Ask a talent acquisition leader how permanent hiring is going and you will get numbers. Time to fill, offer acceptance, source of hire, drop-off by stage, all of it down to the day.
Ask the same person how the contractor population is going and you will usually get one number: spend.
That gap is the whole problem. Contractors, freelancers, and temps are a large and growing share of the working headcount at most companies. A group that size, governed by a single budget line, is not a program. It is an assumption.
Here are the five metrics that tell you whether yours is under control.
One thing before the list. While researching this piece we went looking for industry benchmarks on all five, and found solid survey data for two of them. The rest is mostly vendor content citing other vendor content, and in a couple of cases figures that get repeated everywhere trace back to a survey from 2014. So we are doing this differently: where a real benchmark exists, we name the source. Where one does not, we say so and score against full coverage instead. Three of the five fall into that second category.
TLDR
- 55% of companies estimate up to 20% of contingent workforce spend falls outside their authorized program. The median sits at 11% to 20%, and 9% estimate more than half (SIA Workforce Solutions Buyer Survey 2025).
- Among high-performing programs, half report off-program spend of 10% or less. Among laggards, over a fifth report 41% to 50%.
- In light industrial staffing, top performers reach redeployment rates of 40% to 60%, while firms below 25% are paying to replace workers they already had (SIA).
- The US Department of Labor published a proposed rule on 26 February 2026 weighting two core factors: control over the work, and opportunity for profit or loss. Comments closed 28 April 2026 and no final rule has issued as of September 2026 (DOL).
- From 6 April 2026, UK agencies and end clients can be liable for PAYE underpayments where workers are supplied through umbrella companies, even where the umbrella runs payroll (RSM UK).
- Three of the five metrics below have no credible published peer benchmark. We say so rather than inventing one.
Score your own program
Five sliders, each tagged with the strength of the benchmark behind it. It runs entirely in your browser, nothing is sent or stored, and there is no form.
How much of your contingent spend runs outside the program?
This is the one metric in the set with genuinely good data behind it, and it is the one most worth knowing.
SIA's 2025 Workforce Solutions Buyer Survey asked buyers to estimate how much contingent workforce spend falls outside their authorized program. More than half, 55%, put it at up to 20%. The median lands at 11% to 20%. Nine percent estimate that more than half their contingent spend is off-program.
The spread by program maturity is the interesting part. At high-performing firms, half report 10% or less. Among laggards, over a fifth report between 41% and 50%.
The off-program portion is rarely agency supply, because agency supply arrives with invoices and a vendor record. It is direct and independent contractors: people engaged by a hiring manager, paid through accounts payable or a departmental card, never registered anywhere a workforce report would find them.
This is the metric that makes every other metric unreliable. You cannot report a redeployment rate for a population you cannot enumerate.
What share of engagements have a current classification assessment?
No credible public peer benchmark exists here. We looked. Score yourself against full coverage instead, because full coverage is the only defensible standard.
Two words do the work: documented and in date. A determination made at kickoff two years ago, against a test that has since changed, is a piece of paper, not a defense.
The change is live. The DOL published a Notice of Proposed Rulemaking on 26 February 2026, weighting two core factors, the degree of control over the work and the worker's opportunity for profit or loss, alongside additional factors including skill, permanence, and whether the work is part of an integrated unit of production. The comment period closed on 28 April 2026 and a final rule is expected to follow.
You do not need the final text to act. Control and opportunity for profit or loss are assessable today, from contracts and SOWs you already have. The engagements that fail on control are the same ones that would have failed under any version of this test.
In the UK the direction is the same. From 6 April 2026, agencies and end clients can be liable for PAYE underpayments where workers are supplied through umbrella companies, even where the umbrella operates payroll. Liability moved up the chain. The risk did not go anywhere.
A note on numbers you will see elsewhere: composite figures for what a single misclassification "costs," usually quoted as a range running into six figures, are almost entirely vendor marketing rather than published government data. The underlying components are real and checkable. The tidy aggregate is not.
