Five contract names get used as if they meant the same thing: staff augmentation, a statement of work (SOW) project, managed services, consulting and staffing. They don't. One question sorts them: who directs the daily work, and who answers for the outcome? If you run a 5 to 50 person agency you ask it twice, once when you buy overflow capacity and again when a client wants to know which kind of contract they're signing.
The one question that decides the category
In staff augmentation, the client directs the daily work and carries the delivery risk, and the vendor supplies people and bills for their time. In an SOW project, the vendor directs the work toward a defined deliverable, and the contract is measured on acceptance criteria. In managed services, the vendor runs an ongoing function (a hosted app, a support desk) against service levels and answers when they're missed. Consulting sells advice: the consultant recommends, and the client decides and builds. Staffing is a recruiter placing a worker with you, who then works under your direction.
Those are working definitions, and vendors stretch them. A statement of work is just the document, so a staff augmentation engagement can be written on one too. Staffing and staff augmentation overlap, because in both you direct the day. What differs is what you buy: a match from a recruiter in one case, a firm's people in the other, and with a recruiter you should check who employs the person.
Open any proposal and find the sentence that says who assigns tomorrow's work. If it's you, you're buying staff augmentation or staffing, whatever the cover says. If it's the vendor, find the sentence that says what happens when the result is late or wrong. That sentence is either an acceptance criterion (a project) or a service level (a managed service). If neither exists, you haven't got either one.
The five models side by side
| Model | Who directs the work | What you pay for | Who carries delivery risk | What the contract is measured on | When it fits an agency |
|---|---|---|---|---|---|
| Staff augmentation | You | People's time | You, mostly | The people, and swap-out terms | Your lead has spare hours and the backlog won't end |
| SOW project | The vendor, toward a defined deliverable | The deliverable | The vendor, within the scope | Acceptance criteria | A self-contained build with a stable spec |
| Managed services | The vendor, running an ongoing function | The function running | The vendor, against the service levels | Service levels (response, uptime, fix times) | A stable system you want someone else on call for |
| Consulting | The client decides, the consultant advises | The advice | The client, because advice isn't the build | A recommendation or report delivered | You need an answer before you commit to a build |
| Staffing (recruiter) | You | The match, and sometimes the worker's time | You | A worker found and placed | A role that's really a job |
The columns that matter most are the second and fourth, and in a clean contract they point at the same party: you and you for staff augmentation, the vendor and the vendor for a project or a managed service. If a proposal has you assigning the tasks while the vendor carries the risk, one of those cells is wrong, and you'll find out which on the first bad week.
What the FAR and California law look at
Two legal texts draw the same line. Neither is a rule about what to call a private agency contract, so here is what each covers (both read on 21 October 2026).
The Federal Acquisition Regulation (FAR) governs how the US government buys. Its definition of a personal services contract (FAR 2.101) is one that, "by its express terms or as administered," makes contractor personnel appear to be, in effect, Government employees. FAR 37.104 says that happens when contractor personnel are subject to "relatively continuous supervision and control" of a Government officer or employee, and it names the key question: will the Government exercise it? It also separates out the project case. Giving an order for a specific article or service, with the right to reject the finished product or result, is not that kind of supervision. Agencies can't award a personal services contract unless a statute specifically authorizes it.
That is federal civil-service law, and it doesn't govern your client's commercial contract. I'm citing it because it's the plainest written version of this post's line: directing the result is one thing, directing the person day to day is another. If one of your clients is a federal agency, it stops being background reading, so have counsel read 37.104 first.
California's version is the ABC test in Labor Code section 2775. For purposes of the Labor Code, the Unemployment Insurance Code and wage orders, a person providing labor or services for pay is an employee unless the hiring entity shows three things. The person is free from the control and direction of the hiring entity, "both under the contract for the performance of the work and in fact". The work is outside the usual course of the hiring entity's business. And the person is customarily engaged in an independently established trade, occupation or business of the same nature as the work. The first prong asks about control in fact, not just on paper, so a clause saying a contractor controls their own methods won't help if your project manager assigns their tickets every morning. Section 2776 sets 2775 aside for a bona fide business-to-business relationship, but only if the contracting business demonstrates all twelve criteria.
So the label doesn't decide. Both texts look at how the work is directed: "as administered" in one, "in fact" in the other.
A worked example: one client, one week
Picture a 16-person product agency (an invented example, not an ivector client). On a Monday in November its biggest retail client emails four asks: a loyalty points screen in the app before 15 December, someone to answer when checkout alerts fire overnight, a recommendation on whether to leave the current storefront platform, and "two more developers until January."
The owner's first instinct is one time-and-materials addendum covering all four. Read against the question, they're four different contracts.
The loyalty screen has an end and a way to judge it, so it's an SOW project: the agency directs, and the client accepts against written criteria. The overnight alerts are an ongoing function, so they're managed services, and they need response times, a named person on call and a definition of fixed, or the agency has made an unpaid promise. The platform question is consulting: a short written recommendation, priced as advice, and the client decides. The two developers are staff augmentation. The client's product manager directs them, and the agency carries the people, not the outcome.
Sign the single addendum and the week goes like this. By Wednesday the client's product manager is assigning tickets to the agency's developers, which is fine. By Thursday the checkout alerts are texting the agency's tech lead at 2am, with no service levels, billed by the hour. And the loyalty screen has no acceptance criteria, so it's done whenever the client says so.
Split it instead. One SOW for the screen, a short staff augmentation addendum for the two seats, a separately priced service-level schedule for the alerts (or a polite no), and a one-page note for the recommendation. The agency can buy overflow behind any of them, under the same model it sold. The two seats are staff augmentation, and ivector's page for that describes engineers who work in your tools and standups while you direct the work. The screen is closer to a project, and the partners page describes a fixed-price pilot task first, then a monthly engagement for the engineer or two-person pod.
When each model is the wrong choice
Staff augmentation is the wrong choice when your tech lead has no spare hours. Added people need tickets, reviews and answers from you. It's also wrong when the client is buying an outcome you can't supervise, such as overnight uptime.
An SOW project is the wrong choice when the scope moves every week, because each change becomes a negotiation over what done means. Fixed price vs time and materials vs dedicated team covers that trade.
Managed services is wrong when the system is still changing fast, because there's no stable baseline to write service levels against. It's also wrong when nobody at the vendor will answer at 2am. A service level nobody staffs is a liability with a number on it.
Consulting is the wrong choice when the client wants the thing built, because a recommendation without a builder tends to end up in a folder. Staffing is the wrong choice when you need someone to answer for results, or when you can't say who employs the person you're about to put in your standup.
What to do before you sign
Here's the opinion I'd defend on a call: name the contract after how the work will run in week three, not how you hope it runs in week one. If your client's product manager will assign tickets every morning, sign staff augmentation and plan the management time it takes. Don't write it as an SOW to make it sound like the vendor holds the risk. They won't, and the paper stops matching the practice by the first Thursday.
Before you sign anything, in your own contract or a client's, run the test from the top (who assigns tomorrow's work), then check three things.
- 1.For a project, find the acceptance criteria. For a managed service, find the service levels and who's on call.
- 2.For every individual, find out who employs them on paper. If the answer is nobody, because they're a freelancer, read the business-to-business criteria before the first invoice.
- 3.Write the model into the contract by name, along with what would change it.
This is general information, not legal advice, so have a California employment lawyer read your template once.
We sell staff augmentation and agency overflow capacity, so weigh that. The earlier comparison of freelancers, outsourcing and white-label covers the other half of the buying decision. If you can't tell which row a client's ask belongs in, tell us the scope.
