AI dashboard · Facebook Ads
Facebook Ads dashboard, built by AI on your Data Model
Facebook Ads spend, clicks and conversions next to the orders the store actually recorded, campaign by campaign. Claude built it through OWOX MCP, and every number comes from a governed Data Mart, not from the AI's guess.
Three questions, one Data Model
Every view answers one question a marketing or sales lead asks. Between them they read four of the twelve Data Marts in the e-commerce Data Model below, and each view names the ones it uses, so you can open the definition behind any number.
Summary
What does Facebook claim, and what did we get?
Opens with what the numbers mean: what Facebook leaves after ad spend, how far its claims run ahead of the store, and which campaign returns most per $1. Then spend, clicks and claimed conversions against the store's orders, with what Facebook claims per $1 next to what the store recorded.
Claims vs store
Which campaign is over-claiming?
Each campaign's claimed conversions against the store's orders, net sales and profit after ad spend on the same campaign name, with how many conversions it claims per order. The campaigns add up to the Summary.
Delivery
What are we paying for reach?
Impressions, clicks, CTR, CPC and CPM by campaign, and how each campaign's spend moved on last year.
The metrics on this dashboard
Each one is defined once in the Data Model and read the same way in every view. Follow a chip for what the metric means and how to calculate it.
Facebook ROASFacebook CPAClick-through rateCost per clickCost per mille
An AI dashboard is only as true as what it reads
Most AI dashboards start from an exported file. That works until the second export.
On a CSV
The AI guesses what revenue means, every time
Each export is its own version of the truth, so two dashboards give two numbers. The file goes to whoever has it, and it is out of date the moment it is saved.
Which one goes to the board?
On your Data Model
What Facebook claims sits next to what the store recorded
Unified Ad Spend carries the platform's own claim; Sessions and Purchases carry the store's orders on the same campaign name. Neither replaces the other.
Two sources, one campaign name.
Connecting the AI straight to the warehouse is not enough either. A warehouse rarely holds the business context, so the AI still guesses what the numbers mean. The Data Model is where that context lives.
From your warehouse to a dashboard in five steps
Your analyst does step two once. After that, anyone with access asks their AI for the dashboard they need.
- 01Your warehouseBigQuerySnowflakeDatabricksRedshiftAthenaYour data stays where it is. OWOX runs on it and keeps no copy.
- 02Data ModelData MartsjoinsmetricsYour analyst defines once what net sales, an order and a channel mean.
- 03OWOX MCP and APIgovernedpermissionsThe only door the AI gets: the Data Marts it may read, nothing raw.
- 04Claude or ChatGPTyour own AIAsk for the dashboard in plain words. It builds it on the Data Model.
- 05AI dashboardReportevery runKept as a Report, refreshed every time it runs, like Sheets or Excel.
Build this on your own data
Connect your warehouse, describe your business once, and let your AI build the dashboards your team asks for.