VOI Research Brief
VOI (Value of Idea) estimates early-stage startup potential using structured market data, competitive density, and momentum signals. We don't guess—we calculate.
Every evaluation runs through the same auditable pipeline. Quantitative layers are computed first; qualitative copy must align with those numbers—never the other way around.
Core inputs
Before a score exists, we assemble four evidence lanes: the size of the prize, the intensity of incumbent pressure, the velocity of market pull, and the contextual guardrails that keep estimates honest.
Total Addressable Market (TAM) sketches the full theoretical demand curve. Serviceable Addressable Market (SAM) is the sober slice you can reach with your actual product, geography, and motion. VOI rewards disciplined SAM sizing—boundary constraints matter more than headline hype. Inflating SAM to borrow credibility from an entire category is treated as a modeling failure, not a flex.
We ingest a ranked competitor matrix and attach funding pressure to each name. A handful of well-funded giants behaves differently from a long tail of bootstrapped clones. The engine tracks how concentrated war chests are—average versus peak funding—and uses that shape to estimate how violently crowded the arena feels before you ever ship v1.
Signals compress trend velocity, buyer urgency, and category tailwinds into a 1–5 scale. This is not vibes—it is the model's structured read on whether the market is pulling ideas like yours forward or standing still. Extreme scores require defensible differentiation; generic wrappers cannot borrow AI buzzwords to smuggle in a 5.
Stage assumptions, cohort placement, and the plain-language brief you provide anchor extraction. Optional founder inputs (audience, named alternatives) sharpen competitor discovery and keep the AI from hallucinating a market that sounds plausible but isn't yours. Garbage-in still means garbage-out—but structured prompts dramatically narrow the failure mode.
Formula stack
VOI is a weighted additive model: forty points for market scale, forty for signals, twenty for competitive moat—then a hard tarpit cap when the idea is a saturated wrapper. No exponentials, no mystery curves: you can recompute the score from the published intermediates.
Rewards realistic serviceable markets. When SAM is expressed in millions of USD, $1B+ of serviceable market (≥ 1,000M) earns the full forty-point allocation because it signals room to build a serious company; smaller SAMs scale the contribution linearly so fantasy TAMs cannot smuggle in a free lunch.
MS_pts = min(SAM / 1000, 1) × 40
Evaluates early validation and market pull on a structured 1–5 scale. The engine divides by five and enforces a floor so even weak tailwinds still register—then multiplies into the forty-point lane. This prevents binary hype while still punishing ideas with no demonstrable momentum.
SIG_pts = max(0.1, signals / 5) × 40
Measures how crowded the space feels using competitor count and the distribution of funding. When many well-funded players cluster, saturation approaches one; sparse or underfunded matrices drift toward zero. The ratio of average to maximum funding captures whether a single oligarch dominates the cap table or power is diffuse.
CS = (1 − 1 / (1 + n × (F_avg / F_max)))²
Turns saturation into defensibility. One minus CS becomes the moat factor—clamped so it never collapses to zero—and scales the twenty-point lane. Founders facing gentle competition keep more of that slice; red oceans with concentrated billions see it collapse even if the press release still sounds cheerful.
Moat_pts = max(0.1, 1 − CS) × 20
Thin API wrappers, generic chatbots, and other tarpit ideas are flagged upstream. Their summed weighted raw score is hard-capped before the final mapping so buzzwords cannot manufacture false positives. Everyone else flows through a transparent clamp: round, bound between 1 and 100, and ship.
Raw = MS_pts + SIG_pts + Moat_pts If is_tarpit: VOI_pre = min(Raw, 35) Else: VOI_pre = Raw Final_VOI = round(max(1, min(VOI_pre, 100)))
This is the final convergence point for the three pillars. Each normalized component score is multiplied by its assigned institutional weight—40 for Market, 40 for Signals, and 20 for Moat—and summed to produce the pre-penalty raw VOI rating.
Total = (SAM_norm × 40) + (SIG_norm × 40) + (Moat × 20) SAM_norm = min(SAM / 1000, 1) SIG_norm = max(0.1, signals / 5) Moat = max(0.1, 1 − CS)
The deterministic finalization of the assessment model. The raw score is subjected to the transparent tarpit clamp. Ideas without defensibility are capped at 35 points, while other models are rounded and bound within the 1–100 scale, creating an auditable and reproducible investment decision tool.
If is_tarpit: VOI_pre = min(Total, 35) Else: VOI_pre = Total Final_VOI = round(max(1, min(VOI_pre, 100)))
Data engine
The narrative layer is grounded in retrieval. We document where numbers came from, downgrade junk automatically, and resolve conflicts with recency-aware rules—so diligence teams can trace the chain of evidence.
$