Marketing Mary Platform Help
Marketing Mary (marketingmary.ai) is an AI marketing co-pilot for B2B marketing teams whose
standout module is interactive buyer personas — synthetic personas built from your real customer
data (CRM, Google Analytics 4, email platform, ad accounts) that you converse with to test
subject lines, probe objections, and validate messaging before a campaign ships. Beyond personas it
runs a research → strategise → create → publish → measure → maintain loop and can publish to HubSpot
or WordPress. It maps B2B buying committees (~10 stakeholders across IT/ops/finance/end-users) and
auto-updates personas as new behavioral data flows in.
Its differentiator vs the prompt-only persona generators (InstantPersonas, PersonaGen, Personadeck) is
that personas are grounded in your live data and conversational, not generated from a one-line
description. It's a subscription (Starter/Growth/Agency), UK/EU-based with GDPR built in, and
currently waitlist-stage — treat all features/pricing as best-effort and unconfirmed.
Note: as of research the product is in waitlist/pre-release, so details below are best-effort from the
live site and third-party sources — verify at marketingmary.ai. There is no public developer API, no
webhooks, and no developer MCP server (the "MCP Integration Addendum" in the footer is a legal
data-processing document, not a Marketing Mary MCP server).
Step 1 — Gather context
If references/learnings.md exists, read it first for accumulated platform knowledge.
Ask only what you can't infer:
- What do you want from Marketing Mary?
- A) Build an interactive persona from your real customer data
- B) Talk to a persona to test subject lines / objections / messaging before a campaign
- C) Map a B2B buying committee (multi-stakeholder targeting)
- D) Connect it to HubSpot / Salesforce / GA4, or publish content
- E) Understand the Starter/Growth/Agency tiers (or the waitlist)
- F) Automate or export personas programmatically (API question)
- What data can you connect? Personas are only as good as the CRM/analytics/email/ad data behind
them — a thin or low-traffic account yields a thin persona.
Skip-ahead: if the user wants the validate-before-building method or to compare persona/idea tools
across the market, that's a /sales-idea-validation question — route in Step 2.
Step 2 — Route or answer directly
| If the user's question is about… | Route to |
|---|
| The validate-before-building method, or comparing persona/idea tools across the market | /sales-idea-validation {question} |
| A prompt-only persona generator (persona from a short description, no live data) | /sales-personadeck or /sales-instantpersonas or /sales-personagen {question} |
| Turning persona insights into a content plan across tools | /sales-content {question} |
| Deeper HubSpot-side setup (the CMS/CRM it publishes to) | /sales-hubspot {question} |
| Growing a pre-launch waitlist / capturing real demand | /sales-audience-growth {question} |
When routing, give the exact command: "This is a {domain} question — run: /sales-idea-validation {original question}"
Otherwise, answer Marketing Mary-specific questions using Step 3.
Step 3 — Marketing Mary platform reference
Read references/platform-guide.md for the full reference — the persona-creation phases and required
data sources, how the interactive/conversational persona works, buying-committee mapping, the
research→publish co-pilot modules with their API/webhook/UI-only tags, the Starter/Growth/Agency tiers
and best-effort GBP pricing, the integration/connector surface (HubSpot/Salesforce/GA4 — bidirectional
vs read-only), and the no-public-API automation reality with an LLM-API + HubSpot-API workaround.
Answer using only the relevant section — don't dump the full reference.
Step 4 — Actionable guidance
- A persona you converse with is a model of your DATA, not a real customer — say so every time.
Whatever the user asks, make clear that "talking to" a Marketing Mary persona (testing a subject
line, probing objections) is a synthetic simulation of your existing data, so it's directional
input, not validated demand. It can sound convincing and still be wrong. Keep the objections and
angles the conversation surfaces — that's its real value — but prescribe confirming the go/no-go with
real behavior — a smoke-test click, a reply, an A/B send, a pre-sale — and route the real test to
/sales-idea-validation / /sales-audience-growth.
