Service Annual Survey / Quarterly Services Survey

It is the federal statistical system's measurement of the U.S. service economy, assembled by the U.S. Census Bureau and delivered here as one normalized shelf.

The live half is the Quarterly Services Survey, the only source of timely quarterly indicator performance for selected service industries. Each release publishes estimated total revenue for employer firms - seasonally adjusted and not seasonally adjusted - with measures of sampling variability attached, and the Bureau of Economic Analysis uses the series as an input to GDP estimates. That makes it both a market read and a load-bearing component of the national accounts.

The historical half is the Service Annual Survey, which produced national annual estimates of revenue and expense for select service sectors - NAICS 22, 48-49, 51-54, 56, 61, 62, 71, 72 and 81 - drawn from a probability sample of employer firms on the Business Register. SAS has transitioned into the Annual Integrated Economic Survey, which began collection in March 2024; the final standalone SAS reference year is 2022, re-released on September 25, 2024 on a 2017 NAICS basis, and the 2013-2022 annual series remains intact behind it.

Both halves resolve NAICS 54, professional, scientific, and technical services, to the level decisions actually get made at: legal services (5411), accounting and payroll services (5412), architectural and engineering services (5413), computer systems design (5415), management, scientific, and technical consulting services (5416), scientific research and development (part of 5419), and advertising and public relations (5418).

Real rows from the 2Q2026 cut:

# Service Annual Survey / Quarterly Services Survey -- 2Q2026 revenue estimates, employer firms
# revenue in millions of US dollars

naics            : 54
kind_of_business : Professional, scientific, and technical services
rev_2q2026_mm    : 818997            ($819.0bn)
rev_1q2026_mm    : 790044
yoy_pct_change   : +6.4

naics            : 5416
kind_of_business : Management, scientific, and technical consulting services
rev_2q2026_mm    : 114771            ($114.8bn)
rev_1q2026_mm    : 112112
yoy_pct_change   : +5.5

naics            : 5411
kind_of_business : Legal services
rev_2q2026_mm    : 115213
rev_1q2026_mm    : 107198
yoy_pct_change   : +8.7

naics            : 5415
kind_of_business : Computer systems design and related services
rev_2q2026_mm    : 194768
rev_1q2026_mm    : 192352
yoy_pct_change   : +2.8

naics            : 5412
kind_of_business : Accounting, tax preparation, bookkeeping, and payroll services
rev_2q2026_mm    : 58264
rev_1q2026_mm    : 67078
yoy_pct_change   : +4.9

Read the middle block as the whole value proposition in five lines: an industry everyone argues about, pinned to a NAICS code, carrying a dollar level ($114,771 million for the quarter), a prior-quarter comparison ($112,112 million) and a year-over-year rate (+5.5%). The spread around it is the argument-settler - legal services growing fastest of the group at +8.7%, computer systems design slowest at +2.8%, accounting actually down sequentially on the quarter while still up 4.9% on the year. Get a sample of this dataset.

Five fields carry each row. 2017 NAICS code and kind_of_business establish identity, pairing a machine-joinable classifier with its human-readable name at three- to six-digit detail. total_revenue_millions holds the estimate itself - employer-firm revenue in millions of dollars for the reference period. percent_change carries the movement, quarter-over-quarter on some tables and year-over-year on others. coefficient_of_variation quantifies sampling variability, which doubles as the suppression trigger: cells exceeding publication standards are withheld rather than printed.

Everything else rides under additional fields on request, described in the next section.

Three families. First, the second sampling-variability measure - the standard error of the percent-change estimate - published beside the coefficient of variation on the variability tables. Second, the seasonally_adjusted_flag, the marker distinguishing the adjusted series from the raw one so both vintages can share a single schema. Third, the annual detail-table columns from the thirteen SAS-era tables: revenue by tax status, expenses, sources of revenue, export revenue, inventories, revenue by product line and class of customer, revenue from electronic sources, patient visits, health-record system expenses and telemedicine revenue.

None of these pad the default extract. Name the ones you need when requesting the sample and the file arrives shaped around them.

Geographically the coverage is strictly United States national totals - no state, metro or county cuts in either series. The dataset answers 'how big and how fast', never 'where exactly'.

Temporally the quarterly series reaches back to the survey's 2004 start, roughly two decades of readings, published in advance and revised vintages so a point-in-time snapshot and a best-current-estimate view are both reconstructible. The annual series spans 2013 through 2022.

Granularity sits at employer firms aggregated to selected service industries, resolved at NAICS 3- to 6-digit level within sectors 22, 48-49, 51-54, 56, 61-62, 71-72 and 81. Roughly ninety industry rows fill each quarterly table, which is why the whole service economy reads as one flat series rather than a stack of industry-by-industry lookups.

API, files, or your warehouse. Daily, weekly, or hourly.

The cadence is yours even though the series themselves are quarterly and annual - land each vintage the moment it matters to your pipeline instead of waiting on someone's download page. What arrives is normalized: naics_code, kind_of_business, revenue, percent change and variability measures typed identically whether the row is the current quarter or 2004, so a twenty-year diff is one query rather than an archaeology project across changing spreadsheet layouts. Suppressed cells arrive flagged rather than silently blank, and the seasonally adjusted and raw series ride the same schema with the flag making the difference explicit.

Request the sample and the extract comes carrying exactly the columns you named - the verified five-field core plus anything folded under additional fields.

Five jobs dominate.

1. Market sizing. Anchor a consulting TAM model on the measured $114,771 million for NAICS 5416 rather than a consultant's round-up, quoting the same basis the national accounts use.

2. Momentum tracking. Read the sector's direction each quarter from the percent-change column: professional, scientific, and technical services up 6.4% year over year in 2Q2026, with legal at +8.7% pulling the average up and computer systems design at +2.8% holding it back.

3. Firm benchmarking. Put a consultancy's own growth next to its industry series and the review conversation becomes evidentiary - ahead of the line or behind it, by points.

4. Macro modeling. Key forecasts to the same series the Bureau of Economic Analysis folds into GDP, so model outputs reconcile to the national accounts instead of drifting from them.

5. Cycle baselining. Two decades of quarterly history supply pre- and post-shock baselines for 2008-09 and the pandemic in services revenue specifically, not in aggregate GDP.

Market Researchers & Consultants get their own profession measured by the government - quotable dollar figures at the exact NAICS line a client occupies, refreshed every quarter.

Investors & Quants get a timely official read on professional-services demand that leads company filings and reconciles to GDP inputs, usable as a factor or as context around earnings season.

Strategy & Corporate Development teams get market-entry sizing at 3- to 6-digit NAICS detail with year-over-year movement, separating growing niches from crowded ones before capital commits.

Journalists, Academics & Students get citable federal statistics on the service economy, standing behind every dollar figure and growth rate they print.

If the question is 'how big is this profession and which way is it moving', this shelf is the shortest route from question to defensible answer.

Three things. First, geography stops at the national line - there are no state or metro cuts here, so regional questions need a sub-national companion dataset beside this one.

Second, suppression is part of the design. Estimates whose coefficient of variation exceeds publication standards are withheld, so thin industries can disappear from individual quarters; a gap means 'too variable to publish', never zero. Build pipelines that expect the flags.

Third, mind the seam at 2022. The standalone Service Annual Survey ends with reference year 2022 and continues inside the Annual Integrated Economic Survey, so confirm how the SAS-style service-industry tables map into AIES releases before anchoring anything on post-2022 annual figures. The quarterly side carries straight through. Get a sample of this dataset and see the exact slice you would build on.