How do you validate a financial product before launch?
A financial product is validated before launch by running a bounded synthetic-audience study on the proposition. emable's Market Validator puts one proposition in front of up to 100 AI respondents in Quick or Deep mode and returns a verdict with evidence in minutes (capability status live, 2026). The evidence is labelled directional; it does not replace a regulated approval process or a live market test.
Updated . Figures keep their period, source and reliability grade.
Key figures
| Metric | Value | Period | Source | Grade |
|---|---|---|---|---|
| AI respondents per study, cap | 100 synthetic respondents | 2026 | emable Market Validator capability | A |
| Study modes | 2 modes (Quick and Deep) | 2026 | emable Market Validator capability | A |
| Advisor register records for market context | 1,288,561 register records | Live, 5 September 2026 | emable public EU overview API | A |
The answer
In 2026, a financial product is validated before launch by running a bounded synthetic-audience study with up to 100 AI respondents. emable's Market Validator does this: one proposition, Quick or Deep mode, a verdict with evidence in minutes (status live). The evidence is labelled directional. It shows where a proposition is weak before real customers, distribution partners and compliance costs are involved. It does not replace a regulated product approval process, and it does not replace a live market test.
The numbers
Every figure behind this answer sits in the key figures table above, with its value, unit, period, source and reliability grade: A official statistics, B reconciled from published sources, C modelled estimate, D indicative.
How the number is built
The 100 is a cap, not a sample size claim. A study is bounded: a fixed respondent count, one proposition, and a server-side cost check before anything runs. That is what makes the result repeatable and the charge predictable.
Step one is to write the proposition as a customer would read it. Who it is for, what it costs, what it promises, what it excludes. A vague proposition produces a vague verdict.
Step two is to define the audience: segment, country and situation. The respondents are synthetic, built to that definition. They are models, not customers, which is why every finding is labelled directional.
Step three is to choose a mode. Quick is the shorter study, Deep the longer one. Both return a verdict with the evidence behind it, so a reader can trace each claim to respondent output.
Step four is to change the proposition and run again. The value is in the comparison between variants, not in one score.
Each run passes through the shared lifecycle: queued, running, succeeded, failed, refused, awaiting_approval. A trial needs a login and no card.
What it means for your team
Product, distribution and compliance can read the same verdict before a pilot is scheduled. A weak variant is dropped on evidence rather than on opinion. A strong variant goes into the pilot with the objections already known.
Pair the study with register data when the question is size. The Validator says how a proposition lands. The EU register aggregate says how many advisor records exist in the target country, at a date, with a coverage level. Neither replaces the other.
Treat the verdict as directional in writing. A synthetic audience finds weak spots quickly; it does not measure conversion. Statistical significance is not claimed and should not be quoted.
Keep the study bounded on purpose. More respondents do not turn a synthetic audience into a market. A second variant does more for the decision than a larger first run.
Go deeper
- Validate before shipping use case: the outcome, inputs and charging unit.
- Market Validator capability: status, schemas and the safe result.
- MCP connector docs: run a study from Claude, Codex or another AI client.
- What is embedded intelligence for financial distribution?: where validation sits in the platform.
Related questions
- Are the respondents real customers?
- No. They are synthetic respondents built to the audience definition. That is why every finding is labelled directional and statistical significance is not claimed.
- What does bounded mean?
- A fixed respondent count with a cap of 100, one proposition per study and a server-side cost check before the run starts. The result is repeatable and the charge predictable.
- How do I run it from an AI client?
- The Emable Finance Connector at https://api.emable.ai/mcp exposes Validator workflows over MCP with OAuth. Every run passes through the shared lifecycle from queued to succeeded, failed, refused or awaiting_approval. A trial needs a login and no card.
Methodology and limitations
The respondent cap and the two modes are capability parameters published by emable for the Market Validator. A study is bounded by a fixed respondent count, one proposition and a server-side cost check, and its verdict is labelled directional because respondents are synthetic. The register count is quoted for market context only and is not an input to the study.
Primary sources
See the validate-before-shipping use case