Prompt guide

    Prompt Guide

    Write clearer prompts for AI clients connected to Populous through the MCP server.

    Read MCP docsSee the current tool workflow behind these prompts.Connect to Populous MCPConnect Populous to Claude or Codex before using the guide.

    This guide is for humans prompting an AI client that has access to the Populous MCP. You do not need to know the tool names. You do need to be clear about the question, the audience, and what material respondents should actually see.

    The basic pattern

    Use this structure for most prompts:

    1. State the decision or question Populous should help answer.
    2. Name the audience in plain English.
    3. Describe what respondents should see or interact with.
    4. Keep strategy notes and constraints in background context.
    5. Ask the AI client to plan the run before launching.
    Use Populous to test [the decision or question].

Audience: [who should be simulated].
Respondents should see or interact with: [ad, page, mockup, URL, document, or nothing].
Background context only: [notes, strategy, constraints, prior assumptions].

Plan the run first and show me the summary before launching.

    A good prompt tells the AI what business question Populous should answer. It does not need to tell the AI which MCP tool to call.

    What to include

    Be specific about the decision you are trying to make.

    Weak:

    Test this landing page.

    Better:

    Use Populous to test whether IT directors understand the value proposition on this landing page and whether they would book a demo.

    Name the audience in plain English.

    Weak:

    Run this on buyers.

    Better:

    Audience: mid-market HR leaders at US companies with 500-5,000 employees who influence employee onboarding software decisions.

    Separate background context from respondent-visible stimulus.

    Background context only: our ICP is HR leaders, the product replaces manual onboarding checklists, and price sensitivity is a concern.

Respondents should see: the attached landing page screenshot.

    Say whether Populous should build fresh resources or reuse something saved.

    Use a fresh audience and experiment unless I explicitly say otherwise.

    Or:

    Reuse my saved "US millennials" population if it exists.

    Saved population reuse is the supported reuse case in the current MCP workflow. Experiments and prior setups are generated from the new prompt; use details from prior work as background context instead of asking the AI to bind a saved experiment.

    If you have files

    If the AI needs local files, screenshots, PDFs, mockups, or ad creatives, ask it to create an upload link before it plans the run.

    I have files for this test. Create a Populous upload link first. I will upload the files there, then you should confirm the files and plan the simulation.

    Put related files in one upload session. Do not ask for one upload link per file unless the files belong to totally separate tasks.

    Tell the AI which uploaded files are for planning and which are for respondents.

    The strategy memo is background context only. The two ad images are what respondents should evaluate.

    Do not paste base64, data: URLs, or raw file contents into the prompt. Use the upload link.

    If you are comparing variants

    Say what is being compared and what outcome matters.

    Use Populous to compare two homepage hero variants for B2B SaaS buyers.

Audience: VP-level operations leaders at US software companies.
Respondents should see: both uploaded hero mockups, treated as competing variants.
Outcome: which version is clearer, more credible, and more likely to generate demo intent.
Plan the run first and show me the summary before launching.

    Do not manually specify the experimental arms unless you have a strong reason. The planner should decide the right structure from your comparison framing and the uploaded stimulus.

    If you are comparing audiences

    List the audiences clearly.

    Use Populous to compare urgency for this product across three audiences:

1. SMB owners with 10-100 employees.
2. Mid-market operations leaders at 500-5,000 employee companies.
3. Enterprise transformation leaders at 10,000+ employee companies.

Question: who feels the most urgent need, and what objections differ by audience?
Use a fresh population for each audience.

    This is different from comparing variants. Audience comparisons ask Populous to simulate multiple populations; variant comparisons ask Populous to compare different things shown to the same or similar audience.

    If the website requires login

    Say that the page is private or logged in before the AI plans the run.

    Use Populous to test whether trial users can find the billing export in our logged-in dashboard.

The site requires login. Start a live sign-in flow first, wait for me to sign in, then use that saved website session when planning the run.

Audience: finance operators at US B2B SaaS companies.
Respondents should interact with: https://app.example.com/dashboard
Outcome: whether they can find the export, where they hesitate, and what wording confuses them.

    Do not just paste a private URL and ask Populous to run it. The AI needs to create and save the login session first, then plan the website test with that session attached.

    If you want to reuse saved resources

    Only ask for reuse when you actually mean it.

    Reuse my saved "US millennials" population for this run.

    Good reuse prompts name the saved thing and explain what should change.

