A probationary period is a contractual window in which an employer can end an employment relationship with reduced procedural burden. It is not an assessment. It produces no standardised score, it observes only the candidates you already hired, and it is almost never acted on: US federal agencies took adverse action against 0.4% of newly hired supervisors during their probationary year.
Most hiring managers treat probation as a second screening stage, which quietly licences a weak first one. "If they're not good, we'll know in six months" is the load-bearing assumption under a lot of under-evidenced offers. Four jurisdictions have now put a hard edge on that window, and the evidence on what happens inside it is unflattering.
The six-month convergence
The practical observation period before an employment relationship becomes expensive to exit is settling at roughly six months across very different legal systems.
- European Union — Article 8 of Directive (EU) 2019/1152 requires that where an employment relationship is subject to a probationary period, "that period shall not exceed six months." Recital 28 records that a substantial number of member states already set a general maximum of between three and six months. Longer periods are permitted only exceptionally, for example for managerial, executive or public-service posts.
- United Kingdom — Under the Employment Rights Act 2025, the qualifying service needed to bring an unfair dismissal claim falls from two years to six months on 1 January 2027, and the cap on compensatory awards is removed entirely. The Department for Business and Trade's own summary confirms both changes and the date; the CIPD notes the same six-month threshold applies to employees who have that service on 1 January 2027.
- India — The Model Standing Orders, 2026, notified on 8 May 2026 under section 29 of the Industrial Relations Code, 2020, prescribe a six-month probation extendable by up to three further months on performance assessment.
- Philippines — Labor Code Article 281, renumbered as Article 296, allows probation "for a period not exceeding six months from the date of hiring," and a worker allowed to work after it "shall be considered a regular employee." Regularisation happens by operation of law, not by a manager's decision.
Two of these raise the price of being wrong at the end of the window, not just the speed. Uncapped UK awards from 2027 mean a badly handled dismissal at month seven no longer has a known worst case, and Philippine regularisation is automatic if nobody acts.
The 0.4% problem
The deeper issue is that the probation decision mostly isn't made at all. The US Merit Systems Protection Board examined the supervisory probationary period and found that in fiscal year 2007, 28,731 supervisors were hired and 117 actions were taken across all federal agencies for unsuccessful completion of probation: 73 reassignments to non-supervisory roles and 44 separations. That is four-tenths of one percent.
The same study found only 64% of first-level supervisors had been told they would serve a probationary period at all, and fewer than 47% said their probationary performance was actually used to decide whether they should keep the role. MSPB's own framing is the useful part: research on new leaders typically reports failure rates of 20% or higher, so a 0.4% action rate is not evidence that selection was excellent. It is evidence that the mechanism went unused.
That pattern is structural rather than negligent. Continuation is the default and requires no action; ending probation requires paperwork, a difficult conversation, and a manager publicly conceding that the person they chose was the wrong choice.
Even when the decision is made, the evidence is thin
Probation feels like the gold standard because it observes real work. But the instrument doing the observing is usually a single manager's judgment, collected under the worst possible conditions for accuracy.
Administrative ratings are the least reliable kind
A 2019 meta-analysis by Salgado and Moscoso in Frontiers in Psychology separated supervisory ratings by purpose and found interrater reliability for overall job performance of .45 when ratings were collected for administrative purposes (K = 18, N = 3,632), against .69 when collected for research (K = 201, N = 12,814). Task performance was worse still: .38 administrative versus .62 research. A probation verdict is the most administrative rating in the system.
And they are predictably lenient
Jawahar and Williams' meta-analysis in Personnel Psychology (k = 22, N = 57,775) found that ratings gathered for administrative purposes ran about one-third of a standard deviation higher than ratings gathered for research or development, d = .32 corrected, rising to d = .41 in field settings where real supervisors rated real subordinates. The direction of the error on a probation sign-off is not random. It is "pass."
Six months does not generalise to year three
Even accurate early performance is a limited forecast. Sturman, Cheramie and Cashen's meta-analysis found the stability of performance ratings over a one-year lag ranged from .85 to .67, and test-retest reliability from .83 for subjective measures in low-complexity jobs down to .50 for objective measures in high-complexity jobs, with correlations declining as the interval widened. The more complex the role, the less a six-month read tells you about year three.
There is no comparison group
Probation observes only people you hired. It can say something about this person's first few months; it says nothing about the applicants your front door rejected. The costly selection errors are invisible to it by construction, which is exactly why probation cannot substitute for pre-hire measurement.
What a six-month window actually requires
- Put the decisive evidence before the offer. Anything you could have measured with a work sample or a scored scenario should not be left to month five. AssessAll's AI-graded scenario assessments with integrity bands produce a documented, standardised score at application stage on pay-as-you-go credits, which is the kind of record a compressed window needs.
- Define the probation decision as a criterion, not an impression. Three to five behaviourally anchored outcomes, each with a named evidence source, written into the offer letter rather than reconstructed in month six.
- Schedule the decision earlier than the legal threshold. Structured reviews at weeks 4, 8 and 12 make a month-four decision possible. A review calendar that first produces evidence at month six has already run out of room.
- Separate "did not perform" from "was not enabled." Record what onboarding, system access, manager time and task variety the person actually received. A mis-hire conclusion that cannot survive that record is a management finding, not a selection one.
- Use more than one rater. The .45 figure above is a single-rater number. Two independent raters against the same anchors is the cheapest available reliability gain.
- Treat extension as a decision, not a drift. Where the law provides for it, as India's rules do, an explicit extension with written reasons beats letting the clock run.
- Keep contemporaneous records. With the UK compensation cap gone from 2027, documentation quality moves from an HR preference to a financial control.
When probation is genuinely the right instrument
Probation earns its place for constructs that are hard to observe before hire: sustained reliability and attendance over months, response to corrective feedback, safety behaviour under real conditions, and whether someone works well with a specific team rather than a hypothetical one. It is also the honest fallback where no valid pre-hire instrument exists for a narrow construct. The error is not having a probationary period. The error is letting it carry a decision that a scored, standardised, pre-hire assessment should have made — and the regulatory direction of travel in the EU, UK, India and the Philippines is to give you less room to make that mistake.
Six months is not a screening stage; it is a window in which you will probably do nothing, measured by an instrument with .45 reliability and a known bias toward passing. Move the decisive evidence to the front door, and use probation for what only probation can see. Evidence standards for both are set out in the *Standards for Educational and Psychological Testing* and the EEOC's guidance on employment tests and selection procedures.