Platform
Every ad is the end of a chain: a fact, a claim, an audience, a creative, a placement, a result. Marketing OS makes each link explicit, gated, and auditable — so when something works you know why, and when something is wrong you know where.
The decision layers
Each layer only consumes what the previous one has validated. That ordering is what makes the whole system trustworthy — and what most marketing stacks structurally lack.
Researched market facts with sources and dates
Every fact that could back an ad — prices, regulations, market structure, seasonality — is researched and recorded with its source and date. Facts are scoped to a market, so multiple brands in the same category share what is true.
What the brand is allowed to say
Claims are validated on two independent axes that never collapse: factual state (is it true — research decides) and market state (does it work — experiments decide). Only supported claims can carry advertising.
Personas as decisions, not decorations
Audiences are modelled as a factor tree — occupancy, budget, geography, intent — so a targeting choice is a recorded decision with reasons, not a hunch that lives in someone’s head.
Copy, stills, video and music as gated drafts
The studio drafts headlines, scenes, video takes and music beds against the approved claims and personas. Character limits are validated per network; drafts stay drafts until a human approves.
One pipeline, six networks
Approved content fans out to Facebook, Instagram, X, LinkedIn, YouTube and TikTok with per-network metadata validation. The composed request is stored verbatim before anything is sent — auditable to the byte.
Honest read-back that feeds the next round
Engagement and performance metrics return platform-native. Verdicts require statistical power; small samples say “insufficient evidence” rather than declaring winners. What is learned annotates the claims and creative it came from.
The gates
A rule that lives in a process document is a suggestion. These four live in the system's structure — schema, pipeline and wire checks — where no deadline can talk them out of the way.
No ad angle can exist on an unsupported claim.
Enforced by the database schema itself. An angle built on a rejected claim is refused at write time — there is no code path around it.
Nothing publishes without a recorded human approval.
Client brands additionally require a note recording how sign-off arrived. Editing an approved post drops it back to proposed automatically.
Media with uncleared music cannot leave the system.
Every generated track is born internal-only. Clearing licensing is an explicit operator action with a mandatory note; exports of uncleared audio are refused.
Staged media is verified on the wire before publish.
Files are staged to public URLs and each one is fetched back and checksummed before a network is told to use it. “It should be there” is not a state.
Channels
Accounts connect through each platform's own authorization flow, completed by the account owner — and can be disconnected there at any time.
FAQ
Yes. Accounts connect through each platform’s own authorization flow, completed by the account owner. Access can be revoked at any time in the platform’s settings — tokens are stored encrypted and die with the connection.
The system says so. “Insufficient evidence” is a first-class verdict with its own colour in the cockpit — separate from “biased measurement”, which more data cannot fix. Neither is ever silently rounded to a winner.
No. Your accounts, content and metrics are used to run your marketing, and for nothing else. Nothing is sold, shared or profiled.
