⚡The three models, defined precisely
These terms get blurred in sales decks, so here are precise definitions. The distinguishing variable is always the same: who manages the work.
Staff augmentation means individual engineers from a vendor join your team, work in your tools, attend your standups, and take direction from your leads. The vendor supplies vetted people, handles payroll and replacement, and otherwise stays out of the way. You are buying capacity, and everything else — prioritization, process, quality bar, delivery — remains yours.
A managed team (also called a dedicated team) is a self-contained unit: engineers plus a team lead or engineering manager supplied by the vendor, working exclusively on your product, with their own internal process. You manage the what — roadmap, priorities, acceptance — and the vendor lead manages the how: task breakdown, code review, day-to-day supervision. You are buying a durable unit of delivery capacity with a management layer included.
An agency engagement buys a defined outcome: a scoped project or a stream of deliverables, with the agency owning project management, delivery risk (to the extent the contract transfers it) and the team composition behind the scenes. You interact with a project manager and a delivery plan rather than with individual engineers. You are buying a result, priced accordingly.
See how our dedicated teams are structured
| Dimension | Staff augmentation | Managed / dedicated team | Agency (project-based) |
|---|---|---|---|
| What you buy | Individual capacity | A self-managing unit | A defined outcome |
| Who manages the work | You | Vendor lead, you set priorities | Agency PM |
| Who owns the process | Yours entirely | Vendor internal, aligned to yours | Agency process |
| Pricing shape | Monthly rate per person | Monthly team rate | Per project or milestone |
| Management burden on you | High | Medium | Low day-to-day, high at scoping |
| Typical duration | 3–24 months, flexible | 6+ months, durable | Weeks to months, bounded |
💵The cost structure of each model
The headline price ordering is predictable — augmentation cheapest per hour, agency highest — but the honest comparison is total cost of a unit of delivered work, including your own management time. The table below uses labelled market ranges as of writing (September 2026); verify against current quotes before committing budget.
Augmentation rates look like salaries plus a margin, because that is what they are: the vendor carries recruiting, vetting, payroll, benefits, equipment and replacement risk, and charges 25 to 50 percent over the local employment cost. There is no management layer in the price, which is the entire point — you are either supplying one or nobody is.
Managed team pricing adds the lead (typically 15 to 25 percent of team cost) plus the vendor’s delivery infrastructure. The effective premium over augmentation is usually 15 to 30 percent, and it buys you the supervision, code review and coordination you would otherwise pay for with your own engineering leadership hours — which are the most expensive hours in the comparison.
Agency pricing carries the heaviest load: project management, discovery, QA, delivery risk padding and account overhead. Per hour it is the most expensive model, often 1.5 to 2 times augmentation rates, and it is still frequently the cheapest total cost for a bounded project — because you are not paying your own leadership to run it, and because a fixed-scope agency contract transfers the estimation risk to the party best positioned to carry it.
| Cost component | Staff augmentation | Managed team | Agency |
|---|---|---|---|
| Senior engineer, monthly (offshore, labelled range) | $4K–$9K | Included in team rate | Bundled into project price |
| Management layer | Yours — not in the price | Vendor lead, 15–25% of team cost | Agency PM, built into the rate |
| Effective premium over raw payroll | 25–50% | 40–70% | Highest — includes risk and overhead |
| Your internal management cost | High — 10–20 hrs/week per 3–4 engineers | Medium — 3–6 hrs/week | Low — 2–4 hrs/week at steering level |
| Recruiting / replacement cost | Vendor carries | Vendor carries | Invisible to you |
| Scaling cost | Linear — add a person | Stepwise — grow the unit | Renegotiate scope |
Never compare models on vendor price alone. Add your own management hours at internal cost to each option. Augmentation routinely stops being the cheapest model the moment you price the engineering lead it requires — and frequently remains the right choice anyway, for control rather than cost.
🧭The management burden, honestly itemized
This is the dimension buyers underestimate most consistently, so it deserves its own section. The management burden is not a soft factor — it is a weekly time commitment from specific, expensive people on your side, and if those people do not exist or do not have the hours, the model fails regardless of vendor quality.
With augmentation, you need a working engineering management function: someone who onboards the engineers, writes the tasks, reviews the code, runs the process and answers questions same-day. Budget ten to twenty hours per week of a competent lead for every three to four augmented engineers, plus a product owner who can make priority decisions. Teams without this function should not buy augmentation — they should buy the management with the people, which is exactly what a managed team is.
With a managed team, your burden drops to direction-setting: roadmap, priorities, acceptance criteria, and a weekly cadence with the vendor lead. Three to six hours a week, concentrated in one or two people. With an agency, the burden concentrates at the front — a real scoping effort, crisp acceptance criteria — and then drops to steering: demos, decisions, unblocking. Two to four hours a week for most of the engagement, with spikes at milestones.