    Reuse my saved "US millennials" population, but test the new pricing copy instead of the prior message.

    If you do not mention reuse, the AI should let Populous build fresh resources from your prompt.

    If you need raw results

    For most runs, ask first for the conclusion, key findings, and what to do next.

    After the run finishes, summarize the verdict, strongest evidence, objections, and recommended next test.

    If you need a spreadsheet-style export, ask for it after the run is complete.

    After the run finishes, give me the summary first. Then create a raw CSV download for the enterprise audience only.

    Use raw exports when you need to inspect rows, compare one audience or variant slice, or move results into another analysis tool.

    For very large runs, ask for a focused slice instead of asking the AI to paste everything into chat.

    After the run finishes, summarize the verdict first. Then inspect the results table for enterprise respondents only and show the strongest objections with supporting rows.

    This lets the AI query the completed run server-side and bring back the rows that matter.

    Good prompt examples

    Landing page test

    Use Populous to test whether finance leaders understand this landing page and would request a demo.

Audience: CFOs and finance operations leaders at US mid-market companies.
Respondents should interact with: https://example.com
Background context only: the product automates monthly reporting and competes with spreadsheet-heavy manual workflows.

Plan the run first and show me the summary before launching.

    Logged-in product flow

    Use Populous to test whether trial users can find and download the billing export in our logged-in dashboard.

The site requires login. Start a live sign-in flow first, wait for me to sign in, then use that saved website session for the run.

Audience: finance operations managers at US B2B SaaS companies.
Respondents should interact with: https://app.example.com/dashboard
Outcome: whether they find the export, where they get stuck, and what labels they misunderstand.

Plan the run first and show me the summary before launching.

    Ad creative test with uploaded files

    I have two ad creatives to test. Create a Populous upload link first.

After I upload them, use Populous to compare which ad is more credible and more likely to drive interest.
Audience: US small business owners who buy payroll or HR software.
Respondents should see: the two uploaded ad images.
Outcome: trust, clarity, and likelihood of clicking.

Plan the run first and show me the summary before launching.

    Strategy memo plus mockup

    Create a Populous upload link. I will upload a strategy memo and a product mockup.

Use the memo as background context only. Respondents should see only the product mockup.

Question: would ecommerce founders understand the product and want to try it?
Audience: founders or heads of growth at US ecommerce brands doing $1M-$20M in annual sales.

Plan the run first and show me the summary before launching.

    Message positioning

    Use Populous to test which positioning angle resonates more with HR leaders:

Angle A: reduces onboarding admin work.
Angle B: improves new-hire retention.

Audience: HR leaders at US companies with 500-5,000 employees.
Outcome: which angle feels more urgent, credible, and differentiated.
Use a fresh audience and experiment.

    Reuse a saved population

    Use Populous to test this new pricing message, but reuse my saved "US millennials" population if it exists.

Question: does the pricing feel fair, and what concerns come up?
Respondents should see: the pricing copy below.

[paste pricing copy]

    Follow up on a prior run

    Use Populous to run a follow-up based on sim_key [paste sim_key].

Question: the prior run suggested buyers were confused about implementation effort. Test a revised explanation that directly addresses implementation time and required integrations.
Audience: use the same audience description as the prior run if you can infer it from the results; otherwise ask me before planning.
Respondents should see: the revised explanation below.

[paste revised explanation]

    Common mistakes to avoid

    • Do not say "attach the image later." Files need to be uploaded before planning.
    • Do not mix background material with what respondents should see.
    • Do not ask the AI to use old tool names.
    • Do not paste raw file bytes or base64 into chat.
    • Do not ask to reuse saved resources unless reuse matters.
    • Do not ask the AI to bind a saved experiment; use prior experiment details as background context for a new plan.
    • Do not ask Populous to test a logged-in page without first telling the AI to create a live sign-in session.
    • Do not ask for results immediately after launch; simulations need to finish first.

    A strong one-message prompt

    Use Populous to answer whether this product concept is compelling enough for a buyer interview.

Audience: heads of customer support at US B2B SaaS companies with 100-1,000 employees.
Respondents should see: the product concept below.
Background context only: we are deciding whether to prioritize this concept for discovery next week.
Outcome: clarity, urgency, credibility, objections, and what questions buyers would ask.

Use fresh Populous resources. Plan the run first and show me the summary before launching.

[paste concept]