The impact layer

News tells you the mood. We tell you the move — per stock.

One story can lift one company and sink another. A single sentiment score can't see that. Here's how MarketDX reads the news differently — and why it matters for anything you follow or hold.

Try it in the playground → See real examples ↓

THE PROBLEM

A news story isn't one thing to everyone in it.

Most news tools score the whole article — one number, one label. It's an average. And an average is wrong for every specific stock the story actually moves. Take one real headline:

NEWSAlcoa to buy South32's aluminium assets for $4.1 billion

ONE HEADLINE · SENTIMENT SAYS “POSITIVE” · +0.68

AA▲ poscapital

Alcoa buys the assets — the deal is accretive to earnings. Good for the buyer.

S32▼ negcapital

South32 sells its bauxite, alumina & aluminium operations. The seller loses them.

Same story. Opposite outcomes. If you hold South32, a “positive” score points you the wrong way.

HOW IT WORKS

So we score the impact per stock.

For every company a story genuinely moves, you get a small, structured object — four fields that turn “the news” into something you can actually act on.

TRADITIONAL — ONE SCORE / ARTICLE

{
  "sentiment": "positive",
  "score": 0.68
}

MARKETDX — ONE OBJECT / STOCK

{
  "ticker": "S32",
  "direction": "neg",    // which way for THIS stock
  "aspect": "capital",  // the channel · the WHY
  "relevance": 1.0,    // how central · 0–1
  "reason": "Divests its aluminium ops"
}

directionwhich way

Up, down, or unclear — for this specific stock. Not a mood for the whole article. The buyer is pos, the seller is neg, a passing mention is ambiguous.

aspectthe channel

Why it moves — the mechanism, not the reaction. One of eight channels: demand, capital, technology, competition, regulation, supply, pricing, geopolitics. “Apple raises prices” is pricing; an analyst upgrade is capital; a drug approval is regulation. The channel tells you if it's durable or a blip.

relevancehow central

How central this stock is to the story — 0 to 1. The subject of the piece scores ~1.0; a name dropped for comparison scores ~0.2. It filters the noise, so a passing mention never looks like a headline event.

reasonin plain English

One line, grounded in the actual article. Not a black box — the specific development that moves this company, in words you can read, quote, or hand to a model.

WHY IT MATTERS

Three things a single score can't do.

Split the winners from the losers

In a merger, a lawsuit, a price war — one side gains, the other loses. Per-stock direction shows both. A blanket score averages them into something true for no one.

Catch the competitor off-screen

An Intel upgrade is quietly negative for TSMC — a rival gaining foundry ground. We surface that competition link even when TSMC never appears in the headline.

Tell a catalyst from a blip

“Stock up” could be real product demand or just a valuation call. The aspect tells them apart — so you know whether it changes the story or just the day.

What you can do with it

Whatever you follow, it answers the question you actually have.

→HOLDYou own a stock. Is today's news good or bad for your position — and why? Answered per name, with the reason.

→HUNTYou're looking for ideas. Filter by direction + channel — “who's winning on demand in AI” — a screener that explains every name it returns.

→BUILDYou're building something. Structured JSON per stock — feed it to a model, an app, or an alert. No opaque score to reverse-engineer.

Stocks, private companies (OpenAI, Anduril), forex and crypto are all first-class — so a story's full reach shows up, not only the names that happen to be listed.