Augmentation failure mode
An augmented engineer waiting three days for a task or a review is a meter running on idle. The model collapses without responsive internal ownership — this is the most common way augmentation engagements quietly die.
Managed team failure mode
Unclear priorities from your side. The lead can run the team but cannot decide what the product should do. A managed team with a vague roadmap produces motion, not progress.
Agency failure mode
Sloppy scoping and absent acceptance criteria. Everything an agency delivers is tested against what you wrote down in week one — if that was vague, every dispute is renegotiated at milestone time.
🏆When each model wins
With definitions, costs and burdens on the table, the fit patterns are fairly crisp. None of these are rules — they are the configurations where each model has structural advantage rather than merely being usable.
Discuss which model fits your roadmap
| Your situation | Winning model | Why |
|---|---|---|
| Strong internal eng leadership, need 1–3 specialists fast | Staff augmentation | Your management exists; buy only the capacity |
| Product company, no spare management bandwidth, long roadmap | Managed team | Management comes with the unit; durable context |
| Bounded project with a clear deliverable | Agency | Outcome pricing transfers estimation risk |
| First-ever external hire, small company | Agency or managed team | You have no process to augment into |
| Core product, multi-year horizon, want eventual insourcing | Managed team with conversion path | Durable knowledge, contract option to hire the team |
| Spiky, unpredictable workload | Staff augmentation | Scale headcount monthly without renegotiating scope |
| Highly regulated or IP-sensitive core | Augmentation into your environment | Maximum control over access, tooling and audit trail |
🧮The decision matrix
Score your situation against each row — be honest about the management row, it is the one everyone overestimates — and the model usually selects itself. If the rows split, default to the model that matches your weakest row, because engagement models fail at their constraint, not their average.
| Question | If yes → | If no → |
|---|---|---|
| Do you have an engineering lead with 10+ hrs/week to spare? | Augmentation is viable | Managed team or agency |
| Is the work a bounded deliverable with a definable done? | Agency is viable | Augmentation or managed team |
| Will this work run longer than six months? | Managed team gains an edge | Augmentation or agency |
| Is scope likely to evolve significantly? | Avoid fixed-scope agency pricing | Any model works |
| Do you need delivery risk transferred contractually? | Agency (fixed price) or managed team (SLA) | Augmentation |
| Is per-hour cost the binding constraint? | Augmentation | Consider what you are trading away |
| Do you need to start within two weeks? | Augmentation (bench) or managed team | Any model works |
The single best predictor of engagement success is not the model — it is whether the buyer honestly assessed their own management capacity. Every failed augmentation we have seen post-mortemed had the same root cause: nobody on the client side actually owned the augmented engineers’ week.
🔀Hybrid structures that work in practice
Real engagements rarely stay inside one box, and the hybrids below are standard practice rather than exotic arrangements. The principle behind all of them: match each slice of work to the model that fits it, instead of forcing the whole relationship into one contract shape.
Agency for discovery, managed team for build. The agency runs a fixed-price discovery and design phase — bounded, outcome-shaped, low risk — and the findings define the backlog that a durable managed team then executes. This is the structure we recommend most often for new products, because it prices each phase by its actual risk profile.
Managed core plus augmented specialists. A stable managed team owns the product; augmented specialists (a security engineer for an audit window, a data engineer for a migration, an ML engineer for a model integration) drop in for bounded stints under the team lead’s coordination. The core keeps context; the spikes cost less than carrying specialists permanently.
Managed team with a conversion clause. If there is any chance you will want to insource later, negotiate the option up front: after twelve to eighteen months, you may hire some or all of the team for a defined conversion fee. The fee is real — it compensates the vendor for recruiting cost — and it is still usually cheaper than hiring cold into a system you do not yet understand.
Explore our dedicated team modelAsk about hybrid and conversion structures
📜Contract terms that matter more than the rate
Whichever model you choose, four clauses determine how the relationship actually behaves. Replacement terms: how fast a departing engineer is swapped and at what cost — a two-week replacement SLA with a knowledge-transfer obligation is the standard to hold vendors to. Notice periods: thirty days is healthy; ninety-day locks favor the vendor. IP and access: all work product assigned to you on payment, credentials and repositories in your accounts from day one, in every model. Rate adjustment: how and when rates change, indexed to something objective, so year two does not open with a surprise.
And one non-clause: the kickoff artifact. Before any model starts, require the first two weeks in writing — who does what, in which tools, with which rituals. Most engagement failures are visible by day ten, and only to people who wrote down what day ten was supposed to look like.