- Ground the persona in real data — output tracks the data you connect. Its edge over prompt-only
generators is that personas are built from CRM/GA4/email/ad data, so tell the user a thin or
low-traffic account yields a thin persona; connect real sources first, and note personas
auto-update as new data flows in (they're not a one-time doc).
- For a B2B buying committee, build the whole committee, not one persona. Marketing Mary maps ~10
stakeholders (IT, ops, finance, end-users) with interdependencies — advise mapping each role and
differentiating the message per stakeholder rather than targeting a single blended buyer.
- Present pricing as best-effort and waitlist-stage; point to the live pricing page. Name the tiers
— Starter (
£99/mo, 1 user, 2 personas), Growth (£299/mo, 3 users, 10 personas), Agency (~£999/mo,
unlimited users/personas, multi-client + white-label) — but say the product is pre-release, so
confirm current tiers, seat/persona caps, and which integrations are plan-gated at marketingmary.ai
before relying on any figure.
- There is no public developer API or MCP server — don't plan an integration around one. If asked to
automate, export, or batch personas, state plainly there's no documented public API, no webhooks,
and no developer MCP server (the footer's "MCP Integration Addendum" is a legal/data-processing
doc, not an MCP server you can call), and note it's pre-release/waitlist-stage so there's no
supported export endpoint today. It's a UI tool with native connectors (HubSpot, Salesforce,
GA4, email, ads). Workarounds: read/write the synced data through the HubSpot/Salesforce API, or
generate personas programmatically by calling an LLM API directly (see
references/platform-guide.md).
- It's a co-pilot, not just personas — but publishing to HubSpot/WordPress isn't demand either. The
research/create/publish/measure loop speeds execution; a campaign informed by a persona is
still not evidence anyone will buy. Keep the go/no-go on real behavior.
If you discover a gotcha or tip not in references/learnings.md, append it there with today's date.
Gotchas
Best-effort from research (2026-07) — the product is in waitlist/pre-release, so features, tiers, GBP
pricing, and the connector list are unconfirmed and move; verify at marketingmary.ai.
- "Talking to a persona" feels like validation but isn't. The persona is a synthetic model of your
existing data — pressure-testing a subject line against it is a directional gut check, not a stranger
taking an action. Confirm with a real demand test.
- Persona quality tracks the data you connect. A low-traffic or sparse CRM/GA4 account produces a
generic persona (competitors like Delve AI effectively need thousands of monthly sessions). Connect
real, rich sources first.
- It's waitlist-stage. Pricing, seat/persona caps, and which integrations exist may differ from any
marketing copy — confirm live before committing.
- No public API / webhooks / developer MCP server. Don't build a pipeline around it; the "MCP
Integration Addendum" is a legal document, not a callable server. Automate via the HubSpot/Salesforce
API on the synced data, or an LLM API for persona text.
- Personas auto-update. Unlike a static PDF persona, these change as data flows in — a message you
validated last quarter may map to a shifted persona; re-check before reusing.
- B2B committee ≠ single buyer. Targeting one persona in a ~10-stakeholder committee misses the
approvers/blockers — map and message the whole committee.
Related skills
/sales-idea-validation — The tool-agnostic validate-before-building method + the full persona/validator tool landscape (use this to actually decide build-or-not; a persona you talk to is not demand)/sales-personadeck — Personadeck platform help (a prompt-only persona generator — a styled persona card from a short description, User vs Negative type, PDF-only; UI-only/no API)/sales-instantpersonas — InstantPersonas / User Persona platform help (a prompt-only buyer/user-persona generator + website-perception "Insights" + competitor personas; UI-only/no API)/sales-personagen — PersonaGen platform help (a prompt-only persona generator — detailed personas + content-generation prompts, organized into projects; UI-only/no API)/sales-hubspot — HubSpot platform help (the CRM/CMS Marketing Mary syncs to and publishes into; use for HubSpot-side API/automation)/sales-content — Content-marketing strategy across tools (turn persona insights into a content plan)/sales-audience-growth — Grow a pre-launch waitlist and capture real demand/sales-do — Not sure which skill to use? The router matches any sales objective to the right skill. Install: npx skills add sales-skills/sales --skill sales-do -a claude-code
Examples
Example 1: "Can I test my cold-email subject lines by talking to a persona?"
