Recruitment Models Compared: In-House vs Agency vs RPO vs Embedded RPO

A recruitment model is how you resource hiring: who does the work, where they sit, and how you pay for them. There are four in common use. You build a team in-house, you use an agency, you outsource the process to a classic RPO, or you embed an external team inside your own. Most companies eventually run a hybrid of two or more.

None of them is best in the abstract. Each is the correct answer to a different problem, and each is an expensive mistake applied to the wrong one. This guide compares all four on how they work, what they actually cost, where they break, and when they are the wrong choice, using real numbers rather than vendor claims.

The four recruitment models at a glance

The rest of this guide takes each in turn, then compares the two that are most often confused: classic RPO and embedded RPO.

In-house recruitment

In-house means building your own talent function: recruiters, sourcers and coordinators on your payroll, reporting into your business.

How it works

You hire recruiters as employees. They own the full process, sit in your meetings, learn your product, and build relationships with hiring managers over years. Over time they accumulate something no external partner can hand you: institutional memory of why past hires worked and why others did not.

What it costs

The salary is the visible part. The real number includes employer costs, tooling, ATS and LinkedIn Recruiter licences, job advertising, and management time. It also includes the vacancy: a senior in-house recruiter typically takes two to four months to hire and several more to reach full productivity, and the roles they were meant to fill stay open in the meantime.

Where it wins

• Steady volume. If you hire consistently all year, a permanent team is almost always the lowest cost per hire.
• Institutional knowledge. Nobody will ever know your business like someone who has worked in it for three years.
• Employer brand. Your recruiters are your brand in the market, permanently, not for the length of a contract.
• Control. You set priorities directly, with no commercial conversation in between.

Where it breaks

In-house is a fixed cost against a variable need. Hiring plans move; payroll does not. The model struggles when volume is uneven, when you need a skill your team has never hired for, when you enter a market where nobody knows your brand, or when the plan needs recruiters working this month rather than in four months. And when hiring slows, you are left choosing between carrying idle cost and making redundancies in the team whose job is hiring.

Capacity is the other constraint. SHRM's 2025 Recruiting Benchmarking Report puts the median recruiter at 20 job requisitions a year, rising to about 60 at extra-large organisations. Divide your hiring plan by that number and you have a rough sense of the team you would need to build.

The recruitment agency model

An agency works outside your business and introduces candidates for a fee. It is the oldest model and still the right one more often than the industry likes to admit.

How it works

You brief an agency on a role. They search their network and the market, and send you a shortlist. You pay when someone is hired (contingency) or in staged instalments for an exclusive search (retained). The recruiter works your role alongside other clients' roles.

What it costs

Contingency fees typically run 15% to 25% of first-year salary. Retained search usually runs 25% to 33%. The model is genuinely low-risk on a single role: on contingency, if they do not deliver, you do not pay.

Where it wins

• Low volume. For one or two hires a year, no other model competes on total cost. A fixed monthly fee cannot beat a one-off percentage.
• Confidential or executive search. Retained search exists for this and is better at it than anything else on this list.
• Genuinely rare skills. A specialist agency that has spent a decade in one niche has a network you cannot rent by the month.
• Zero setup. No onboarding, no internal owner, no systems access. You send a brief and wait.

Where it breaks

The economics invert as you scale. Every additional hire costs the same percentage again, so the model punishes exactly the companies doing well. At 30 hires a year on 20% fees, agency spend becomes one of the largest lines in the talent budget, and you own none of the process that produced it.

The second problem is context. An agency recruiter splits their week across several clients. They cannot sit in your planning meeting, cannot tell your hiring manager the salary band will not clear the market, and often will not name you to candidates in case they apply directly. Candidate experience is delivered by someone whose relationship is with the transaction, not with your brand.

Classic RPO (recruitment process outsourcing)

Classic RPO, sometimes called core or enterprise RPO, means handing a defined hiring process to an external provider who runs it as their own.

How it works

You outsource a scope: a region, a function, or all volume hiring. The provider brings their recruiters, their methodology, often their technology, and their reporting. They run the process end to end and report against agreed KPIs. Contracts are usually annual or multi-year, and the scope is defined in a statement of work.

