product
VCs and operators doing TAM and market-sizing work. They arrive with a market in mind ("AI sales-ops agents, US, mid-market") and a deadline: a partner meeting, an IC memo, a board deck. Their context is high-stakes and time-boxed. They are numerate and skeptical, they have been burned by a $50B-because-the-deck-said-so figure before, and the first thing they do with any number is ask "where did that come from?"
The job to be done: get to a market size they can defend in the room in minutes, with the sourcing and the method visible enough to survive scrutiny. Not a research project. A decision input.
Strata is a community-curated market-sizing engine that replaces IBISWorld / Statista / Gartner-lite reports by exposing the primitives (atomic sourced numbers) and the frameworks (parameterized recipes that combine them into TAM), plus an expert layer for correction.
It exists because market sizing today is either a black box (a number with no visible derivation) or a slog (weeks of analyst spreadsheet work). Strata makes the derivation the product: pick a market, pin its scope, watch competing methods triangulate, and walk away with a hash-pinned, forwardable artifact where every figure links to its source.
Success looks like an operator pasting a /tam/<hash> link into a memo and the
recipient being able to click any number down to a regulator filing, then
disagree with a specific assumption instead of the whole figure.
Editorial, considered, quietly authoritative. Three words: rigorous, composed, candid.
The voice is that of a sharp analyst who respects your time: states the number, shows the work, names the disagreement, and stops. The Instrument Serif display type already sets this register, premium and intellectual, closer to a finance publication than a SaaS dashboard. Calm confidence, never hype.
The defining tension to hold: the feel is editorial and unhurried, but the job is fast. Considered design must never read as slow or precious. A fund partner pulling a number 20 minutes before a meeting should feel the restraint as clarity, not ceremony. Premium restraint that still gets out of the way.
Emotional goal: the reader trusts the number because they can see why it is trustworthy, not because the UI insists on it.
- Generic SaaS template. Inter for everything, purple-to-blue gradients, cards nested in cards, the rounded icon tile above every heading, the hero-metric template. This is the 2026 AI-slop default and the fastest way to read as untrustworthy to a numerate audience. Strata's editorial type system exists precisely to escape this lane.
- Legacy research reports (IBISWorld / Statista / Gartner). Dense PDF walls, opaque "our analysts estimate" methodology, stock-photo corporate gloss, zero provenance you can click. Strata is the inversion of this: the methodology is the interface.
- Academic / dry. Wall-of-text, joyless, no hierarchy, rigor performed rather than delivered. The audience is operators on a clock, not reviewers of a thesis. Hierarchy and pace are not optional.
Not on the list but worth stating: not crypto/web3 either, no neon-on-black or speculative hype. Strata is sourced and sober.
- Provenance is the product. The number and its source are one object, not a figure with a footnote somewhere else. If a figure cannot show its derivation, it should not assert itself like one that can.
- Disagreement gets a page, not a footnote. When methods or sources conflict, surface the spread as a first-class view ("lower $X, higher $Y, N× apart, here's why"). Triangulation visible beats a false single number.
- Honest about limits. Never fabricate a figure to fill a slot. "Unsizable", "dark estimate", "modeled input", and "world-aggregate needs a paid key" are shown plainly. Candor is a feature for this audience, not a weakness.
- Fast to a defensible answer. Optimize the path from market query to a number the user can stand behind. Editorial restraint serves speed and clarity; it never adds ceremony between the operator and the figure.
- Restraint as authority. Earn trust through composure: tight hierarchy, deliberate type, no decoration that does not carry meaning. The interface should feel like it knows the answer, not like it is selling one.
Target WCAG 2.2 AA. Body text ≥4.5:1, large text ≥3:1, against the light
#FAFAFA paper, including the muted --fg-3 metadata tier (verify, do not
assume). Full keyboard navigation and a visible focus ring on every
interactive element, including the ReactFlow pyramid graph nodes. prefers- reduced-motion honored everywhere, this is load-bearing given the count-up
TAM headline, sensitivity tornado, and graph/scenario-tree animations: each
needs a crossfade or instant-state alternative. Do not encode meaning in color
alone (source-tier badges already pair hue with a label, keep that invariant).