Skip to main content
Product5 min read

Same Number, Every Surface: TradingFlow's Governed Metrics Layer

Written by

The AI assistant, Cookbooks reports, MCP, and TradingFlow's own pages all compute Net DEX, DEI, and sentiment from one small, governed registry — not four separate implementations that could quietly disagree. Here's how, and why it's deliberately small.

Here's a failure mode most data platforms have and never talk about: the dashboard says a stock's flow is bullish, the chatbot says it's neutral, and an exported report says something in between — because three different pieces of code computed "sentiment" three slightly different ways. Nobody notices until someone compares them side by side. Once they do, they stop trusting all three, not just the wrong one.

TradingFlow's answer to that failure mode isn't a bigger team of people manually keeping numbers in sync. It's a small, deliberately narrow piece of infrastructure: one governed metrics registry that every surface — the app's own pages, TradingFlow AI, Cookbooks reports, and MCP — reads from instead of each computing its own version.

Rank Symbols table showing Net DEX, Net DEI, Bullish DEI, Bearish DEI, and Sentiment columns for five names

Net DEX, Net DEI, and Sentiment on Rank Symbols aren't three separate calculations happening to agree. They're the same governed definitions the rest of this piece describes.

What "governed" actually means here

A metric in this registry isn't just a formula. Each one is a small bundle of facts, defined once:

  • What it's called and what question it answers — Net DEX answers "which way did directional flow lean," DEI answers "how big was that relative to the name's normal size."
  • Where it comes from — which underlying data it reads, and on what schedule that data updates.
  • How it's signed and scaled — so "positive" means the same thing everywhere it appears.
  • Who's allowed to see it — a metric that's paid-only stays paid-only no matter which surface asks for it.

That bundling matters more than it sounds like it should. It's the reason you'll see the same freshness caveat next to GEX everywhere it shows up on TradingFlow — not because someone remembered to retype it in four places, but because the caveat is attached to the metric's definition itself, not to any one screen showing it.

One registry, four consumers

Diagram: a governed metrics registry compiles once into SQL, consumed identically by app pages, the AI assistant's get_metric tool, MCP's semantic_query tool, and Cookbooks recipe blocks

Four very different surfaces, reading from the same place. None of them can write back to it — the registry only ever answers questions, it never changes what happened in the market or in your account.

Concretely, four things draw from the identical registry:

  • App pagesRank and Option Trades compute their columns from it directly.
  • TradingFlow AI — when you ask the assistant a data question, it doesn't write its own SQL. It calls a tool that composes an answer from the same named metrics, so its numbers can't quietly drift from what the page next to it shows.
  • Cookbooks — a recipe like Daily Market Recap is built from named queries against this same catalog. See the day's premium and sentiment split computed this way in a real run below.
  • MCP — for readers who connect their own AI tools to TradingFlow's data, the same governed catalog is what those tools query too. Nothing external gets a side door to different numbers than the app shows.

Daily Market Recap header showing session KPIs: total premium, call premium, put premium, bullish premium, bearish premium

The premium and sentiment split at the top of a Daily Market Recap run — the same governed definitions as the Rank Symbols columns above, just aggregated to the whole session instead of one name.

Why the registry stays small on purpose

It would be easy to read "governed metrics layer" and picture something sprawling — a full data catalog, a graph of every entity in the business and how they relate. TradingFlow's is deliberately not that. It covers a short, closed list of metrics (premium, net premium, Net DEX, DEI, sentiment, and a few trade-level counts), and it does exactly one job: compile a question into a single, safe, read-only query.

Two restrictions are the actual point, not limitations to apologize for:

  • It only reads. Nothing in this layer — not the AI assistant, not MCP, not a Cookbooks recipe — can use it to change data, place a trade, or alter your account. It answers questions; it never acts.
  • The model can't write its own SQL against your data. When TradingFlow AI or an MCP client asks a question, it's choosing from named, pre-approved metrics and filters — not composing arbitrary queries. That's a narrower capability than a general text-to-SQL tool, deliberately: narrower is what makes "the numbers can't drift" a claim TradingFlow can actually stand behind, instead of a hope.

A bigger, more general system would have been easy to build and harder to trust. This one is small enough that "does the assistant's number match the page" is a question with a mechanical answer, not a QA process.

What this means for you, practically

You don't need to think about any of this to use TradingFlow — that's the point. But it does mean a few things are true whenever you're comparing evidence across surfaces:

If you see this metric here......it means the same thing here
Net DEX on Rank SymbolsNet DEX in a Cookbooks report's exposure chapter
Sentiment on Rank SymbolsSentiment in TradingFlow AI's answer to "what's the sentiment on this name"
DEI on Rank ContractsDEI pulled through MCP by your own connected tool

If two of those ever disagree, that's a bug worth reporting — not an inherent feature of comparing a dashboard to a chatbot. That's the bar the registry is built to hold.

Read the concepts behind the metrics

None of this is investment advice, and consistency across surfaces isn't the same thing as a trading signal — it just means you can trust that when TradingFlow shows you the same number twice, it's actually the same number.

Ready to see it for yourself? Open TradingFlow.

Related Articles

Product

Daily Market Recap: Read a Full Options Session in TradingFlow Cookbooks

TradingFlow's Cookbooks turn a live options-flow session into a guided report. Daily Market Recap walks eight chapters — tone, money flow, timing, DEX/DEI, volatility, GEX, movers, and a headline name — each with real evidence next to the explanation, not baked-in prose.

Product

Options Chart That Matters: GEX, Walls, and Flow—Not a Pretty Candlestick

Looking for an options chart? Map gamma exposure, call/put walls, and same-session option flow in TradingFlow Rank—not just another price chart.

Product

Why Alerts Alone Fail: Rank → Tape → Structure

Reddit tool threads and X promo screenshots often stop at alerts. TradingFlow’s differentiator is a linked rank → tape → structure loop with labeled data horizons.