A ranked answer to “what should we build next?”, refreshed as the audience changes.
The engine generates candidate experiences from your actual audience, scores each on seven dimensions, writes a thesis you can argue with, and attaches reasons that can be checked against your own records. A score you cannot interrogate is a horoscope.
- Four-night coastal retreat · LisbonTravel·$84,600 contribution · break-even at 22 of 6087
- Twelve-seat private dinner series · 4 citiesGather·$61,400 contribution · 6 weeks to launch81
- Annual membership with quarterly dropsMembership·$18,900 monthly recurring at 9% conversion74
- Weekend festival, owned venueGather·$212,000 gross · $164,000 fixed before a ticket sellsRisk 7861
- Seven-night charter cruiseTravel·Deposit to operator due 11 months outComplexity 9143
Example product surface. Figures are illustrative.
Seven dimensions, each shown with its inputs.
The composite is a weighted roll-up of these seven. Complexity and risk score inversely — a low number is good — and are shown that way rather than quietly flipped behind the scenes.
Revenue potential
78How much gross could this plausibly produce at a realistic sell-through?
Tier prices × modelled mix × capacity, discounted to the Base case.
Margin
74How much of that gross survives contact with the costs?
Contribution after marginal cost per unit and fixed costs.
Demand
92Is there evidence anyone wants this, or is it a hunch?
Observed engagement, past purchase behaviour, and live demand signals.
Audience fit
88Does the audience you actually have match the audience this needs?
Segment overlap, geography, and prior willingness to travel or pay up.
Complexitylower is better
41How many moving parts, vendors, jurisdictions and dependencies?
Blueprint defaults plus leg count, vendor count and travel involvement.
Risklower is better
38What is committed before revenue arrives, and how much of it is recoverable?
Non-refundable deposits, fixed cost timing, and cancellation exposure.
Time to launch
63How long from decision to doors open — and can the audience wait that long?
Critical path through the dependency graph for this experience type.
Example product surface. Figures are illustrative.
It does not only suggest more of what you already do.
Candidates are generated from experience blueprints across eight families. A musician’s audience might rank a retreat above a tour date; a chef’s might rank a membership above another pop-up. The engine has no opinion about your category — only about your audience.
- Blueprints carry sensible defaults — duration, capacity band, cost model, price anchors — which the model then overrides with your real numbers
- New experience types can be added without a schema change, so the catalogue grows without a migration
- Every candidate is shortlisted, dismissed or promoted, and dismissals inform the next run
- GatherFestival, showcase, weekend event, private dinner series, meet-and-greet
- ImmersionRetreat, camp, residency, writing or training week
- AccessBackstage tier, studio visit, small-room upgrade, listening session
- ParticipationTournament, competition, collaborative build, cohort challenge
- TravelDestination experience, cruise, multi-city series, tour package
- CommerceDrop, limited run, collector edition, bundled release
- MembershipAnnual membership, subscription tier, alumni programme
- EducationMastermind, cohort course, workshop series, certification
Reasons you can check, not adjectives.
Each opportunity stores a thesis and a list of reasons. Every reason carries its own provenance and, where it can, the evidence behind it — so a sceptical reader can go and verify the claim in their own data.
- Observed
511 members have travelled to an event of yours before.
travelled_for_event = true, last 24 months
- Observed
412 members registered deposit intent within 72 hours of the test opening.
demand_signals, kind = deposit
- Calculated
Median lifetime spend in the Traveler segment is 4.1× the audience median.
$1,840 vs $449
- External
Three vendors quoted the four-night package within 8% of each other.
quotes: $92,400 / $95,000 / $99,600
- Modeled
Base case clears break-even at 72% sell-through with $47,000 contribution.
break-even 22 of 60
Confidence 0.58 — Moderate. Ceiling 0.60, set by the modeled reason. Convert deposits to orders and the ceiling rises to 0.95.
Example product surface. Figures are illustrative.
- ObservedRecorded directly from your data. Not an estimate.0.95
- CalculatedArithmetic on observed data. Deterministic and reproducible.0.90
- ExternalSourced from a third party. Accuracy depends on that source.0.75
- ModeledAn estimate from assumptions you can inspect and change.0.60
- AIGenerated by a language model. Judgement, not measurement.0.50
Confidence is capped by the weakest reason supporting a claim — enforced in code.
A high score is a hypothesis. Test it before you spend.
The engine’s output is an argument for spending money. So the step after ranking is never a purchase order — it is a cheap test whose result flows straight back into the model.
Waitlist
Zero-friction interest. Sets an upper bound on demand, nothing more.
Refundable deposit
The strongest pre-capital signal there is. Money moves; the obligation is tracked.
Presale
Real orders against real inventory, before the public window opens.
Tier test
Which price they choose, not whether they would buy. This moves the tier mix.
Each signal is tied to a tier, so a deposit at the $3,600 level does not just increment a counter — it changes the modelled mix, the blended price, the break-even point and the confidence ceiling on the whole opportunity.
Let the engine make its case.
Connect a source and see what ranks first — with every factor and every reason open to inspection.