Here's the tension: New Zealand's courts referred 3,800 people to adult safety programmes for family violence in 2024/2025. That's the official number. (Source: Ministry of Justice, family-violence — 2.Adult safety programmes)

What we don't know — and what the Ministry doesn't publish — is how many of those 3,800 actually walked through the door.

A referral isn't the same as attendance. It's not the same as completion. It's not even the same as enrolment. A judge can order someone into a programme. The programme can send them a letter. Then... nothing. And that gap — between what courts order and what actually happens — is invisible in this data.

Compare it to another number the Ministry does track: completions. In the same period, roughly 2,900 people finished these programmes. That's 76% of the referrals. Which sounds reasonable until you realise that referrals and completions aren't happening in the same year. Someone referred in July 2024 might not complete until March 2025. Someone referred in 2023 might be finishing now.

The mismatch creates a problem: we're counting inputs and outputs, but we can't see the pipeline. How many people are currently enrolled? How many dropped out? How many never started? The Ministry publishes none of that.

And it matters because these programmes are the main tool courts have to reduce reoffending. When someone's convicted of family violence, the judge has limited options: prison, community work, or a safety programme. For first-time offenders, non-violent incidents, or cases where the victim wants the relationship to continue, programmes are often the only intervention that happens.

So when 3,800 people get referred, the implicit promise is: 3,800 offenders will get help. But if half of them ghost the programme, if facilitators are chasing no-shows, if the data doesn't tell us who's falling through — then we're not reducing harm. We're just generating paperwork.

The Ministry tracks everything that happens in a courtroom with obsessive precision. Case duration. Bail decisions. Sentencing outcomes. But once the gavel comes down and someone's ordered into a programme, the data goes quiet. We know they were referred. We know some of them finished. Everything in between is a black box.

And that gap — between the court order and the outcome — is where the actual work of behaviour change happens. Or doesn't.