What it costs

Commercial models vary: a fixed management fee, a fee per hire, a cost-per-slot for the recruiters deployed, or a blend. At genuine scale it is materially cheaper per hire than agencies, which is the entire reason the category exists.

Where it wins

• Large, stable volume. Hundreds of similar roles a year is exactly what a repeatable, industrialised process is for.
• Process maturity. If your hiring process is genuinely broken, a provider with a proven methodology can install one.
• Compliance and reporting. Regulated, multi-country hiring benefits from a partner who does audit trails and workforce reporting for a living.
• One throat to choke. A single accountable partner across a big programme, rather than fifteen agency relationships.

Where it breaks

Classic RPO takes the process away. That is the point, and it is also the cost. The know-how accumulates on the provider's side, so at the end of a three-year contract you are often less capable than when you started, and switching is expensive because the capability leaves with them.

It is also slow to flex. When your plan changes in month four of a three-year agreement, you are in a commercial negotiation rather than a conversation. And because the provider runs their process in their way, it can quietly become a separate function that your hiring managers experience as a vendor rather than a team.

Embedded RPO

Embedded RPO places the provider's recruiters inside your team for a set period. They use your systems, run your process, and carry your employer brand. To candidates, and often to your own hiring managers, they are simply part of your team.

How it works

A partner deploys senior recruiters and sourcers into your talent function, typically for three, six or twelve months. They work in your ATS, from your email domain, on your LinkedIn Recruiter seat. They join your Slack, attend your planning, and take intake directly from hiring managers. You pay a monthly fee for the team, not a percentage of each hire. For a deeper treatment of the mechanics, see our guide to what embedded recruitment is and how it works.

Where it wins

• Continuous or scaling hiring. A flat monthly fee means cost per hire falls as volume rises, inverting the agency curve.
• Speed to capacity. Recruiters can be working your roles in one to two weeks rather than the months an internal search takes.
• New markets. Local partners bring the network, salary data and cultural context you do not have yet.
• Capability stays. Because they run your process in your systems, the workflows and market data remain with your team at handover.
• Flexibility. Scale the team up for a launch quarter and down when the plan changes, without redundancies.

Where it breaks

Embedded needs an internal owner. It multiplies a functioning hiring process; it does not replace one. If no hiring manager will attend intake or return interview feedback, an embedded team will not save you, it will simply surface the problem faster and with better dashboards.

It is also wrong at low volume. If you are hiring one or two people a year, a fixed monthly fee cannot compete with a one-off agency fee, and we will tell you that on the first call. And it is not a hands-off model: if what you want is a shortlist in your inbox with no involvement, you want an agency, and there is nothing wrong with wanting that.

Classic RPO vs embedded RPO: the difference that actually matters

This is the comparison most buyers get stuck on, partly because the industry uses the words interchangeably. Embedded RPO is a form of RPO, not an alternative to it. Both outsource recruitment capability. The difference is proximity, flexibility, and where the capability ends up.

Put simply: classic RPO is a better answer when the process is the problem and you want someone else to own it at scale. Embedded is a better answer when the capacity or capability is the problem and you want to keep owning the process.

Agency vs RPO: what is actually different

An agency sells you candidates. An RPO sells you recruitment capability. That is the whole distinction, and everything else follows from it.

An agency is paid per successful placement, so its incentive is to close the fillable roles quickly and move on. An RPO is paid for capacity or process, so its incentive is throughput and quality across the plan, including the awkward roles an agency would quietly deprioritise. Agencies are transactional by design, which is a feature at low volume and a liability at high volume.

What each model really costs

Cost is where these models separate most sharply, and where the sales conversation is least honest. The useful question is not which is cheaper, but which is cheaper at your volume.

The published benchmarks

In the US, SHRM's 2025 Recruiting Benchmarking Report shows median cost-per-hire has split in two: non-executive hires fell to $1,200, down 27% since 2017, while executive hires climbed to $10,625, up 113%. Routine hiring is getting cheaper; senior hiring is getting rapidly more expensive. That gap is where model choice starts to matter, because it is the senior end that agencies price as a percentage.

The cost of getting it wrong is larger than the fee. Gallup puts the cost of replacing an employee at roughly 40% of salary for frontline roles, 80% for technical roles, and up to 200% for leaders. A cheap hire that leaves in eight months is not a saving. See more recruitment statistics for the wider 2026 picture.

