, C→D 25%, D+→IPO 20%, D+→M&A 20%. The IPO and M&A states are absorbing.
Burn rate uses median net-burn from the Bessemer State-of-the-Cloud SaaS benchmarks: $0.5M/mo at Seed,