User says: "Marketing Mary lets me chat with a buyer persona — if it likes my subject line, is that a green light to send?"
Skill does: Explains that the persona is a synthetic model built from your CRM/GA4/email data, so
its reaction is a directional gut check, not demand — it can approve a line real recipients would
ignore. Recommends using the conversation to surface objections and angles (its real value), then
validating the winner with real behavior — an A/B send, reply rate, or a smoke-test click — and
points the real test to /sales-idea-validation and /sales-audience-growth. Notes personas
auto-update, so re-check before reusing an old winner.
Result: The founder mines the persona for angles but takes the go decision from real recipients.
Example 2: Grounding personas in real data + mapping a committee
User says: "We sell to a buying committee. How do I get Marketing Mary to build something useful, not a generic persona?"
Skill does: Says persona quality tracks the data connected — wire up the real CRM, GA4, email,
and ad accounts first (a low-traffic account yields a thin persona), and let personas auto-update.
Explains Marketing Mary maps ~10-stakeholder B2B buying committees (IT/ops/finance/end-users) and
advises building the whole committee and differentiating the message per role rather than one blended
buyer. Frames each persona as a hypothesis to confirm against real stakeholder conversations.
Result: A committee map with role-specific messaging, grounded in live data instead of a one-liner.
Example 3: Automating persona export via API (developer/automation)
User says: "I want to pull our Marketing Mary personas into our data warehouse automatically — what's the API?"
Skill does: States plainly there's no documented public API, no webhooks, and no developer MCP
server (the footer's "MCP Integration Addendum" is a legal data-processing doc, not a callable
server), and that the product is waitlist-stage — so there's no supported export endpoint today.
Offers the real path: since personas are bidirectionally synced with HubSpot/Salesforce, read the
persona/contact data through the HubSpot or Salesforce API into the warehouse; and for programmatic
persona generation, call an LLM API directly (see references/platform-guide.md for a cURL/Python
sketch). Notes the thing worth automating is a real demand signal, not a synthetic persona.
Result: The user avoids building on a non-existent API and pipelines via the connected CRM instead.
Troubleshooting
"The persona agreed with my messaging — does that mean it'll work?"
Symptom: A Marketing Mary persona reacts positively to a subject line / pitch and the user reads it as validation.
Cause: The persona is a synthetic simulation of your existing data, not a real prospect — it
pattern-matches to plausible agreement and can be confidently wrong.
Solution: Keep the useful parts (objections surfaced, angles to try), discard the "verdict," and earn
a real one — an A/B send, reply/click data, or a pre-sale. Cross-check with a real demand test via
/sales-idea-validation.
Persona feels generic despite connecting my data
Symptom: The generated persona is vague or could describe anyone, even after linking accounts.
Cause: The connected sources are thin or low-traffic — data-grounded personas need real volume
(analytics-connected tools broadly need thousands of monthly sessions to differentiate).
Solution: Connect richer, higher-volume sources (full CRM history, GA4 with real traffic, email
engagement, ad data), and let personas auto-update as data accumulates. Add more first-party signal
before trusting the profile.
"Where's the API / how do I automate or export personas?"
Symptom: Looking for API docs, a webhook, or an MCP server to export or batch personas.
Cause: Marketing Mary has no public API, no webhooks, and no developer MCP server (the footer
"MCP Integration Addendum" is a legal doc), and it's pre-release.
Solution: Don't build a pipeline around it. Read the persona/contact data via the HubSpot or
Salesforce API it syncs to, or call an LLM API directly for programmatic persona generation (see
references/platform-guide.md). Automate the real demand signal instead — see /sales-idea-validation.
