The Census Bureau's Quarterly Services Survey, released on 9 September 2026, shows revenue from government customers at computer systems design and related services firms up 7.5% in the second quarter of 2026 against the same quarter of 2025. Revenue across all customers in that industry grew 3.2%. If you run an agency, the tempting reading is that public-sector work is where the money is and you should chase it. That is half right, and the half that is wrong is the expensive half.
The numbers are real and they come from the publisher itself. The caveats are real too: they are preliminary, they carry a wide margin of error, and they describe an industry that is bigger than custom software. This post goes through what the table says, what it does not, and the staffing question an owner has to answer before bidding on government work: how much of the job can you hand to someone else?
What does the Census Bureau's services survey measure?
The Quarterly Services Survey (QSS) is the Census Bureau's quarterly estimate of revenue for employer firms in selected service industries. The table used here, Table 13 in the Q2 2026 release, covers Professional, Scientific, and Technical Services and breaks revenue out by class of customer. Our row is NAICS 5415, computer systems design and related services, which includes custom programming along with systems integration, facilities management and similar work. If you are a 15-person web and product agency, you are one small cell in that row.
Three details matter. The estimates are not seasonally adjusted and not adjusted for price changes. The second-quarter 2026 figures are marked preliminary. And the Bureau suppresses some cells entirely: in this industry the business-customer line is withheld in every quarter shown, which will matter in a moment.
What did government customers do in the second quarter of 2026?
Here are the Census figures for NAICS 5415, in millions of dollars, straight from Table 13:
| Quarter | Total revenue | Government customers | Household and individual customers |
|---|---|---|---|
| 2Q 2025 | 189,379 | 58,383 | 10,005 |
| 1Q 2026 | 192,352 | 60,605 | 9,422 |
| 2Q 2026 (preliminary) | 195,524 | 62,783 | 9,391 |
Government customers were 62.8 billion dollars of 195.5 billion in the second quarter of 2026, or 32.1%. A year earlier they were 58.4 billion of 189.4 billion, or 30.8%. Year over year, government revenue rose 7.5% and total revenue rose 3.2%. For the first half of 2026 against the first half of 2025, the Census table has government up 7.6% and total up 4.9%.
The more interesting number is the one the Census does not print. Business customers are suppressed, but if you subtract government and household revenue from the total, what is left is roughly 123.4 billion dollars in the second quarter of 2026 against 121.0 billion a year earlier, growth of about 1.9%. In the first quarter, the same subtraction gives about 6.5% year over year. That is my arithmetic, it assumes the three customer classes add up to the total (the table does not state that), and it should be read as an approximation. But it points at what happened: government revenue held steady near 7.5% growth while everything else slowed sharply in the spring.
How much should you trust a 7.5% growth figure?
Less than the headline suggests. Census publishes sampling variability in Table 14 of the same release. For NAICS 5415 government revenue, the standard error on the year-over-year change is 2.5 percentage points, and the coefficient of variation on the second-quarter level is 13.3%. The standard error on the change in total revenue is 1.7 points.
Put plainly, the true government growth rate could comfortably sit anywhere from the low single digits to the low double digits, and the gap between 7.5% and 3.2% is suggestive rather than settled. The first-quarter comparison is a good check: government was up 7.7% and total was up 6.6%, a much smaller gap. What changed between quarters was mostly the non-government side. So the safer reading is not that government work is booming. It is that government demand stayed put while commercial demand softened, which is a different thing to put in a pitch deck.
What does that mean for an agency's week?
Picture a 14-person agency whose revenue is mostly three commercial retainers. One client trims its monthly hours. A second delays a build. The owner sees a state or county RFP for a case-management portal and thinks about bidding, because the Census says government is where growth held up.
The portal is probably more work than the agency can staff at the size it has today, and that is the real decision. A bid commits the agency to a price, a schedule and, often, a subcontracting limit written into the contract. On a federal small business set-aside, for example, the SBA's 50% rule caps how much of the money the agency can pass to firms that are not similarly situated. A different federal rule, covered in our post on the DOJ data security rule, can turn on where one of your subcontractor's engineers lives. Neither is a reason to avoid public work. Both are reasons to know the staffing plan before you sign.
Should you hand the extra capacity to a subcontractor?
My view: use outside capacity for the part of the plan the contract lets you subcontract, and keep the pieces that carry the compliance risk with people you employ. That usually means the interface work, integrations, test suites and internal tooling go to a partner, and the architecture, anything touching the sensitive records and the final acceptance stay in house.
The economics are the same ones that govern any overflow, covered in our post on white-label development economics. The difference with public work is that payment is slower and the paperwork is heavier, so the cash needed to carry a partner for a quarter before the first milestone payment is larger than on a commercial job.
When is a subcontractor the wrong answer here?
Three cases. If the solicitation is a set-aside and your own payroll cannot cover at least half the work, adding a subcontractor does not fix that, it makes the problem worse, and you should either team with a similarly situated firm or pass. If the contract restricts where the work can be done or who can touch the data, check that before you ask any outside bench, including ours, to take a share. And if the only reason to bid is that Census says government is growing, skip it. A 7.5% growth rate across an entire industry says nothing about whether you will win this one.
Census data tells you where the market moved last quarter. It does not tell you your win rate.
If you are weighing whether to staff an upcoming job with an outside engineer, our agency overflow offer starts with a small fixed-price pilot in your own repository, so you can judge the work before you commit to a monthly arrangement. We will tell you plainly when it does not fit. If you want to talk through a specific bid, get in touch, and if you are still deciding how to staff growth, our build-your-team page sets out the options.
