Define the core capability every part of the venture reuses. Score each segment against two evidence-weighted tests. Watch the ranking move over time, and win one before you expand.
Mustard plans a venture as a set of market segments built on a single shared capability — then decides, with evidence rather than enthusiasm, which one to win first. This guide explains every part and gives you worked examples to fill all eight slots, whatever you're building.
The core is the one thing every segment is built on. The most common mistake is naming a market ("people who travel," "small businesses") instead of a capability. A market defines who you serve; a capability is the shared method that makes every segment recognizably yours.
"We serve music fans." · "We help small businesses." · "Anything to do with travel."
"Mint a physical presence moment into an owned asset." · "Install one AI workflow, then hand over the keys." · "Match idle local supply with verified presence."
Test your draft: could two very different segments both be built on it? If yes, you have a core. If it only describes one product, it's a segment.
Every segment must clear two gates to belong on the same venture. They're capability questions ("is this really the same business?"), not market questions ("is this big?").
Does it reuse the core capability?
If it needs a brand-new engine, it's a different company wearing your logo.
Does it feed the same center?
Does it add to one shared thing — a data pool, a network, a trust layer — every other segment also feeds? If it stands alone, park it.
A green light you flip on a hunch is how plans lie to founders. So a test isn't pass/fail — it's your honest read (0–100) times how good your proof is. Drag the sliders and change the evidence tier to feel it.
The segment's score is the weaker of its two adjusted tests — a segment that aces one gate and flunks the other still isn't the same business.
The bar is the score a segment must clear to count as Qualified. Set it once and hold it constant, so every segment is judged the same way. A useful default is 60.
Clears the bar on both gates with real-enough evidence. Eligible to be the beachhead or a later expansion.
Qualifies, but the beachhead isn't won — so it waits.
Below the bar — usually thin evidence, not a bad idea. Go get proof.
Eight is a ceiling, not a quota. Generate many candidates, keep only the ones that share your core. Here's how to generate them, three worked examples to copy, and a starter bank.
The move: brainstorm 12–15 candidates across these lenses. Score each. Keep the eight that share the core and clear (or could clear) the bar. Everything else goes on the parked shelf — recorded, not deleted.
Same framework, three kinds of venture. Notice how many segments sit parked — that's the evidence discount working, not a flaw. A real pre-launch plan has proof for one or two and beliefs about the rest.
Stuck on empty slots? Borrow and adapt. These are seeds for the who-you-serve column — still run each through your two tests.
Of the qualified segments, exactly one is the beachhead: the single market you win completely before expanding. Mark it won only when you truly own it. Until then, every other qualified segment is "next, not now."
Highest evidence-weighted score (not biggest market) · smallest segment you can dominate (not broadest you can reach) · the one whose customers your next segment already trusts.
Hard to dislodge — repeat usage, references, a real moat — not "we have some customers." Flip it too early and you've only fooled yourself.
Expansion isn't "do the next segment." It's a sequence where each stage consumes the previous stage's output, so the whole thing converges into one venture.
See each worked example's build order above for how this plays out concretely.
You don't know the right segment up front — you discover it by watching the market re-rank your guesses. Mustard treats the plan as a living thing.
Put a segment in front of the market and gather the one decision signal you chose.
Score every segment and watch the order move — take a dated review so next time you see the deltas.
Promote a winner to beachhead, park laggards, revive a parked one when conditions change.
The two habits that make it real: review snapshots (so "I feel like this is working" becomes "its score went 55 → 68"), and a parked shelf (so you never re-try a dead idea blind).
The whole method in one breath
Name the core. Gate the segments with evidence-weighted tests. Cap it at eight. Win one before you expand — and let the ranking, not your enthusiasm, tell you which one.
Ready? Open the Canvas tab up top to build yours — or the FAQ for the questions that come up along the way.
Want the deeper manual — the reasoning behind each test and the beachhead discipline? Read the Beachhead Field Guide ↗ (password-protected).
The questions that come up while you're working — answered to keep you moving and keep you honest. Search, or filter by topic.
Score the proof you have, not the future you want. Mustard can't verify a single number you enter — it organizes your thinking and exposes unproven confidence. The truth still lives in customer conversations. Use the canvas to decide which segment to win first; use the guide to learn the method; use the conversations to find out what's real.
Every term the framework uses, in one place — each with a quick definition and a worked example. Search, or filter by group.
Examples use MINY (a proof-of-presence music venture) and AnyGame (an illustrative games venture — swap in your own). The vocabulary is the same whatever you build.