Methodology: readiness and simulation engines

Readiness engine

The Commercialization Readiness Dashboard pulls together the insight sections you work through across the app. Each section - for example vitamin vs painkiller, size of prize, business model, go to market - can be scored from 1 to 5 on a set of dimensions. The dashboard converts those scores into percentages, blends them with your hypothesis progress (validated, refuted, still testing), and rolls everything up into a single readiness level.

The AI does two things here. It reads your transcripts and focus group activity to draft commentary and suggested scores for each section, and it summarizes the overall picture. You stay in control: you can edit notes, override any score, and regenerate at any time. The readiness level is a weighted average of the sections you have completed, so it improves as you analyze more areas and capture more evidence.

How the 16 readiness dimensions work

Custard assesses commercialization readiness across 16 structured dimensions drawn from established customer discovery and I-Corps methodology. These dimensions span the full arc from problem validation through to scaling:

• Problem significance: Is the problem a painkiller (must-have) or a vitamin (nice-to-have)? How frequently does it occur and how severe are the consequences?
• Customer segment clarity: Have you identified a beachhead segment? Do you understand who is the economic buyer versus the end user?
• Size of prize: Is the addressable market large enough to justify investment? Is the growth rate compelling?
• Solution fit: Does your proposed solution directly address the root cause of the problem? Have customers confirmed they would switch?
• Competitive advantage: What is your defensible moat - IP, network effects, switching costs, or something else?
• Business model: Is the revenue mechanism clear? Have you tested willingness to pay?
• Go-to-market: Do you have a realistic channel strategy and customer acquisition plan for the first 100 customers?
• Hypothesis progress: What share of your core hypotheses are validated, refuted, or still open? Validated hypotheses lift the overall score; open hypotheses dampen it.

Each dimension is rated 1-5. A score of 1 indicates a dimension is untested or very weak; 5 indicates strong, evidence-backed confidence. The AI drafts an initial score and short commentary for each dimension by synthesizing your transcripts, notes, and focus group responses. You review and override as needed.

How scores are calculated

Each insight section contributes a percentage score derived from its dimension ratings. Sections you have not yet completed are excluded from the average - the readiness level reflects only the ground you have actually covered, so it tends to rise as you do more analysis rather than starting artificially high.

Hypothesis progress acts as a weighting signal layered on top of the section scores. A project with many validated hypotheses and strong section scores will score higher than one with similar section scores but mostly untested assumptions. This design aligns the readiness metric with the I-Corps expectation that evidence quality matters as much as analytical coverage.

The Business Value Calculator (simulation engine)

The Business Value Calculator is Custard's simulation engine. It lets you model the financial and strategic impact of your innovation before you have real revenue data to rely on. You set the inputs - adoption rate, price point, deal size, implementation cost, productivity gain - and the engine runs a Monte Carlo-style probability sweep across a plausible range of outcomes.

The result is a distribution of outcomes rather than a single point estimate. You see a P10 (pessimistic), P50 (median), and P90 (optimistic) scenario, with the key drivers ranked by their contribution to the spread. This is intentional: most commercialization decisions fail not because the median case was wrong but because the downside tail was worse than expected.

What the simulation models

• Customer adoption curve: How quickly will your target segment adopt the product, given your go-to-market strategy and competitive context?
• Revenue potential: What is the realistic annual contract value or price per unit, accounting for the willingness-to-pay signals you captured in customer interviews?
• Cost to deliver: What implementation, support, or training costs does the customer bear, and how do those affect net value delivered?
• Time to value: How long before the customer realizes measurable benefit? Longer payback periods reduce perceived value and increase churn risk.
• Competitive displacement: If your customer is replacing an existing solution, what switching costs and incumbent advantages must you overcome?

How the AI uses your research data

When you run a simulation, Custard pre-populates the input fields using signals it has extracted from your interview transcripts, focus group sessions, and insight sections. If you have captured willingness-to-pay data from interviews, that becomes the default price range. If you have validated the problem frequency, that informs the adoption rate prior.

You are always free to override any pre-populated value. The AI-suggested inputs are a starting point derived from your evidence, not a locked forecast. After each simulation run, Custard flags which inputs have low confidence - either because the underlying evidence is thin or because the signal is conflicting across interviews - so you know where to focus additional discovery.

Limitations and responsible use

• Outputs are probabilistic estimates. They reflect the assumptions and evidence you provide. Weak or thin inputs produce weak conclusions. A simulation run on five interviews is directionally useful, not conclusive.
• AI-drafted commentary and scores are decision support, not decisions. Always validate against direct customer conversations and primary source data before presenting to investors, licensing committees, or grant panels.
• The readiness score is not a universal standard. It is calibrated for early-stage university and research-institution commercialization contexts. It may not translate directly to corporate venture or government procurement pathways without adjustment.
• Simulation ranges are illustrative. The P10-P90 spread is computed from a parametric model seeded by your inputs, not from historical comparable-deal databases. Treat the output as a structured thinking tool rather than a financial forecast.
• Evidence quality is your responsibility. Custard can synthesize what you give it, but it cannot interview customers on your behalf or verify whether the people you spoke with are genuinely representative of your target segment.

Responsible AI notice: All readiness scores, simulation outputs, and AI-generated commentary are probabilistic decision-support tools. They are not a substitute for primary evidence, legal advice, or professional financial guidance.