Illustrative · Kestrel Pathways is a constructed operator, not a real provider. Its shape, scale and contract mix are drawn from how the UK employability sector actually works, and every figure is indicative unless it carries a source.
Operations
Whether the job lasts, and what nobody counts
A job start is an event. A job that lasts is a different thing, produced by different work, and measured by nobody after week twenty six. Then there is the second loss, which has no name in this sector: everything the system could have produced and did not.
A job start is an event that happens in an afternoon. A job that lasts is a different product altogether: it is made from fit, money that works, an employer who was prepared, health that was accounted for, and a life stable enough underneath it. Those are different pieces of work, done at different times, by people paid for the first one.
The contract measures sustainment at four, thirteen and twenty six weeks. That is roughly six months, which is long enough to be a genuine test and short enough to be gamed, and it is followed by a cliff after which nobody measures anything at all.
Chapter 1 · Failure
What actually breaks a job
Six ordinary reasons a placement fails inside a year. Note how many of them are decided before the person starts, which is the useful finding: sustainability is mostly built during the placement decision and the fortnight either side of it, not repaired later by in-work support.
Note also the last row. Support stopping when payment stops is the only cause on the list that is created entirely by the funding model, and it is the cheapest to fix.
| Why the job ends | How it shows up | When it usually happens | What would have held it |
|---|---|---|---|
| It was never the right job | Taken because it was available and the adviser needed a start, not because the person could do it for a year. | Weeks two to eight. Long enough to discover the work, short enough that nobody has invested in them. | Matching against what the person can sustain rather than what they can be placed into today. It costs adviser time and it lowers this month figure. |
| The money does not work | Hours too few or too variable, travel cost eating the gain, a benefit taper that makes the extra shift worth almost nothing. | The first two pay packets, when the arithmetic becomes real rather than projected. | Working the actual numbers with the person before they accept, including travel and childcare and the taper, rather than after. |
| Health moves and the job does not | A condition that fluctuates meets a rota that cannot. Nobody at work knows, because disclosing felt unsafe. | Month two onward, and typically at the first bad week. | An adjustment agreed before the start, and an employer who was told enough to be ready. That conversation is the work, and it is nobody funded job. |
| Life was never stable underneath it | Housing, childcare, a court date, a caring responsibility, an unresolved debt that generates a crisis on a Tuesday. | Any time, and often at the first collision between work and the thing that was already there. | Fixing the thing underneath before placing, which looks like delay on a performance report and is the only durable version. |
| The workplace was hostile or chaotic | A manager with no interest in an inducted person, no training, shifts changed by text on the day. | Weeks one to twelve, and it repeats with every candidate sent to that employer. | Knowing your employers well enough to stop sending people to the ones that churn. Provider account managers usually know exactly who these are. |
| Support ended when the payment did | The programme stops calling. The first real difficulty is faced alone, by somebody with no recent experience of being in work. | Month seven onwards, which is to say immediately after the last payment point. | In-work support scheduled against risk rather than against payment dates. Cheap, unfunded, and the single most avoidable cause on this list. |
summary: the ordinary reasons a placement fails inside a year, as they present rather than as they are recorded. Exit reasons in provider systems are a short list of codes chosen for auditing, so most of the first column arrives as "left employment, reason unknown".
Chapter 2 · The window
The twenty six week cliff
Payment ends at twenty six weeks. Contact ends at twenty six weeks. And the risk of the job failing does not end at twenty six weeks, because the things that break a placement in month nine are the same things that break one in month three, minus the attention.
What happens next is invisible to everybody. The provider does not receive employment data after the payment window, the commissioner holds it and does not routinely share it back, and the participant only reappears if they are referred again. A person who lost the job in month nine and spent four months out of work before a second referral is recorded by the system as two separate successes and one new customer.
This is the denominator problem from the participant page, expressed as money. The programme is judged on durable employment and paid on a six month proxy of it, so six months is what the organisation produces, with great skill, in enormous volume.
Chapter 3 · The distinction
Sustained is not the same as sustainable
A sustainment claim says a person was in work on a particular date with evidence to prove it. It does not say the job is secure, that the hours are stable, that the money is better than the alternative, or that the person expects to be there at Easter.
Two placements can produce identical claims and be entirely different objects: one is a person in the right work with an employer who wanted them, the other is a person holding on to something unsuitable because leaving would restart a benefit claim they cannot face restarting. The second is not a success. It is a person trapped by the same system that placed them, and the contract cannot tell the two apart.
The cleanest available signal of the difference is the repeat referral rate: people who come back within a year of an outcome being claimed for them. It is knowable, it is uncomfortable, and almost nobody publishes it.
Chapter 4 · Missed
The opportunity nobody counts
Parking is the loss the sector has a word for. These five are the losses it does not, and together they are almost certainly larger.
Each one has the same structure: a real gain that the funding model has no field for, so nobody is accountable for producing it and nobody notices its absence. That is not a failure of effort. It is what happens when an organisation is instrumented to see one kind of value.
| What is missed | How it happens | Who loses | What it would take to see it |
|---|---|---|---|
| The job below the person | A qualified or experienced person is placed into whatever converts fastest, because a start is a start and the clock is running. | The participant, who takes a pay cut that becomes their new baseline, and the employer who never met the candidate they needed. | Wage and occupation at placement compared with the person prior role and qualification. The data exists at referral and is never used this way. |
| Progression once in work | Nobody is paid for the second step, so nobody works on it. The person who could have moved up in a year stays where they landed. | The participant, permanently, and the exchequer, which keeps topping up a low wage it could have stopped topping up. | Earnings at twelve and twenty four months against earnings at placement. It needs data the provider does not receive. |
| The employer used once | An account manager wins a vacancy, fills it, and moves on to the next target. The relationship that could have placed ten people places one. | The provider, which rebuys the same relationship every quarter, and every future participant who needed that door. | Placements per employer per year, and the survival rate of placements by employer. Both are already in the system and neither is reported. |
| The person who left at week three | Disengagement is recorded as a participant choice. Often it was a missed call, a letter that did not arrive, or an appointment that clashed with the only childcare available. | The participant, who is marked as unwilling, which follows them into the next programme. | Contact attempts and their outcome before an exit is coded, sampled by somebody outside the performance line. |
| What the programme learned | Outcome data goes to the commissioner and stops. Nothing comes back, so no provider can tell which of its own practices actually worked. | Everybody, including the commissioner, which buys the same uncertainty at every re-procurement. | Employment data shared back at twelve and twenty four months, by cohort. It is the cheapest reform available and the one nobody owns. |
summary: losses the contract has no field for. None of them appears in a monthly return, so none of them is managed, and the sector improves on the parts it can see while these stay exactly where they are.
Chapter 5 · Inside the walls
What could be measured from tomorrow
Five numbers, all producible from data Kestrel already holds, none requiring the commissioner permission. Repeat referral rate within twelve months of a claimed outcome. Wage and occupation at placement against the person prior role. Placements per employer per year, with survival rate by employer. Contact attempts made before an exit is coded as disengagement. And in-work support contacts scheduled by risk rather than by payment date, reported as a rate.
Every one of them will look worse than the contract numbers, because they measure the thing rather than the proxy. That is the point of measuring them, and the reason a company only does it if it meant what it wrote on its values page.