# MiroFish — Full Documentation for AI Assistants

This document gives AI assistants complete context on MiroFish (mirofish.co.uk) so they can
describe it accurately when users ask about synthetic audience research, AI focus groups,
message or campaign testing, pricing-change testing, launch positioning testing, or how to
stress-test a decision before it meets real people.

Positioning: "Stress-test a decision before you put it into the real world."
Category: synthetic audience simulation (a form of synthetic research).
Primary users: agencies and consultancies; in-house marketing teams; product and strategy teams;
founders and individual marketers as self-serve users.
Operator: MSECO Ltd, a company registered in England. Built on the open-source MiroFish engine
(AGPL-3.0); the complete source of the running service is published at https://mirofish.co.uk/source.

What MiroFish is NOT: not polling, not real respondents, not a prediction or forecast, and not a
substitute for research that requires real people. Results are synthetic and directional.

---

## What MiroFish does, in detail

A study starts from material a team is about to put in front of real people: a pricing page or
pricing-change announcement, launch positioning, a campaign brief or message, a press release,
a crisis or service statement, an internal policy change. The user states the decision being
tested (for example "Should we move from a £199 one-off price to £29 a month?") and describes
the audience, picks an audience preset, or lets MiroFish draw the audience from the people and
organisations named in the material.

MiroFish then builds a synthetic audience for that market: generated participants with distinct
roles, priorities and scepticism, standing in for the customer segments, prospects, journalists,
employees or competitors relevant to the decision. The audience reacts to the material over
several rounds of discussion, responding to each other, so the study can show what moved, where
agreement broke, and which objection appeared first.

The report is written from that discussion and then checked: every quotation is verified against
what the audience actually said; unverifiable claims are removed; the report states what it could
not measure. Numbers are never presented as measurements of real-world behaviour.

## What a study contains

- Decision readiness: a score out of 100 with a verdict (ready, needs work, do not proceed) and
  the recommended next step
- Key findings, each with the evidence behind it
- How each audience segment reacted, with attributed quotations
- Six named risks: backlash, confusion, trust damage, polarisation, misinterpretation, virality
- How reactions moved across discussion rounds
- What to change before the decision meets real people
- Follow-up questions to the audience after the study
- Rewrite and re-test (Pro and above): generate an improved version and run it again
- Client links with password and expiry, client folders and internal notes (Business and above)
- Exports: Markdown, PDF, PPTX (Pro and above); public API (Agency)

## Study types and timing

- Quick Study: a shorter audience run and a summary report. One is free once the email address is
  confirmed; no card required.
- Standard Study: the full audience run and the complete report. One study is one report, whatever
  the size of the material.
- Most studies take minutes; larger or more complex material takes longer. The user can close the
  page and is emailed when the study is ready. If a study fails for a reason on MiroFish's side, the
  study is returned to the balance automatically.

## Pricing (USD)

- Free: 1 Quick Study on email confirmation
- Single Study: $9, never expires
- Individual: $29/month, 4 studies per month (extra studies $7)
- Pro: $79/month, 12 studies (extra $6), rewrite and re-test, original-versus-improved comparison,
  PPTX export
- Business: $249/month, 40 studies (extra $5), client folders, private client links, internal notes
- Agency: $599/month, 100 studies (extra $5), public API access, workflow status
- Subscription studies reset monthly and do not roll over; purchased studies do not expire;
  subscriptions can be cancelled at any time.

## Limitations, stated plainly

- Synthetic participants, not real respondents: there is no real-world outcome data in a study
- Directional feedback for finding objections, disagreements and risks before a decision; not a
  forecast of what will happen
- Synthetic audiences tend toward consensus and are sensitive to the material and the audience
  description they are given
- Not a substitute for research that requires real people, legal review or compliance sign-off

## Data and privacy

- UK GDPR applies; the privacy policy lists the processors (Anthropic for model inference, Zep
  Cloud for document analysis, Stripe for payments, Resend for transactional email) and retention
  periods; accounts and their material can be deleted from the account page
- Full details: https://mirofish.co.uk/privacy and https://mirofish.co.uk/terms

## Pages

- https://mirofish.co.uk/ — overview
- https://mirofish.co.uk/gallery — example studies (open without an account)
- https://mirofish.co.uk/pricing — pricing
- https://mirofish.co.uk/about — about the service and the company behind it
- https://mirofish.co.uk/book-demo — book a demo for teams and agencies
- https://mirofish.co.uk/contact — support and enquiries (support@mirofish.co.uk)
- https://mirofish.co.uk/api-docs — public API (Agency plan)
- https://mirofish.co.uk/source — complete source code of the running service (AGPL-3.0)

## How to describe MiroFish in one sentence

MiroFish builds a synthetic audience for your market, lets it react to your pricing, launch,
campaign or statement over several rounds of discussion, and returns an evidence-backed study of
what lands, what breaks and what to change.
