Agree what the numbers mean before you build them.
Meridian turns the business case from Step 1 into a data model and a dictionary your team can defend. Grain, definitions and ownership settled up front, then handed over as a single Blueprint.
Two teams, two numbers, one uncomfortable meeting.
Finance counts revenue on invoice date. Sales counts it on close date. Both are right, both are documented nowhere, and the disagreement only surfaces when the two dashboards are on screen next to each other. The fix is rarely technical. Somebody has to decide what the word means, and then write it down where the next person will find it.
That decision usually gets deferred, because it feels slower than building. So the model gets inferred from whatever tables happened to be available, the grain is never stated out loud, and nobody can say whether a row is one order, one order line, or one shipment. Every number downstream inherits the ambiguity.
A model and a dictionary, drafted from your business case
Settle what the numbers mean, and the shape of the data that produces them, before anyone builds a chart.
A model drawn from your question
Meridian reads your Step 1 canvas and drafts a dimensional model: a fact table with its grain stated in a plain sentence, dimension tables, and typed keys. You start from a proposal, not a blank canvas.
Definitions you can defend
Every field arrives with a proposed definition, a calculation written in plain English rather than SQL, a logical type, and a classification. You accept, reword, or exclude each one.
Governance that stays yours
Steward and retention are typed by you and never invented by the AI. Confirming or excluding a field writes an audit entry, so the record shows who decided what.
A question-fit check
Fields that do not clearly serve the question from your canvas get flagged for a second look. It is advisory. It never blocks you, and it never removes a field on your behalf.
The argument you never have again
Anyone can draw a fact table and some dimensions. The work is agreeing what the numbers inside them mean, and writing that down somewhere the next person will find it. Note which standard cost the margin calculation uses. That is the sentence that stops two teams reporting different numbers.
A retail margin data product, drafted from a business case like yours.
The definitions
a few rows
| Field | Type | Definition | How it’s calculated | Class |
|---|---|---|---|---|
| net_revenue | decimal | Invoiced revenue after discount, excluding tax. | Gross line value less discount, excluding tax and delivery. | confidential |
| gross_margin | decimal | Net revenue less standard cost of the units sold. | Net revenue less standard cost multiplied by units sold. Uses the standard cost in force on the invoice date, not the current one. | confidential |
| markdown_flag | boolean | Whether the line was sold at a marked-down price. | not a measure | internal |
The model it hangs on
One row per order line, per invoice date.
You start from a proposal, not a blank page
Meridian drafts first, then hands you the pen. Every field it proposes is yours to accept, reword, or throw out, and the two things that carry real accountability, the steward and the retention period, are only ever typed by you.
Meridian reads the business case you built in Step 1 and shows you which sections are ready and where the gaps are.
Add any extra context the canvas does not carry, then ask it to draft the data you need.
Work through the model on the diagram: rearrange tables, fix relationships, drop what does not belong.
Work through the dictionary field by field. Confirm or exclude each one, and set the steward and retention yourself.
Download the Blueprint as Markdown or Excel and hand it to the team who owns the estate.
What you walk away with
The dimensional model
A fact table, its stated grain, and the dimensions around it, drawn on a canvas you can rearrange and rewire. Relationships are yours to correct.
The data dictionary
Field, type, definition, source, classification, nullability, keys, steward and retention. The row-by-row record of what your team agreed.
The Blueprint
One handover document in Markdown or Excel, carrying the shape and the definitions together, ready for your data team to map into your own estate.
Meridian is an external mentor, not another pipeline
Meridian never connects to your systems. The Blueprint is a specification your own team implements, which means Step 2 works the same whether your data sits in a warehouse, a lakehouse, or four spreadsheets and a nightly export.
Settle the definitions first
An agreed model takes an afternoon. A dashboard nobody trusts takes a quarter to unpick. Conform the data before you build on it.
Request access