Factors
Why factors matter
Overview
Theme returns are never “pure narrative.” When a theme jumps or sells off, part of that move is usually shared with the broad market, growth vs value, rates, credit risk appetite, or other common drivers. Factor scores help separate those shared forces from behavior that looks more specific to the basket—so you can read a move with more context.
On stockthemes.ai, every theme gets a compact Factor Profile. The Factors tool lets you compare themes, inspect rankings, and chart factor spreads over time.
Why this matters for themes
A theme can look like it “worked” when the move had little to do with that basket’s thesis and almost everything to do with a shared factor pull—AI / innovation, growth, market beta, and so on. Same calendar window, different driver: that is what factor scores are trying to surface.
Factors don’t replace reading constituents, filings, or news. They answer a narrower question: what kind of risk has this theme been behaving like?
- Theme +22.6%
- MTUM +23.6%
- After momentum -3.1%
Space Supply Chain: Space & Satellite Components from the start of 2026: rode momentum up into the mid-June peak, then gave most of it back in the summer MTUM selloff—same path up and down. After subtracting momentum daily, only -3.1% is left (SPY was quieter). Looks like a space-components story; most of the move was the factor.
All series rebased to 100 at the window start. “After momentum” compounds (theme daily return − MTUM daily return). Teaching illustration, not a tradable hedge.
What we measure
Each non-sector factor is built from liquid ETF proxies (or a long/short ETF spread). For example, market beta tracks SPY; momentum compares MTUM vs SPY; growth/duration compares IWF vs IWD; AI / innovation compares AIQ vs SPY. Sector factors use sector ETFs with the market effect residualized out.
We then score how each theme’s daily returns line up with those factor series. Scores are mapped to a 0–100 scale for readability. Higher means stronger positive exposure to that factor; lower means weaker (or inverse) exposure.
Themes are also ranked against the full theme universe for each factor, so “82” comes with context like “#28 of ~1,300 themes,” not just an absolute number.
Co-movement vs incremental
Co-movement (standalone) asks: how closely does this theme track the factor on its own? It’s the simpler “does it move with AIQ / MTUM / SPY?” read, and it matches the factor compare charts on the Factors page.
Incremental asks: after we account for overlap with the market, growth, sectors, and other factors in the model, how much exposure is still left? Themes often co-move with several related factors at once; incremental scores reduce that double-counting.
On theme pages, the Higher / Lower Factor Profile columns are ordered by incremental exposure (and show those scores). The Factors tool can switch between co-movement and incremental rankings when you want both views.
Example: same theme, two score types
Illustrative numbers — co-movement can look “high” while incremental is only mid-range once overlapping drivers are removed.
AI / Innovation
Momentum
Quality / Defensive
AI / Innovation still co-moves a lot, but much of that is shared with market/growth—so incremental drops. Momentum stays high on both. Quality stays low on both. Theme-page Higher / Lower follows the incremental column.
Higher vs lower exposure on theme pages
Higher exposure lists the theme’s strongest incremental factor tilts—the drivers that still stand out after controls. Lower exposure lists the weakest incremental tilts—factors the theme is less tied to, or that it has tended to move against.
A factor can co-move strongly with a theme and still land in Lower exposure if most of that co-movement is already explained by market, growth, or other overlapping factors. That’s intentional: the card is answering “what’s distinctive,” not “what correlates in a vacuum.”
Closest sector signal
Themes often behave like an industry even when they aren’t a single GICS sector. Closest sector signal picks the sector factor the theme tracks most closely (after removing broad market overlap), with a score and universe rank.
Wording scales with how extreme that fit is—from a modest match, to a very tight match, to the clearest sector fit in the universe when a theme is #1.
- Theme +24.4%
- Financials +16.4%
- Technology +31.7%
Seat-based Pricing is a software basket—easy to peg as Technology. Over this window Tech ran ahead while the theme stayed closer to Financials (XLF). Closest sector signal agrees: Financials, score 64, #30—after market overlap is removed.
All series rebased to 100 at the window start (theme vs XLF vs XLK). Teaching illustration of co-movement; live closest-sector scores use sector residuals vs the market.
Live signal on Software: Seat-based Pricing: Financials · score 64 · #30.
Factor families we publish
Market & style: Market beta, momentum, small-cap, growth/duration, value/cyclical, quality/defensive, speculative beta, unprofitable growth.
Macro & rates: Falling-rate sensitivity, credit risk-on, inflation / commodity, strong dollar, oil sensitivity.
Narrative / risk appetite: AI / innovation, crypto, retail speculation, meme sensitivity.
Sectors: Industry ETF residuals vs the market (used for closest-sector signal and sector rankings).
Plain-language definitions for each factor also appear as tooltips on Factor Profile rows and in the Factors UI.
How to use this
Start on a theme page: glance at Higher / Lower and the closest sector line before diving into constituents. Ask whether today’s move looks like broad beta, a style factor, or something closer to the theme’s own story.
When comparing several themes, open the Factors page—leaderboards, makeup compare, and factor spread charts are built for that workflow.
Treat scores as a lens, not a ranking of “good” vs “bad.” High speculative beta can be exactly what you expect for one theme and a warning flag for another.
Limitations
Scores are statistical estimates from recent daily returns. They can be noisy for young themes, thin baskets, or periods when factor ETFs themselves are unstable. Confidence can be low when the fit is weak—those profiles may be hidden or flagged.
ETF proxies are imperfect stand-ins for economic ideas (especially narrative factors like AI or meme risk). Classification of themes remains judgmental; factors don’t prove what a company “is.”
Nothing here is investment advice or a forecast. Use factors alongside primary filings, earnings, and your own process. See also methodology limitations.