Portfolio concentration risk: how much is too much?
There is no single safe percentage. What counts as concentrated depends on how many holdings you own: in a book of 10, the Clavix reference bands put an ordinary largest position between 21% and 30%, with the top tenth of books starting at 35%. In a book of 3, 33% is arithmetic, not a decision.
What concentration risk is
Concentration risk is the risk that one company, or one sector, decides the outcome of your whole portfolio. It is not the risk that a name falls. It is the risk that when a name falls, nothing else you own is large enough to matter. Diversification is not a moral position on investing. It is arithmetic about how much of your result any one event can claim.
What makes it hard to see is that it lives in the weights, not in the names. A portfolio of eight excellent companies where one of them is 41% of the money is more concentrated than a portfolio of eight ordinary ones at even weight. The quality of the business never enters the measure. Neither does whether the position got large because you sized it that way on day one or because it grew into the number while you were not looking.
It also compounds with every other exposure in the book. A large position in one name is usually a large position in that name's sector, in its customers' cycle, and in whatever macro force moves it, so the same 41% can be one bet or four depending on what sits next to it. Which means the question worth asking about your own book is not whether the biggest name is a good company. It is what share of your money that name represents, and whether the share is larger than a portfolio of your size would carry by accident.
How to measure it
The measure
Three numbers cover most of it, and all three are division.
- Largest position. The market value of your biggest holding divided by the total value of the portfolio, times 100.
- Top three combined. The same calculation for your three biggest holdings, summed. This catches the book that looks fine on its single largest name and is still only three decisions wide.
- Largest sector. Group every holding by sector, sum the weights inside each group, and take the biggest group. This is the one that catches four separate positions that all rise and fall for the same reason.
Two details matter more than they look. First, every weight is by value, never by count. Five hundred shares of an $8 stock is a $4,000 position, and twenty shares of a $600 stock is a $12,000 position, so the second one is three times the exposure of the first. Counting positions instead of weighing them is the most common way people misread their own concentration.
Second, funds and unclassified names are left out of the sector measure in the Clavix model, and left out in both directions: a fund is not a sector you are concentrated in, and it is not a sector you are spread across either. Folding it into a sector bucket would flatter a book that has not earned it.
A worked example
Take a portfolio of 8 holdings worth $250,000, labelled A through H so that nothing here reads as a comment on a real company.
- A: $102,500, which is 41.0% of the book
- B: $37,500, which is 15.0%
- C: $30,000, which is 12.0%
- D through H: $16,000 each, which is 6.4% each
Largest position is 102,500 divided by 250,000, so 41.0%. Top three combined is 41.0 plus 15.0 plus 12.0, so 68.0%. Now suppose A, F and G are all in the same sector: largest sector is 41.0 plus 6.4 plus 6.4, so 53.8%.
Read against the 8 holding row in the table below, that one portfolio gets three different verdicts. The largest position at 41.0% clears the limit of 40.9%, so it is in the top tenth of 8 holding books for single-name concentration. The top three at 68.0% sits inside the ordinary band of 60.8% to 71.3%, so on that measure the book is unremarkable. The largest sector at 53.8% is above the band of 33.0% to 48.1% but below the limit of 57.3%, so it is elevated without being extreme.
That spread is the point. A book is not simply concentrated or not. It is concentrated on a specific measure, and the three disagree often enough that reading only the first one is how the sector problem stays invisible.
How much is too much, by holding count
Almost every rule of thumb on the internet is a fixed number: no more than 5% in one stock, or 10%, or 20%. A fixed number cannot be right, because it ignores the only input that determines what is even achievable. A portfolio of 3 holdings has a largest position of at least 33.3% and a top three of exactly 100%, before its owner has made a single decision. A 10% rule would tell that person they had failed at arithmetic.
So the threshold has to move with the holding count. The table below is the reference Clavix measures against. For each holding count it gives an ordinary band, meaning the middle half of the reference books, and a limit, meaning where the top tenth begins. If you want the risk half of the picture on the same list of holdings, the free portfolio grader returns a value-weighted A+ to F risk grade with no account and no email.