How long does it take to get someone compliantly started?
Also no published industry benchmark. Contractor time-to-start is not systematically surveyed, which is itself telling.
Measure it end to end. Not from requisition approval, and not from the point your team gets involved. From "we want this person" to their first compliant working day: approval, contracting, classification check, right to work, systems access.
For a sense of scale, CXC describes an integrated contractor workflow cutting a three-to-four week process down to five to seven days, and notes that contractors in some technical environments wait considerably longer before doing meaningful work. Treat that as illustrative rather than as a benchmark, because that is what it is.
Worksome gets contractors to a compliant start in three days. That is our own standard, not an industry average, and we would rather label it honestly than dress it up as research.
The cost of a slow process is not mainly the delay. It is the candidates you never hear about, because the specialist choosing between you and a client who can start them Monday is not going to explain the decision to you.
What share of finishing contractors do you re-engage?
Redeployment is the highest-return metric on this list and the least reliably benchmarked.
The most useful published figures come from SIA's reporting on light industrial staffing, where top performers reach 40% to 60% and firms below 25% are quietly paying to replace people they already had. Enterprise contingent programs are a different population, so treat those numbers as directional rather than matched to you.
You will also see a widely repeated claim that fewer than 6% of firms measure redeployment at all. It traces back to a 2014 survey. It may well still be directionally true, but it is too old to build a case on and we are not going to pretend otherwise.
We suggest measuring over a 14-day window between assignments. That is our recommendation, not an industry standard. Thirty or sixty-day windows inflate the number and hide exactly the gap you are trying to find.
The economics are not in dispute even where the benchmarks are. A redeployed contractor needs no sourcing, no screening, no cultural assessment, and no ramp. Paying full acquisition cost to replace someone who finished an assignment three weeks ago is the most expensive habit in contingent hiring, and it is almost always a systems failure rather than a judgment call. Nobody decided not to rehire them. They just fell out of view.
Could you contact last year's contractors today?
No published peer benchmark, and scored against full coverage.
Of the contractors who finished an assignment in the last twelve months, what share could you find in a system, contact, and re-engage right now, with compliance records still in date? Not "someone probably has their email." Findable, contactable, and clean.
This metric is on the list because it sets the hard ceiling on the one above it. A 60% redeployment rate is arithmetically impossible if you can only locate 30% of the people who might be redeployed. Teams that set a redeployment target without fixing data continuity first are managing a number they have already capped.
What does a good score look like?
The scorecard places your result on a five-step maturity scale, from Level 1, where there is activity but no real program, to Level 5, where coverage extends to every worker type and location. Those bands are ours. For independent research on how program maturity tracks with outcomes, SIA has published on exactly this, and the short version is that maturity predicts almost everything else.
Most programs that run these five metrics honestly land at Level 2 or Level 3. Level 3 means processes, policies, and metrics exist and get followed, which is a real achievement and also where most programs stop.
Two things matter more than the headline number.
Where the gaps cluster. Low scores on classification coverage and off-program spend are a different problem from low scores on redeployment and data continuity. The first pair is risk. The second pair is cost. Both are worth fixing, but only one of them shows up as a penalty notice.
Whether you could produce the numbers at all. If you had to guess at three of the five, the guessing is the finding. A program you cannot measure is a program you are hoping about.
Where to start
Start with classification coverage and off-program spend, in that order. They carry the most weight on the scorecard for the same reason they matter most in practice: they are the two gaps that generate consequences arriving from outside the business, on someone else's timetable.
Redeployment and data continuity come next, and they come as a pair. Fix the pool before you set the target.
Time-to-start usually resolves alongside off-program spend, because they are the same problem wearing different clothes. A hiring manager who needs a specialist next week and knows the official route takes a month will find another route, every time. Make the compliant route the fastest route and the workaround stops being worth the effort.