RPO gives you capacity. It can’t give you a decision.

The case for RPO is usually made in the language of volume. Hiring has spiked, the internal team is underwater, agencies are producing noise, and something has to give. An embedded team that works to your scorecard and reports into your leaders is a genuinely good answer to that problem, and when the problem really is capacity it works quickly and visibly.

The difficulty is that a large share of stalled hiring isn't a capacity problem at all. It's a decision problem. Candidates are arriving, and they're waiting. Feedback takes 6 days. The panel can't align on the bar. The offer needs an approval from someone who's travelling. Every one of those is a queue in front of a decision, and none of them is relieved by generating more candidates faster.

Buy RPO for a decision problem and the visible result is more candidates waiting longer. Time-to-fill doesn't move, quality complaints go up because the pipeline is now larger and less individually curated, and 18 months later the conclusion is that the provider underperformed. The provider did what they were bought to do. The constraint was somewhere else.

This is a diagnosable question, and it takes about a week to answer properly. Take the last 2 quarters of filled and unfilled roles, and split elapsed time into the intervals that belong to different owners: requisition open to first slate, slate to first interview, interview to feedback, final interview to decision, decision to offer out, offer out to accept.

The shape of that distribution tells you what you're buying. If most of the elapsed time sits in the first interval, you have a sourcing and screening constraint, and capacity will genuinely help. If it sits in the feedback and decision intervals, which is where a surprising share of it turns out to be in most organisations that run this exercise honestly, then recruiting isn't the bottleneck and expanding it won't move the outcome.

Do this before an RPO conversation rather than during one, because the answer also determines what a good engagement looks like. A provider brought in to relieve a screening constraint and a provider brought in to compress decision latency should be given different mandates, different metrics and different access. Sold the same engagement, only one of them can succeed.

Every RPO engagement runs on a scorecard, whether or not anyone writes one down. It's the shared definition of what a good candidate looks like, and it's the thing the provider is actually being asked to reproduce at volume. When it's precise, an embedded team can hit it consistently within weeks. When it's vague, they can't, and the resulting friction gets described as a quality problem.

Vague scorecards are the norm, and not because anyone's careless. Internal teams carry the bar tacitly. They know it when they see it, they've absorbed the preferences of specific hiring managers, and they've never had to write it down because they never had to hand it to anyone. Handing that to an external team makes the tacit part visible for the first time, and quite often reveals that different hiring managers were holding different bars.

This is the most valuable and least advertised effect of a well-run RPO engagement. Specifying the scorecard forces an organisation to agree on a standard it had been improvising. Organisations that treat that as an annoying onboarding step get a mediocre engagement. Organisations that treat it as the work get a better hiring process than the one they had, whoever ends up running it.

The structural choice that determines most RPO outcomes is whether the embedded team sits inside the decision loop or feeds it from outside. Outside looks tidier. The provider produces slates, the internal team decides, the boundary is clean and easy to govern. It also guarantees the provider can never influence the interval where most of the time is lost.

Inside means the RPO recruiters attend the calibration conversations, hear the disagreements between hiring managers directly, chase feedback with the standing of someone who works here, and are trusted to push back on a brief the market is refusing to answer. That takes real access and a degree of institutional trust some organisations are uncomfortable extending to an external team.

The discomfort is worth naming, because it's usually the actual constraint on the engagement rather than anything about the provider. An embedded team that isn't allowed to behave as though it's embedded is an agency with a longer contract and a desk. The word in the model is doing no work.

RPO agreements are almost always one-sided in an odd way. They specify in detail what the provider must deliver, and say nothing about what the client must do for that delivery to be possible. Slates within a number of days, quality thresholds, reporting cadence, all defined. Feedback turnaround, interviewer availability, decision authority, approval routing, unmentioned.

Since the client's side of that exchange is where a large share of elapsed time lives, this is like contracting for a delivery time and leaving the loading dock hours unspecified. The fix is straightforward and slightly awkward to propose. Write the client's obligations into the agreement with the same precision, and give them a named owner senior enough to enforce them on hiring managers who miss.

Providers rarely ask for this, because asking sounds like anticipating your own failure. Clients who offer it get better engagements, because it turns the conversation from a supplier relationship into a shared process with 2 sets of commitments, and because it surfaces, before signature, whether the organisation is actually prepared to change how it decides.

An embedded team is very good at a specific, valuable set of things. Absorbing volume without a permanent headcount commitment. Bringing market intelligence an internal team can't maintain part-time. Holding a defined bar consistently across a lot of roles. Taking coordination load off people whose time is worth more elsewhere. Bought for those, the engagement usually pays for itself and the internal team gets its judgement back.

It's not good at making decisions the organisation hasn't made. It can't agree the bar on your behalf, compel a hiring manager to give feedback, or make a mis-set compensation band competitive. Those are internal problems, and they stay internal problems after the contract is signed: more visible, better instrumented, and still yours.

The diagnosis is the whole thing. An organisation that knows where its days go can buy the right instrument and hold the provider to a fair standard. One that hasn't looked will buy capacity for a decision problem, and 18 months later will have learned something expensive about its own process rather than about RPO.