Cases for the practical state

Aunty Mavis, Whaea Hiki and Novapay

This week, we turn to Novopay and the lesson it offers about how the practical state breaks down when delivery is treated as a technical fix rather than a lived practice. The hypothesis supporting the theory of the practical state is that it is not an institution or reform agenda, but the middle ground where discretion is exercised, trust is earned, and legitimacy is sustained. Over the past eight weeks we have explored this in ordinary times, through iwi and hapū relationships, collaborative governance, and frontline craft. Today Novopay shows what happens when those truths are ignored.

In Aotearoa, public power has always swung between two poles: on one side, concentrated executive authority; on the other, a fragmented push for equality and equity through distributed delivery. As Matthew Palmer (2007) argued, these extremes are held together not by doctrine, but by pragmatism: a kind of constitutional glue that keeps the system functioning in the absence of formal resolution.

This series builds on that idea. It proposes that the holding work once done by pragmatism is now being performed, sometimes explicitly, sometimes instinctively, by what I call the practical state.

The practical state is not an institution or a reform agenda. It is a mode of governing that emerges where public servants and communities meet, not in the context of policy or protest, but in the contested middle where discretion is required, judgment is tested, and trust must be earned daily.

It is what Honig (1993) describes as democratic tension held in practice. It is Lipsky's (1980) street-level discretion, Lindblom's (1959) incremental muddling-through, and Ryan's (2011) negotiated space between mandate and meaning. It operates in the space where, as Pressman and Wildavsky (1973) demonstrated in their foundational study of implementation, policy intentions meet the complex realities of practice: where "great expectations" must navigate the intricate chain of actors, decisions, and circumstances that determine whether programs succeed or fail.

The practical state isn’t nostalgic. It’s just not stupid. It knows that when policy decisions are made, the change eventually lands on someone’s desk. And usually that person has never been part of the original korero or signed up to the perceived political policy problem, so when the change does land, that person better be prepared to roll with it.

Every school in Aotearoa knew who did the payroll. It wasn’t a system. It was Aunty Mavis. Or Whaea Hiki. Or the deputy principal, who was also ordering new library books, in between handing out lunch orders, getting sick kids home, and trying to navigate the latest health and safety obligations that had to be included in the new field trip forms.

Novopay didn’t fail because of software bugs. It failed because no one in power thought to ask how schools work. The state saw payroll. Schools saw trust. The system saw transactions. Aunty Mavis saw whānau being paid on time so they could pay their bills. Whaea Hiki saw good teachers being acknowledged for the extra time they were putting into their and her community.

The Novapay failure wasn’t just a glitch. It was a betrayal of the people who keep schools running, and of the practical wisdom they carry. It took seven years of planning and $182 million to build something that didn’t work. Another $45 million to stop it from hemorrhaging. And for what? So, a minister could say we had a modern payroll system? So the theoretical state could pat itself on the back while principals spent Christmas on hold trying to correct staff payslips?

The theoretical promise was seductive. A web-based payroll solution to replace the existing Datacom system, which, though stable for eight years, was written off as "dated" because it lacked online access and sophisticated analytics.

Talent2, the chosen vendor, was awarded the contract through a competitive tender process focused on technical specs and cost. No one seemed to pause and ask why the same vendor had so recently failed at NZ Post, requiring a total system rewrite after thousands of under- and overpayments. That prior failure was known. It was treated as a solvable nuisance, not a red flag.

This is a classic example of positivism: the belief that good design and rational planning will prevail over messy human systems. That implementation is just a matter of execution. That complexity can be managed through logic, and readiness achieved through training. And so, despite trial testing revealing 147 defects and over 6,000 issues, and despite over half the 731 trial users saying they weren’t ready, the system went live.

What happened next is still staggering. Five thousand people were underpaid. Some not paid at all. Personal data sent to the wrong schools. Some staff being overpaid by tens of thousands. Teachers with hyphenated names were paid twice as much. And all this in the very first pay round.

This wasn’t just a technical failure. It was a delivery model breakdown. One that ignored the relational and contextual truths of education payroll in Aotearoa: that teachers operate under multiple collective agreements, often on mixed hours or fixed terms. Schools adjust those arrangements to respond to the needs. That none of this can be captured in a system that doesn’t understand the concept of a reliever covering a teacher on half-pay for a public holiday. It was Mavis and Hiki, not the machine, that kept the system going by knowing what mattered.

The consequences were enormous and also intimate. Schools lost hours and trust trying to sort the mess. Principals had to act as complaint handlers. Mavis got blamed for a mess she didn’t cause. Hiki watched as good teachers left the profession; it was the last straw. And every one of those experiences eroded the sense that public systems can be relied on. That implementation is about fairness as much as function.

And when the crisis escalated, what came was not humility but more of the same: centralisation, ministerial oversight, technical reviews that refused to apportion blame. A Deloitte report noted that it was not authorised to investigate how these decisions were made in the first place. It simply assessed the present and suggested process improvements. Which, in policy terms, is equivalent to determining flood damage while being banned from inspecting the dam.

Eventually, Novopay was pulled back in-house. The Ministry took complete operational control. The system was stabilised. The error rate dropped below 0.2%. But not before $45 million more was spent rebuilding what had worked perfectly well in the old system: human discretion, relational knowledge, adaptive capacity. The very features that the theoretical state had dismissed.

This is the lesson: the practical state cannot be abstracted. It can’t be deduced from policy diagrams or constructed in vendor deliverables for procurement processes. It lives in the quiet expertise of people who carry too much and ask for too little. It is fragile in the face of managerialism. But it will resist being ignored.

Novopay could have been done differently. Iteratively. Locally. With humility. It could have respected the legacy systems that worked, and the people who made them work. Instead, it became another example of a state in love with its cleverness and deaf to the warning signs shouted from the ground up.

And that’s the final warning. Schools were expected to absorb the shock quietly, through burnout, workaround, and exit. They were expected to adapt. I would say trust is finite. And front-line goodwill, once humiliated and bruised, does not readily volunteer itself again.

Suppose we want a public service that delivers with integrity. In that case, it must also be built on more than abstract models and meaningless theories of change: I am arguing in this series that change must begin where delivery lives: in the hands of those we ask to make the impossible work, day after day, without thanks, without pause, and far too often, without power.

Next week, we return to those quieter cases: here, the practical state holds, not because of pressure, but because people choose to act with care, judgment, and relational depth. We’ll revisit how discretion is exercised when it matters most, and how effective public service emerges not just from design, but also from attention, memory, and trust. I’m going to tell you all about a wahine called Annie Aranui, and a few others like her. The practical state is made up of some extraordinary, practical, and practically wise people.