A business takes a chance on someone with no track record. It pays them, teaches them and gives them skills the market actually wants. Eight months later they walk out of the door to a job paying two-thirds more. It is not hard to see why the business feels it is owed something.

That feeling is entirely understandable. But it is not, on its own, enough to make the employee legally liable.

In Geeks Ltd v Watts [2026] EWCA Civ 889, handed down on 10 July 2026, the Court of Appeal set aside an order requiring a former trainee to repay £8,108 of training costs. Bean LJ gave the leading judgment, with Males LJ and Jeremy Baker LJ agreeing. The decision matters a long way beyond one IT company in Sutton, because the arrangement it struck down is one I see regularly in small business contracts, and because the argument the employer ran is the one on which most of these clauses are built.

The judgment is available from The National Archives.

What happened

Mr Watts joined Geeks in March 2019 as a trainee quality assurance engineer. He had a music degree, considerable debt and had applied for more than fifty jobs in the IT industry before this one came along. His salary was £18,000, rising to £20,000 and £22,000 in years two and three.

On the same day he signed his employment contract, he signed a second document: a Contract of Training Investment. It put a price of £8,108 on his first six months. The figure included a mentor costed at £60 an hour and a hundred hours of study and practice time at £13 an hour. The contract itself conceded that the table was a basic calculation intended to reach an appropriate figure rather than a precise appraisal of the cost to the business.

The mechanics are worth considering because they reveal what the clause was really for. The debt did not start reducing until Mr Watts had completed twelve months’ service. It was then written off at one eighteenth per subsequent complete month. The same document said that, if he repaid by staying, Geeks was under no obligation to increase his pay while the arrangement continued. If his employment ended before the debt had been cleared, the balance became payable in eighteen instalments. Redundancy was the only exception.

Mr Watts asked for a pay rise. It was refused. He resigned after eight months for a £30,000 job elsewhere, and Geeks sued him for the full £8,108. It won at trial before a deputy district judge and won again on appeal before HHJ Evans-Gordon. The Court of Appeal went the other way.

The argument that failed, and why it matters to your clients

Geeks’ best point in the Court of Appeal was that the Training Cost Debt existed from the outset and could be repaid either through continued service or by money. On that analysis, resignation merely determined the method of repayment. It did not create the debt, so the arrangement was not a restraint of trade.

It is a clever argument, and it is the intellectual foundation of much of the clawback drafting I see. Bean LJ dismantled it. If it were right, an employer could require a junior employee to repay a sum roughly equal to their entire first six months’ salary on the basis that they had been trained, with no enquiry into reasonableness at all. A clause reading ‘if you leave within twelve months you must repay your entire gross salary’ would be enforceable as a debt. The public policy problems, including the potential undermining of national minimum wage legislation, were described as obvious.

The doctrine looks at practical effect, not the label on the document.

“a question of substance, not form” (paragraph 49)

“financial disincentives are not exempt as a class from scrutiny” (paragraph 49)

The objection that the clause did not actually stop anyone leaving also got nowhere. No clause could do that because an employment contract cannot be specifically enforced against an employee. The question was whether the financial obligation could hamper the employee’s freedom to work.

There is a second casualty here that anyone advising in this area needs to know about. Many practitioners have been relying on Bacon J’s decision in Steel v Spencer Road LLP [2024] ICR 137 for the comfortable proposition that a clawback triggered by resignation is not a restraint of trade. HHJ Evans-Gordon said that, left to herself, she would have applied Steel to that effect. The Court of Appeal has now confined it. The result in Steel was correct on its facts, concerning a discretionary bonus repayable if the employee left within three months, but it does not support the wider proposition. Bean LJ also disagreed with part of Bacon J’s reading of Marshall v NM Financial Management Ltd [1997] ICR 1065. Advice based on a broader reading of Steel may now need revisiting.

Why this particular clause lost

The Court assumed, without deciding, that Geeks had a legitimate interest in maintaining a stable, trained workforce. That assumption is doing a lot of work and I would not lean on it. Bean LJ pointed out that the scope of that interest has gone undefined for thirty-three years since a case called Ingham v ABC Contract Services.

The clause failed on reasonableness instead, for two reasons.

The first is a drafting failure, and it is the one to take back to your templates. The obligation was indiscriminate. Apart from redundancy, it applied whatever the reason for departure. It caught resignation, dismissal on a week’s notice, a move to a competitor, a move out of technology entirely and a move to nothing at all. Bean LJ’s example was an employee leaving to care for a grandfather with dementia. A clause that catches the carer and the sacked employee on the same terms as the defector is not protecting a stable workforce. It is simply charging people for leaving.

