Profitability margins
A good gross margin: ten sector medians
What a good gross margin is depends on the sector: medians on the 2026-09-07 build ran from 26.54 percent in energy to 61.80 percent in healthcare.
By Javier SanzPublished
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
On the 2026-09-07 distribution build the median gross margin ran from 26.54 percent in energy to 61.80 percent in healthcare. Forty percent is the number that gets quoted as a good gross margin, and read inside a sector it stops meaning one thing.
On the 2026-09-07 distribution build, across 2,401 US companies with filed accounts carrying a gross margin, the median was
39.39 percent (GET /api/public/distribution/gross_margin, build 2026-09-07)and the interquartile range ran from 23.65 to 61.35. So 40 percent is approximately the middle of the market. That is the least useful thing you can know about it, because almost nobody competes against the middle of the market.
The same line, read inside ten sectors
Breakpoints as the distribution endpoint returns them, rounded to two decimals. One row per company, not per listing, drawn from companies whose filed statements are on record.
| Sector | Companies | p25 | Median | p75 | p90 |
|---|---|---|---|---|---|
| Healthcare | 433 | 42.49 | 61.80 | 77.23 | 89.50 |
| Communication Services | 106 | 36.33 | 54.75 | 73.28 | 84.91 |
| Technology | 536 | 31.66 | 49.27 | 70.91 | 80.29 |
| Financial Services | 61 | 24.95 | 46.28 | 70.25 | 93.44 |
| Consumer Cyclical | 334 | 21.42 | 36.47 | 49.88 | 68.89 |
| Consumer Defensive | 172 | 19.13 | 33.42 | 50.23 | 62.39 |
| Utilities | 37 | 20.48 | 31.82 | 61.35 | 72.38 |
| Industrials | 417 | 19.33 | 30.12 | 40.75 | 55.52 |
| Basic Materials | 171 | 15.71 | 27.69 | 39.37 | 55.06 |
| Energy | 114 | 13.79 | 26.54 | 42.69 | 78.44 |
| All sectors | 2,401 | 23.65 | 39.39 | 61.35 | 79.34 |
Healthcare's first quartile is 42.49. A healthcare company on exactly 40 percent is in the bottom quarter of its own sector. Industrials' third quartile is 40.75. An industrial company on exactly 40 percent is a whisker off the top quarter of its own sector. Technology's median is 49.27, so the highest gross margin Apple has ever filed sits below the middle of its own sector too.
The same figure. Two opposite readings. The gap between the highest sector median and the lowest is 35.26 percentage points, which is wider than the entire interquartile range of the industrials cohort. A sector's standing on margin does not carry through to its standing on return on assets.
Why our own 40 percent rule says it is a convention
The open research section publishes a rule called "gross margin stays at or above 40 percent". Its stated source is not a study. It reads:
A convention, not an estimated boundary. No published threshold for a gross margin is defensible across industries, so this one is a line to test rather than a standard to meet.
That is the honest position and the table above is why. The metric registry gives gross margin no good or bad bands at all, for the same reason. A band would have to be one number, and one number is wrong for at least four of the ten sectors here.
The two things the table cannot tell you
The utilities row rests on 37 companies and the financial services row on 61. Both clear the thirty-company minimum this build applies, and both are thin enough that a single unusual filer moves the quartiles. The healthcare and technology rows rest on 433 and 536, and are steady.
Real estate has no row. The build refused it, with the sentence it publishes for every refusal:
Fewer than 30 comparable companies have this figure on file, so a percentile would be noise. We are not showing one.
That refusal is not an accident of sample size alone. Property trusts do not present a cost of revenue line, so a gross margin is not defined for them and the figure is withheld before the cohort is even counted.
What is excluded before the median is taken
Three exclusions, all applied before the numbers above:
617 companies withheld as not applicable. Banks, insurers, equity real estate trusts and mortgage trusts do not present a cost of revenue line. A gross margin computed over their accounts would be a number, and it would not mean anything, so it is refused rather than printed.
51 exact zeros dropped. A gross margin of precisely 0.00 is almost always a missing line coerced to zero somewhere upstream, not a company that sold everything at cost. Those rows are treated as missing.
Winsorising at the first and ninety-ninth percentile. The all-market observations were clamped to -125.95 and 99.45 before the breakpoints were computed. Values are clamped, never dropped, so the cohort size is the cohort that produced the numbers.
The reading that survives
Two sentences do the work of the table.
A gross margin is only informative against the same sector. The peer percentile is the reading, and the absolute number is not.
Where a company sits inside its sector is a fact about the company. Where the sector sits is a fact about how that industry books its costs, and the two get confused constantly. Neither is the same as how far one company's own margin has moved across ten filed years.
What this leaves out
These are breakpoints from one build, dated 2026-09-07. The distribution is rebuilt from filed accounts and the numbers move. The live figures are on /coverage and in the distribution endpoint itself, which stamps every response with its build date and marks itself stale after 35 days.
The cohort is companies whose filed statements are on record, deduplicated to one row per company. It is not the whole US universe, and the counts above are not coverage claims. A count over the companies holding a figure is a different statistic from a count over the market.
Sector is the coarse grouping. Within healthcare, a device manufacturer and a pre-revenue biotech sit in the same row and have nothing in common. Within technology, a software company and a contract manufacturer do the same. The finer industry level would split these, and this build publishes sector only.
The stored gross margin prefers the data vendor's trailing-twelve-month ratio where one is published, so it is not always the ratio you would get by dividing one fiscal year's gross profit by that year's revenue. Where the two differ, the annual recompute is the one traceable to a 10-K.
Nothing above is a view on any company or any sector. A percentile describes where a figure sat among comparable companies. It does not say whether any of them is worth owning.