Creaming and parking

Creaming is concentrating effort on the people closest to a job. Parking is what happens to everybody else. They are usually written as a pair because they are not two behaviours, they are one decision seen from the two ends of a caseload: a fixed number of adviser hours moving towards the cases that convert, and away from the cases that do not.

Parking is the half that does the damage and the half that is hard to see, so it is worth describing precisely. A parked participant is not refused, not sanctioned and not told anything. They are maintained. An appointment every four or six weeks that satisfies the contract minimum service standard, a group session instead of one to one time, a link to a job board, and no adviser ever picking up the phone to an employer on their behalf. They are on the caseload, they are attending, the file is in order and nothing is happening.

From the chair it is nearly undetectable, because a service that has quietly stopped trying looks exactly like a service that is not going well.

It is arithmetic rather than malice, and any account that skips that will not survive contact with the people doing it.

An adviser has a fixed number of hours. The organisation is paid when somebody starts work and stays there. Suppose three hours of good work converts a person who was nearly job ready, and forty hours might convert somebody with a fluctuating health condition, a gap in their record and no recent references, with perhaps a one in five chance of success. Under a performance floor written into the contract, a manager allocating scarce hours moves them towards the first case. Every time, for perfectly defensible reasons, in a meeting where nobody says anything that would look bad in writing.

Multiply that across a few thousand advisers and you have a national pattern that nobody authored. The people who wrote the funding rules never meet the person being parked, which is the general shape of this problem: the party with the most control over the decision has the least sight of what it costs.

The people most likely to be parked are precisely the people the programme was funded to reach. Somebody who would have found work anyway still generates a payable outcome, so the public purse frequently buys a result it was going to get for nothing, while the person it was actually buying help for receives an appointment every six weeks.

That is the deadweight problem and the fairness problem arriving together, and they compound: the measured performance of the programme can be perfectly respectable while its additional effect, the part that would not have happened anyway, is small.

There is also a cost that cannot be recovered. A person who has been parked twice has learned that the process is theatre. The next provider inherits somebody compliant, polite and entirely disengaged, and records it as attendance.

The tell is effort, not outcomes. Plot adviser hours per participant against assessed distance from work. If the line slopes down as difficulty rises, the organisation is parking people regardless of what anybody intends or believes.

Two cautions about that measure. The first is that once effort is reported, effort gets recorded rather than spent, so the number needs checking against something else: a small random sample of case files read against what the participant says happened, by somebody outside the performance line. The second is that a change protecting the hardest third can simply move the harm to the middle third, who were converting, so both groups have to be watched at once.

The other honest signal is the repeat referral rate: people who come back within a year of an outcome being claimed for them. It says something about whether the outcomes were real, and almost nobody publishes it.

Differential pricing that reflects the real cost difference between cases. Most programmes that tried this made the differential too small to change the arithmetic, which is worse than not trying, because it allows everyone to say the problem was addressed.

Minimum service standards audited on what was delivered rather than on what was recorded. Publishing outcomes and effort by cohort, so a parked group becomes visible instead of being averaged into a headline. And measuring further out, so a fast conversion that does not last stops looking better than slow work that does.

Notice how few of those are in the gift of a provider acting alone. That is the uncomfortable finding, and it is why a provider arguing publicly for differential pricing is the most credible witness available: it is arguing against its own short term revenue.