Reference bands by holding count
Each cell shows the ordinary band on the first line and, underneath, where the top tenth of books begins.
| Holdings | Largest position | Top three | Largest sector |
|---|---|---|---|
| 2 | 60.5–82.8over 91.4 | n/a | 61.9–88.9over 100.0 |
| 3 | 47.4–65.5over 75.3 | 100 fixed | 50.9–73.5over 85.9 |
| 5 | 34.0–47.5over 55.7 | 79.2–88.9over 92.5 | 40.0–58.5over 70.0 |
| 8 | 24.9–34.6over 40.9 | 60.8–71.3over 76.4 | 33.0–48.1over 57.3 |
| 10 | 21.4–29.6over 34.9 | 52.9–63.0over 67.9 | 30.3–44.1over 52.6 |
| 15 | 16.0–22.1over 26.0 | 40.6–49.2over 53.5 | 26.6–38.1over 45.3 |
| 20 | 12.9–17.6over 20.6 | 33.4–40.5over 44.2 | 24.7–34.8over 41.1 |
| 30 | 9.5–12.8over 15.1 | 25.0–30.3over 33.2 | 22.6–31.0over 36.2 |
All figures are percentages of portfolio value, from the Clavix concentration model, version concentration-bands-v1. There was no real cross-section large enough to cut percentiles from, so the reference population is simulated from real inputs: 20,000 portfolios per holding count, holdings drawn without replacement from the 1,000 classified single-stock tickers and carrying their real sector, with weights drawn Dirichlet at alpha 1.397, fitted to the 33 real weight vectors available (30 ETF holdings tables plus 3 portfolios of three or more positions). The band is the 25th to 75th percentile and the limit is the 90th. Counts below 2 and above 30 read the nearest end row, because past 30 the curves are flat to within a percentage point. At 2 holdings the largest-sector limit reads 100.0, which nothing can exceed: that cell is a fact about owning two names, not a threshold.
The caption that matters most is the one about what the reference is. Those simulated portfolios were assembled without regard to sector. So the band is not what other Clavix users look like, and it must never be read that way. It is what a portfolio of your size looks like when nobody was concentrating on purpose. Sitting above the band means you are more concentrated than chance. That is a genuinely useful finding, and it is a different sentence from being more concentrated than other investors, which these numbers cannot tell you.
A fourth measure in the same model runs the other way. Sectors represented counts how many named sectors your portfolio touches, and there more is better, so its limit is a floor read from the 10th percentile rather than a ceiling read from the 90th. A 20 holding portfolio assembled by chance touches 8 or 9 named sectors, and the floor sits at 7.
Why concentration hides in plain sight
Nobody decides to put 41% of their money in one company. It arrives by drift. You size a position at 8% because that felt right, the company does well, it triples while the rest of the book goes sideways, and now it is 20% and climbing. Every step of that was success. At no point was there a moment that felt like a decision, which is exactly why the weight never got re-examined. Concentration is usually the shape a portfolio takes when it has been right.
Employer stock is the same mechanism with a schedule attached. Grants vest, a purchase plan runs on a calendar, and the position grows on a timetable nobody set position by position. It also carries an exposure the percentage does not show: if the business struggles, the value of the shares and the security of the paycheck move together. The weight in the portfolio is the visible half of that exposure.
Then there is the counting problem. People count positions and feel diversified: twelve tickers sounds like twelve bets. But twelve positions where one is 40% and the rest split the remainder is closer to one bet with eleven rounding errors attached. A brokerage statement is a list, and a list gives every row the same visual space, so 40% and 4% end up looking like the same kind of thing.
Finally, the largest position tends to be the one you are most confident about, because confidence is usually what made it large. That makes it the position examined the least skeptically, at exactly the moment it deserves the most arithmetic.
How Clavix reads this
Clavix is an iPhone app that grades 1,000 US stocks and ETFs from A+ to F, refreshed every trading day. It scores five independent dimensions: financial health, news sentiment, macro resilience, sector resilience, and price stability. On all five, a higher score means lower risk, and four of the five carry weight in the grade. Holdings are entered manually, by ticker and share count. There is no brokerage login, no OAuth, and no third-party account access. The full grading methodology lays out which four combine into one letter, and why financial health is shown but not weighted.
