First-pass criterion — the idea screen. Per-name questions: thesis, catalyst, kill criteria, sizing math, expected value, asymmetry. This is what ptk-1 through ptk-4 covered. A name that fails this screen never reaches the portfolio-fit conversation.
Second-pass criterion — the portfolio-fit screen. Given the current portfolio (sector exposures, factor tilts, individual position correlations, total drawdown sensitivity), does adding this name at the contemplated size improve the portfolio's expected return relative to its expected risk? The honest version of this question is sometimes uncomfortable, because the answer for a high-conviction name in an over-concentrated bucket may be "smaller size" or "pass."
| Question | Idea screen | Portfolio-fit screen |
|---|---|---|
| What is this asking? | Is this a good investment in isolation? | Does this make MY portfolio better? |
| Inputs | Per-name thesis, catalyst, EV, asymmetry | Current portfolio composition, correlations, factor tilts |
| Output | Pass / interesting / actionable | Add at full size / smaller size / pass on portfolio grounds |
| When it can override the other | Never — failing the idea screen ends the conversation | An interesting idea may still fail this and be passed |
| Most common error | Concluding "interesting" before kill criteria are written | Skipping it entirely — the idea screen does both jobs |
Worked example — Tirebridge Materials (fictional industrial-services name). Idea screen: thesis is margin expansion driven by pricing power, catalyst is Q3 print, EV is $46 vs current $38, asymmetry is 1.6x reward / risk. Idea screen passes — interesting at a 3-4% position. Portfolio-fit screen: the investor's portfolio is already 21% in industrial-services names with similar margin-expansion theses, two of which carry highly correlated downside (same end-market exposure). Adding Tirebridge at 3% pushes that cluster to 24% and adds a third position that draws down with the others in the bear case. Portfolio-fit conclusion: take Tirebridge at 1.5% rather than 3%, OR pass and look for a less-correlated long in consumer staples or healthcare to balance the cluster. Conviction does not justify ignoring the cluster — sizing does.
Correlation is not just sector membership. Two positions in different sectors can correlate through a common driver — interest-rate sensitivity, oil exposure, dollar strength, China revenue, refinancing risk. The portfolio-fit screen is most valuable when it surfaces correlations the sector classification missed. A REIT, a homebuilder, and a regional bank are in three different sectors but correlate strongly through rates; treating them as three independent positions is what "diversification by sector code" gets wrong.
Sit with the ideas.
An investor's portfolio is already 38% energy and industrials. He finds an interesting long on Tirebridge Materials, an industrial-services name with strong margin expansion and a clear catalyst. The thesis is sound on its own merits. What is the most accurate way to think about adding the position?