How the arithmetic actually runs

Agency costs scale linearly with hires. Twenty hires at an average $80,000 salary and a 20% fee is $320,000, and next year it is $320,000 again.

In-house costs are flat but lumpy. Two senior recruiters, fully loaded with tooling, is a substantial fixed commitment, plus two to four months of vacancy before either is productive.

Classic RPO costs are contracted and predictable, but usually locked for the term.

Embedded RPO costs are flat per recruiter per month, so the effective cost per hire falls with every hire that team closes. The break-even against agency fees typically arrives somewhere around five to seven hires a year at mid-market salaries; below that, use an agency.

What the numbers look like in practice: Nebius

Abstract arithmetic is easy to argue with, so here is a real engagement with real figures.

Nebius, an AI infrastructure company, entered 2026 at around 700 people with a twelve-person internal talent team, heading for close to 3,000 people by year end against a roadmap of 361 senior hires across data center, GTM, R&D and corporate functions. Two options were on the table. Build the specialist team internally, at roughly six months per senior recruiter to reach productivity. Or scale agency usage, at $46,000 per senior placement.

The agency arithmetic was the problem: 361 hires at a $46,000 baseline is a $16.6 million annual cost. Matchr's twelve-month embedded fee for the same plan is approximately $1.9 million.

We embedded thirteen senior talent partners across four business lines. They were productive in week one, the first hires landed inside month one rather than month six, and four months in the engagement had displaced $1.3 million in agency spend, eliminating about 80% of the agency motion on the roles we work. Projected across the full year, that is $14.7 million in net displaced spend. Read the full Nebius story.

The honest caveat: those numbers work because Nebius hires at volume, in a specialist market, with an internal team that owns the plan. Run the same model at five hires a year and the maths stops being flattering. Volume is what makes embedded cheap, and that is true no matter who is selling it. More engagements and numbers are in our customer stories.

The hybrid model: what most companies actually run

In practice, very few companies run one pure model. The hybrid model is not a fifth option so much as an honest description of reality: a permanent core plus external capacity where it earns its place.

Common combinations that work:

• In-house core plus embedded surge. A permanent team owns process, brand and business-as-usual hiring; embedded recruiters absorb the peaks, new markets, or unfamiliar profiles.
• In-house plus agency for the edges. Keep volume hiring internal; use retained search for the two executive roles a year where confidentiality matters.
• Embedded plus agency. Embedded for the plan, agency for the genuinely rare one-off nobody in the market can source.
• Classic RPO plus in-house strategy. The provider industrialises volume hiring; a small internal team owns workforce planning and employer brand.

The failure mode of hybrid is the same everywhere: nobody owns the whole. When an agency, an RPO and an internal team all work the same requisition list without one accountable owner, candidates get contacted twice, hiring managers get three different processes, and everyone reports success while time to hire climbs. If you run hybrid, name the owner first.

How Matchr does embedded RPO: Predictable and Flexible

We run embedded RPO in two commercial shapes, because hiring plans come in two shapes. The recruiters and the way they work are identical; only the way you pay changes.

Predictable RPO: fixed monthly cost

Senior talent partners and sourcers embedded in your team at a fixed monthly cost, with clear KPIs and structured delivery, and no placement fees. Your cost is the same whether the team closes three roles that month or eight, which means cost per hire falls as the plan runs. This is the right shape when you know roughly how much you are hiring and want the budget line to stop moving. See how Predictable RPO works.

Flexible RPO: base fee plus a fee per hire

An embedded talent team that scales with your hiring needs: a lower fixed base fee plus a fee per hire. Capacity flexes up and down without committing to a fixed team. This is the right shape when volume is genuinely uncertain, when the plan depends on a funding round or a launch that has not happened yet, or when you want more of the cost to sit on the results side. See how Flexible RPO works.

The choice between them is a question about your certainty, not your budget. Predictable is cheaper per hire if the volume materialises. Flexible carries less risk if it does not. We would rather put you on the one that matches your plan than the one with the larger contract value, because engagements that stop making sense do not get renewed.

How to choose: five questions

Work through these honestly and the answer usually picks itself.