The second is economic reality. Mr Watts was paid not much above the national minimum wage at the time. The effect of the clawback was that he was:

“reduced in retrospect to the equivalent of an unpaid intern” (paragraph 70)

That was not a clause going no further than reasonably necessary.

Two supporting points are worth having on file. The burden sits on the employer to justify the restraint, following Harcus Sinclair LLP v Your Lawyers Ltd [2022] AC 1271. Mr Watts did not have to prove that it was unreasonable. He also had no independent legal advice. The recital saying he had the opportunity to take advice was, in Bean LJ’s phrase, ‘neither here nor there’, because his evidence was that he could not afford it. The absence of advice was not decisive, but it was a relevant pointer away from reasonableness.

The instalment mechanism helped, but not enough. The Court accepted that repayment by monthly instalments was more reasonable than a lump sum demand or an acceleration clause, but a sensible payment mechanism cannot rescue an excessive underlying obligation.

The arithmetic did not escape either

The £8,108 calculation was not a ground of appeal, so nothing was decided about it. Bean LJ commented anyway, and the comments are a warning shot.

He would have found the £60-an-hour mentor rate highly questionable. On Mr Watts’ evidence, it was five or six times what the mentor was actually paid. More damagingly, he observed that, even accepting the claimed mentoring and study hours, the greater part of the working day was unaccounted for, as though what Mr Watts did during that time was of no value to the employer. That was highly artificial when the evidence showed that clients were already being billed for his services.

That is the point I would put in front of any client running one of these schemes. You cannot bill a client for an employee’s time and simultaneously tell a court that the same time was a cost of training them.

What I would actually change

I would not delete every training repayment clause on the back of this judgment. I would audit the purpose, scope and arithmetic of each one, because the version that survives Geeks v Watts is narrower than most templates. A clause recovering a share of genuine, evidenced expenditure is in a decent position. A scheme designed to make resignation financially painful is not.

The questions I would work through are these.

• Does the clause distinguish between reasons for leaving, or does it catch everyone except those made redundant?
• Is the sum tied to identifiable external costs, course fees, examinations, qualifications and third-party invoices, rather than to the ordinary business of employing someone while billing clients for their output?
• Is internal time costed at a defensible rate, or at a figure nobody actually pays?
• Does the taper run from when the cost was incurred, or is the whole sum parked behind a long service cliff, as it was here?
• What is the separate effect under the National Minimum Wage Regulations and the rules on deductions from wages?
• The Court identified the public policy concern but did not decide those statutory issues.
• Finally, is the clause reasonable given the employee’s bargaining position, likely salary and the fact that they probably will not obtain independent advice?

There is a strategic point too for businesses seeking to justify these schemes by pointing to what leavers went on to earn. Reasonableness is judged at the date of signing, without hindsight. Mr Watts’ eventual £30,000 salary could not determine whether the clause was reasonable when signed, and it did not rescue the arrangement. If a client’s defence of its clause depends on how things turned out, it is arguing the wrong case.

Do not overread it

This is not authority for a maximum clawback period, a mandatory taper formula or a rule that repayment only works where the employee joins a competitor. The judgment does not say that external training fees are irrecoverable. A proportionate clause tied to genuine, evidenced costs remains capable of enforcement. An ordinary condition requiring continued employment before a benefit is earned is not automatically a restraint, although its wording and practical effect still matter.

What the judgment does establish is broader and more commercially awkward: a substantial payment obligation triggered by departure can engage the restraint of trade doctrine even when it is drafted as an unconditional debt and expressly says that the employee is free to leave. Once the doctrine is engaged, the employer has to justify the clause, and many of the ones I see would struggle.

For small businesses, training clawbacks remain available. The agreement just has to look like fair cost recovery rather than a retention device in a debt-shaped hat. For those of us drafting them, the useful question is not whether the figure can be explained in a schedule. It is what the clause will do to an ordinary, low-paid employee who needs to leave for a reason that has nothing to do with competing. If the answer is that it would make them feel unable to leave, a tidier definition of ‘Training Cost Debt’ is not going to save it.

This article is general commentary on the judgment and not legal advice on any particular contract or dispute.

If you would like your employment or training contract templates reviewed in light of this decision, please get in touch.