Tool
Most ventures fail because the architecture doesn't work — not because the founders lacked vision or effort.
The Venture Assistant guides you through the full structural design of your venture before a penny is spent building. The process below is what it takes you through.
The cost of an architectural change
Phase 1 — Design
How is the problem currently being solved? What activities, costs, and players are involved in the standard approach? This is your baseline — the architecture you are designing against.
Within the conventional model, identify the single activity that makes the business structurally unviable at scale. This is the constraint that determines whether a new venture can ever be financially sound — not a cost to be reduced, but a constraint to be designed around.
In small claims legal disputes, the critical cost driver was expert legal facilitation — skilled professionals needed to de-escalate conflict and guide both parties to settlement. Expensive, hard to scale, unavoidable in the conventional model.
Use theory — psychology, economics, systems science — to find a mechanism that eliminates or bypasses the critical cost driver entirely. Not a cheaper version of the same thing. A different mechanism altogether. This is where the genuine architectural innovation lives, and where financial modelling confirms whether it works before anything is built.
Instead of expert facilitators, digitally disintermediate the negotiation — keep parties in separate spaces, use an algorithmic auction to calculate the optimal settlement. The same logic powered microfinance: instead of expensive individual creditworthiness checks, lend to peer groups who self-select and hold each other accountable.
Identify what significant cost the product must remove from how customers currently achieve an important objective. Not a marginal improvement — a meaningful reduction in something they genuinely need to solve. The larger the cost eliminated, the stronger the adoption pull.
Calculate the biggest cost that will emerge as you grow, given how you are currently conceiving the product and operations. Find it now — in simulation — before you have built anything and while changing course is still cheap.
Structure the product so customers can easily perceive and extract the value. Friction here — even when the underlying product works — kills adoption. The product needs to communicate what it does and deliver that outcome with minimal effort from the customer.
A gateway partner is an organisation that already has a trusted relationship with your target customer — a financial institution, a professional body, a trade association — and through which your product can credibly be introduced. Rather than building a direct customer relationship from scratch, the venture is designed to be sold through a gateway partner's existing channel. This requires designing a distinct product specifically for the partner, whose own needs must be addressed before they will carry yours.
Identify the single resource that is essential to delivering your core competence — and design a moat around it. The goal is to make your position structurally difficult for competitors to replicate, not just temporarily hard. A proprietary dataset, a certified network, an exclusive partnership: something that compounds over time and cannot simply be bought.
Build a full picture of how product flows to the customer, how marketing and information flows, and how money flows back. Every activity costed. Every role mapped to an at-scale HR structure. The design phase is complete when this model holds together financially — not before.
Phase 2 — Build
Identify the activities where you most need to build confidence — specifically those that must work within tight cost parameters defined during design. Test each independently, before assembling the system. The question is not "can we build this?" but "can we deliver this at the cost the model requires?"
Test clusters of related activities together — the sales process, the delivery process, the support process. Stress-test the interfaces between components before connecting everything into a single system.
Launch the smallest operating unit that is statistically representative of the venture at scale. Not a pilot in the loose sense — a real, functioning miniature of the full system, designed to generate data that is meaningful at scale.
Replicate the proven operating unit. At this point, growth is replication — not reinvention. The structural work is done. The venture is designed to scale without requiring a new architecture at each stage.
Free. Runs in Claude Code. Based on IVE methodology developed at Cornell University.
Get the Venture Assistant