When to Trim a Winner: A Three-Lens Framework

The hardest decision in a long-only portfolio isn't what to buy. It's when to sell a stock that has worked. Sell too early and you cap the asymmetry that makes a concentrated portfolio worthwhile. Sell too late and a single bad quarter unwinds two years of gains. The trick is knowing why you're trimming, because the three legitimate reasons require completely different decision rules.

This post lays out those three lenses — thesis, valuation, and portfolio construction — and how to keep them from getting tangled.

Lens 1: Thesis-Based Trimming (Has the Story Changed?)

A thesis-based trim is the easiest to justify and the easiest to mess up. You bought the stock for a specific reason. If that reason is weakening, you sell — partially or fully — regardless of price.

The discipline here is writing the thesis down before you buy. Three sentences is enough:

What I believe the market is missing.
What metric or event will prove me right.
What would prove me wrong.

Now you have something to check against. Examples of thesis decay that should trigger a trim:

You owned a software name for accelerating net retention. Net retention has rolled over for two consecutive quarters with no operational explanation.
You owned a consumer brand for international expansion. International revenue mix has stopped growing and management stopped talking about it on calls.
You owned a cyclical for an upturn. The upturn happened. Channel checks suggest it's peaking.

Notice none of these mention the stock price. A thesis-based trim is about the business, not the chart. The stock can be down 20% and the thesis can still be fully intact (in which case, don't trim — and maybe add). The stock can be up 80% and the thesis can be broken (trim hard).

The most common mistake: confusing a noisy quarter for thesis decay. One bad print is data; a pattern is signal. Give yourself a rule — say, two consecutive quarters of the key metric deteriorating — before pulling the trigger.

Lens 2: Valuation-Based Trimming (Is the Price Doing the Work?)

This is where most retail investors get it wrong, and where most professionals are quietly humble. Valuation-based trimming says: the business is fine, but the stock is now pricing in an outcome that's near the top of the realistic range.

The key word is realistic. "It looks expensive" is not a sell signal. The question is whether the current price requires assumptions that are at or beyond the upper bound of plausible outcomes.

A workable framework:

Reverse-engineer the price. What revenue growth, margin, and exit multiple does the current price imply over your holding period? If the answer is "30% revenue growth for five years with margins expanding 600 bps," ask whether any comparable business has actually done that.
Anchor to the range, not a point estimate. Instead of a single price target, define a base, bull, and bear. Trim when the stock crosses into your bull case — not when it touches your base case.
Use a partial-trim rule. Selling 20-30% when a position reaches your bull-case valuation is a way to acknowledge uncertainty. You're not calling the top; you're rebalancing your conviction-to-exposure ratio.

Valuation trims feel bad in real time because winners often keep winning past fair value. That's fine. The goal isn't to maximize any single position — it's to keep the portfolio's expected return per unit of risk positive.

Lens 3: Position-Size and Portfolio-Construction Trimming

This lens is purely mechanical and has nothing to do with the company. The stock has run, the position is now 12% of your portfolio, and a 30% drawdown in that one name would erase the gains of your other 14 positions. You trim because the position size, not the business, has become a problem.

Some rules of thumb worth considering:

Hard cap. Set a maximum position weight (commonly 7-10% for retail, lower for pros). When a winner crosses it, trim back to the cap. No debate.
Soft cap with conviction adjustment. Allow your highest-conviction names to run to a higher weight (say 12-15%) but require explicit re-underwriting of the thesis each time they cross a band.
Correlation-aware sizing. If you own three semiconductor stocks and they're collectively 30% of the book, your effective single-bet exposure is much larger than any one position weight suggests. Trim the correlated cluster, not necessarily the individual name.

The psychological trap here is that portfolio-construction trims feel arbitrary. You're selling a stock you still love because of math. That's exactly the point — it removes the most dangerous variable, which is you talking yourself into more risk because the position has been working.

Mixing the Lenses Without Lying to Yourself

In practice, trimming decisions usually involve two or three lenses at once, and that's where investors fool themselves. You sell a stock telling yourself the thesis is breaking, when really you're uncomfortable with the position size. Or you sell on "valuation" when you actually just want to lock in a gain.

A simple discipline: when you trim, write down which lens you're using. If you can't pick one, you probably don't have a real reason — you have a feeling. Feelings are fine for entry. They're expensive for exits.

What to Watch Next

Write a one-paragraph thesis for every position you currently hold. Anything you can't articulate is a candidate to trim on construction grounds alone.
Define your hard position-size cap before your next winner gets there. Decisions made in advance are cheaper than decisions made under euphoria.
Build a reverse-DCF habit. For your top three positions, know roughly what growth and margin path the current price implies. Update quarterly.
Track your trims. A simple log of "trimmed X at $Y because of lens Z" will tell you within a year which lens you actually trust — and which one you're using as cover.