The free grader on this site takes three to twenty holdings and returns a single grade for the whole portfolio. Concentration is one of the three layers that set that grade: the more of the book that sits in its largest names, the further the grade is marked down. What it does not do is print the numbers on this page. It gives you no largest-position percentage, no effective holding count and no band, and it does not connect to your accounts. That is the honest scope: the bands are a published, versioned measurement model, and the arithmetic behind them is three divisions you can run on your own portfolio in about a minute.
One detail cuts against us and is worth stating anyway. A fifth measure, largest industry, was built and then deliberately dropped, because the industry field is null for all 1,047 rows in the ticker metadata table: the data provider returns only one classification string, and that string is already stored as the sector. The row could never carry a value for anybody, and a limit nobody can breach is not a limit.
Check your own portfolio
Concentration decides how much of your outcome any one name speaks for, and the grader weights every holding by its value, so the name that is 41% of your money is 41% of your grade. Concentration then counts a second time, on purpose: the diversification layer marks the grade down again for how much sits in the largest names. Give it tickers with no share count or amount and it weights them equally instead, and says so.
The free Clavix grader
- Takes three to twenty holdings, entered by ticker and either a share count or a dollar amount. Below three it declines to grade rather than grading thinly, because two of the four things it measures describe how holdings behave together.
- Returns one A+ to F grade for the whole book, built from three layers: asset quality, less a penalty for how much of the money sits in the largest names, less a penalty for how much those names move together.
- Measures a fourth, stability, from how the book has actually behaved, and reports it beside the grade rather than inside it.
- Names the weakest of those four layers. It does not break the grade back out into the five per-ticker dimensions; for those, open a name in the stock risk directory.
- Weights by position value when you give share counts or amounts, and equally when you give neither. It tells you which it used.
- Names any ticker it does not cover and leaves it out, rather than dropping it quietly.
- No account, no email, no brokerage connection. Your tickers are sent to be scored and are never stored: no record of your holdings is kept.
Frequently asked questions
What percentage of my portfolio in one stock is too much?
There is no single percentage, because the answer moves with your holding count. Clavix reads the middle half of a measured reference population as ordinary. At 10 holdings that is 21.4% to 29.6% for the largest position, with the top tenth of books starting at 34.9%. At 20 holdings the same band is 12.9% to 17.6%, and the top tenth starts at 20.6%.
Is 40% in one stock too concentrated?
It depends on how many holdings you have. In a book of 8, the reference limit for the largest position is 40.9%, so 40% sits right at the edge of the top tenth of books assembled by chance. In a book of 3, a largest position of 40% is unremarkable: the reference band there runs 47.4% to 65.5%, and 33.3% is the arithmetic floor.
How many holdings do I need to stop being concentrated?
Concentration is about weights, not counts, so no number of holdings settles it by itself. A 20 name book where one name is 40% is more concentrated than a 10 name book at even weights. The reference bands do flatten out, though: past 30 holdings the curves move less than a percentage point per name, which is why the Clavix table stops there.
Does concentration risk apply to ETFs too?
Yes, and it is harder to see. A fund is one line in your account but many companies underneath, so two funds can carry the same large names and your true weight in a single company can sit well above what any one line suggests. The measures on this page treat each line as one position, which makes them a floor on your real concentration, not a ceiling.
Why does employer stock count as concentration risk?
Because the position and the paycheck depend on the same company. If the business struggles, the value of the shares and the security of the income move together, so the effective exposure is larger than the portfolio weight alone shows. Employer stock also tends to grow on a vesting schedule rather than by a decision, so the weight rises without anyone choosing it.
Is concentration risk the same as being undiversified?
They point at the same thing from opposite ends. Diversification asks how many independent bets you own. Concentration asks how much of the outcome the biggest one claims. Concentration is the easier of the two to measure, because it needs only position values and a total. How closely the names move together is a separate question and a separate measurement.
For educational purposes only. This page explains how a risk measure is calculated; it is not investment advice, not a recommendation about any security, and not a prediction of return. Clavix is not a broker-dealer or a registered investment adviser. Consider consulting a licensed professional before making investment decisions.