Skip to content
Leadspring

Methodology

How a Leadspring verdict is computed, what it rests on and where it stops. Thresholds version 2026-10.1.

What a run is

A run takes your brief and your quiz to a panel of 1,000 synthetic respondents from the market you choose: United States, Italy or the European Union.

Each respondent is a simulated person with an age, a place, an income band and habits, sampled from official statistics. They answer the quiz one by one; the answers are aggregated with a margin of error for each option.

Data the panel comes from

United States: US Census Bureau tables (2025). Italy: ISTAT tables (2025). European Union: Eurostat tables (2025).

Area, metropolitan region and age band are drawn from these tables. Each table carries its verification status in the report.

Nine facts before any run

Price, customer, who pays, market, problem, solution, alternatives, channel, and investment with its time frame.

If one is missing the run doesn't start and no quota is used. Answering "I don't know" is allowed and lowers the confidence by 4 points per fact.

How the verdict is set

Eight dimensions are scored from 1 to 10. The overall score sets the verdict: Go from 7.5, Early from 6, Pivot from 4.5, No-go below.

Caps apply whatever the score: a problem below 4 or willingness to pay below 3 limits the verdict to Pivot, and so does an LTV to CAC ratio below 1.

Benchmarks for the economics

LTV to CAC is healthy from 3 and a risk below 1; CAC payback is healthy within 12 months and a risk beyond 18 (David Skok, SaaS Metrics 2.0).

Gross margin is healthy from 70% and a risk below 40% (Bessemer and OpenView SaaS benchmarks). Customer lifetime is capped at 60 months.

Confidence and evidence

Confidence depends on the strongest evidence reached: your statements (E0), the synthetic panel (E1, 30%), sourced market facts (E2, 35%), at least 30 real answers (E3, 45% to 65% at 100 answers) and at least 10 leads with consent (E4, 80%). It never goes above 90%.

Below 40% the result is shown as a preliminary read, not a verdict. A synthetic-only run is always preliminary.

Purchase intent is discounted

People overstate what they would buy. Synthetic "I would pay" answers are discounted by 50% until real answers calibrate the domain (hypothetical bias, Murphy et al. 2005).

What it can't do

A synthetic panel anticipates; it doesn't replace real customers. It can be wrong on new categories, niche professions and taste-driven products.

That's why every report asks you to publish the quiz and shows real answers next to the simulated ones.

Know before you build.

Bring one idea. Leave with a verdict, the numbers behind it and a link to test it with real people.

  • Free plan, no card
  • One full run included
  • Real-people link included