HOW IT WORKS

Seven steps, one first customer.

Paste an idea, let 100 customers from the corpus ask what they would actually ask you, and find the real problem before you spend a dollar. Every number is counted, every decision is yours.

HOW IT WORKS · FIRST CUSTOMER

Seven steps. You answer between them; the corpus does the rest.

Every number on screen is counted from individual personas — never estimated by a model. Verdicts on your own answers stay yours: ducira never grades its own rebuttals.

Start step 1 free
  1. 1

    Questions you’ve heard

    Paste what real customers already asked you. Optional, kept verbatim.

    You type
    3 of your questions
    “How much does it cost?”
    “Does it work with what I already use?”
  2. 2

    Every customer question

    100 people who actually have your problem ask what they would ask you. You see how many of the 100 asked each one.

    84 questions · 63 of 100 on price
    Price63 of 100
    Trust41 of 100
  3. 3

    Hard questions

    Your best answer to each question, then the question that breaks it.

    You decide: refuted, or can’t refute
    26 hard · 9 can’t refute
    “If I’m too tired to cook, why am I not too tired to plan a subscription?”
  4. 4

    Real problems

    Only what you couldn’t refute goes in. We name the problem underneath, then ask the group who has it — and who would pay.

    3 problems · 41 of 100 · 12 pay
    41 of 100 have it · 12 would pay · 26 maybe
  5. 5

    Solution questions

    The questions whose answers define a solution to each real problem.

    You answer
    6 answered
    “What has to be true at 6 pm for dinner to be a non-decision?”
  6. 6

    Your system

    The solving process written as a repeatable system: steps, roles, tools, metrics — and the number that tells you to stop.

    5 steps · stop if waste > 15%
    5 steps · 3 roles · tripwire: wasted meals > 15%
  7. 7

    Sales questions

    What to ask before you explain your system to a real customer — tried on 30 personas first so only the openers survive.

    5 questions · 61% open

    “How many dinners last month did you skip?” · 22 of 30 open

THE CORPUS

One million US personas, each with a household, a job and a Tuesday night.

Built on NVIDIA’s Nemotron-Personas-USA. ducira picks the 100 whose lives match your idea and locks that sample for the whole run, so step 7 asks the same people step 2 did.

#4172 · 34 · Columbus, OH

ICU nurse · nights

“Honestly I’d just want it to be there. The planning part is what kills me, not the money.”

Opens

#8830 · 41 · Fresno, CA

Charge nurse · two kids

“$14 is a babysitter hour. I’d need to see it beats the cafeteria on a bad week.”

Closes

Example personas for “meal-prep for night-shift nurses” — synthetic, not real people. Good at direction, weak at exact prices.

WHAT IT IS, AND WHAT IT IS NOT

Synthetic, not real.

The 100 customers are AI personas built on US Census distributions (NVIDIA Nemotron-Personas-USA). They are good at showing the direction of a concept or a preference. They are weak at brand awareness and exact price points, and they are not a substitute for talking to real people.

Counted, not guessed.

Every number you see is a count of individual persona responses (“37 of 100 asked about price”). We never turn that into a market-size prediction.

You decide.

Only the hard questions you mark as “can’t refute” become problems. The AI never grades its own arguments.