1. How many hires, over how long? One to five a year points to agency. Continuous hiring points to embedded. Hundreds of similar roles points to classic RPO. Steady and predictable points to in-house.

2. Is the problem capacity, capability, or process? Capacity means you need more hands: embedded or agency. Capability means you have never hired this profile: embedded or a specialist agency. Process means hiring itself is broken: classic RPO, or fix it before you buy anything.

3. Who owns hiring internally? If the answer is nobody, no external model will save you. Fix that first; it is cheaper than any of the options here.

4. How certain is the plan? Signed and funded points to a fixed model. Dependent on a round or a launch points to a flexible one. If roles are not approved, wait.

5. What do you want to own at the end? If you want the capability to stay with your team, embedded or in-house. If you would rather someone else own it permanently, classic RPO or agency.

The talent leaders we interviewed for our Talent Acquisition Trends 2026 report, 35 of them, kept returning to the same point: the model matters far less than whether the hiring process underneath it is real. Structure first, then resourcing.

When each model is the wrong choice

The fastest way to choose is often elimination.

• In-house is wrong when volume is uneven, when you need capacity this month, when you are entering a market you do not know, or when you cannot justify permanent headcount against a plan that might change.
• Agency is wrong when you are hiring continuously, when fees have become your largest talent line, or when candidate experience and employer brand matter more than speed on a single role.
• Classic RPO is wrong when you want to keep owning your process, when your plan changes faster than a contract cycle, or when the scope is too small to industrialise.
• Embedded RPO is wrong when you hire twice a year, when you want a shortlist with no involvement, when nobody internally owns hiring, or when the plan is not approved yet.

Any partner who tells you their model is right for every company at every stage is selling, not advising. That includes us.

Frequently asked questions

What are the main recruitment models? In-house, recruitment agency, classic RPO, and embedded RPO. Most companies run a hybrid of two or more: a permanent core plus external capacity where it is needed.

What is the difference between RPO and a recruitment agency? An agency sells candidates and is paid per placement, usually 15% to 25% of first-year salary. An RPO sells recruitment capability and is paid for capacity or process. Agencies are cheaper for one-off roles; RPO is cheaper once hiring is continuous.

What is the difference between classic RPO and embedded RPO? Embedded RPO is a form of RPO. Classic RPO takes a defined process and runs it externally, often on a multi-year contract. Embedded places recruiters inside your team to run your process, in your systems, usually on 3, 6 or 12-month terms, and leaves the capability with you.

What is hybrid RPO? Running more than one model at once, most commonly a permanent in-house core with embedded or agency capacity layered on for peaks, new markets, or specialist roles. It works when one person clearly owns the whole requisition list.

Which recruitment model is cheapest? It depends entirely on volume. Below roughly five hires a year, agency wins. Once hiring is continuous, fixed-fee embedded RPO wins and the gap widens with every hire. In-house is cheapest at steady, predictable volume, once you count the full loaded cost.

What is the difference between Predictable RPO and Flexible RPO? Both are embedded models from Matchr. Predictable RPO is a fixed monthly cost with no placement fees, best when you know your volume. Flexible RPO is a lower base fee plus a fee per hire, best when volume is uncertain.

When should I build an in-house recruitment team? When hiring is steady and year-round, when you can justify permanent headcount against an approved plan, and when the institutional knowledge of your business is worth more to you than speed to capacity.

So which recruitment model should you choose?

Start with volume and certainty, because those two answers eliminate most of the options. A handful of hires a year is an agency question. Steady year-round hiring is an in-house question. Hundreds of similar roles is a classic RPO question. Continuous hiring, a new market, or a plan moving faster than you can staff it is an embedded question.

We build our business on embedded RPO because, across the companies we work with, it is the model that most often solves the problem they actually have: not a shortage of CVs, but a shortage of capable people close enough to the work to hire well. That is a genuine position, and it is not universal. At low volume we will point you at an agency. If your process is broken, we will tell you to fix that before you buy anything from anyone.

The model is not what makes hiring work. Structure, judgment, and a hiring team that shows up are what make it work. The model just decides who is standing close enough to apply them.

Want to work out which is right for your plan? See how our embedded model works, or let's get in touch and talk it through against your actual numbers.