Research How do you measure a theme?

How do you measure a theme?

AI, defense, biotech, robotics — themes are the stories markets tell. Stories can't be charted, so most theme investing runs on vibes. Theme Compass is our attempt to replace the vibes with three measurable things.
Themes August 2026·5 min read
Share

The problem: themes don't have tickers

"AI" is not a stock. It's a supply chain — chipmakers, data centers, power, networking, model labs, software — plus a crowd of companies that put the word in their earnings calls. Buy a theme ETF and you're buying someone's opinion of that supply chain, diluted by index rules. Follow the news and you're measuring hype, not exposure. The first job of Theme Compass is definitional: write down, publicly, which companies constitute each theme and each sub-theme — the AI theme alone splits into nine, from GPUs to power to the application layer — so that every number downstream refers to something inspectable.

Purity: not every "AI stock" is an AI stock

Within each basket, companies aren't equal. A pure-play whose entire business is the theme tells you more about the theme than a conglomerate with one relevant division. So constituents carry a purity tier — pure, core, or merely linked — and basket weights follow purity, with any single name capped so no giant can impersonate an entire theme. When NVIDIA sneezes, the compute sub-theme should move; the AI theme shouldn't catch pneumonia by construction.

Momentum, measured against the world

A theme basket rising 10% means nothing in a market that rose 12%. Every basket is therefore scored on excess return against a global benchmark, blended across one-week to six-month windows, then ranked against every other sub-theme we track. The output is a 0–100 score with a plain meaning: this is how strongly the market is currently paying up for this specific story, relative to all the other stories on offer.

Breadth: the honesty check

A theme where everything is rising is a trend. A theme where one stock is rising is a stock.

Momentum alone is easy to fake — one mega-cap can drag a basket's return while the rest of the theme quietly rolls over. So every score comes with breadth: the share of constituents trading above their own moving averages. The dangerous combination gets its own warning light — narrow leadership, a high momentum score standing on low breadth. That pattern is how theme rallies tend to end: the story is still winning, but fewer and fewer stocks are carrying it.

Stages, with the evidence shown

On top of momentum and breadth sits a rules-based cycle-stage classifier — emerging, advancing, euphoria, digestion — that folds in the same machinery the rest of the site runs, including the theme's own regime state. One design rule is non-negotiable, and it's the same rule the honest-backtest note argues for: the label always ships with the list of conditions that fired. You never get "euphoria" as a verdict; you get "euphoria, because the basket is this far above its 200-day average, three-month returns are this stretched, and news buzz is this many deviations above normal." Labels you can argue with are worth more than labels you have to trust.

Open Theme Compass — every theme, scored and staged →
Research notes discuss ideas, not recommendations. Nothing here is investment advice — see the Disclaimer and Methodology.

Keep reading

RegimeRegime investing: what the risk-averse actually getRiskTurbulence, absorption, and the anatomy of fragilityIndiaOne market, one desk: reading India on a single page