What breaks on day two of a "successful" merger?

When Auckland's eight councils became one in 2010, the merger was complete on paper in a single day. One council. One mayor. One chief executive. Clean.

Underneath it sat more than 3,500 separate computing systems.

The programme to consolidate them, NewCore, ran for years. It was, by Auckland Council's own description at the time, one of the most significant projects it had undertaken since amalgamation. Not one of the most significant technology projects. One of the most significant projects, full stop. Six years after the merger it was still in final testing, with further rollout scheduled the year after that.

That gap is the whole subject of this article. The structure merged on day one. The organisation took the better part of a decade, and arguably is still going.

The boundary is not the system

Head Start proposals are being written right now against a 9 August deadline. They are, necessarily, about boundaries. Which councils, which functions, one unitary authority or two, where the lines fall. The assessment criteria are practicality, simplicity, value for money, effective representation, timeliness and fit with the new resource management system. Reasonable criteria. But they describe a shape, and a shape is not a system.

Consider what a council actually is. It is a consenting process that depends on a property database, which depends on a GIS layer, which depends on a rating system that talks to a finance system, which feeds an annual plan, which sets the levels of service the consenting process is measured against. Change one and the others move. Now do that four times over, with four different vendors, four different data standards, four different definitions of what counts as a property, and four sets of people who each believe their version is the sensible one.

The Spinoff made an observation in June that deserves more attention than it got. More important than the maps is how the organisations are structured: what assets they control, what powers they have, how their balance sheets work. Those specifications make more difference to daily life than the boundary lines do. That is a systems argument, and it is correct. The map is the least consequential thing on the table.

Why the savings arrive late, or not at all

The 2022 Infrastructure Commission review of Auckland found little evidence of cost improvement in core services such as road maintenance and building consent processing. Total operating costs went up. Auckland Council maintains it has saved $2.4 billion since 2010, and both things can be true at once, which is exactly the point.

Ōpōtiki's mayor put the sceptical version plainly. One of the best performing councils in the country happens to be one of the smallest. If bigger were automatically cheaper, Auckland would have the cheapest rates in New Zealand.

The systems explanation for this is not mysterious. Savings live in the connections, not the boxes. You cannot decommission four consenting systems until you have one that works. You cannot have one that works until you have agreed a single definition of a consent, a single workflow, a single set of rules, and a single answer to the question of whose local variation was load-bearing and whose was just habit. Every one of those agreements is slow, contested and invisible from the outside.

Meanwhile the costs arrive immediately: integration work, dual running, transition teams, people leaving with knowledge in their heads that was never written down. Councils have been told there is no additional money for this and that developing proposals is business as usual. Ministers have discouraged hiring large numbers of consultants to do the work.

So the pattern is costs front-loaded and certain, benefits back-loaded and contingent on a set of technical and operational agreements that nobody is currently discussing, because everyone is discussing boundaries.

A question for anyone writing a proposal this month. What is the single number in your benefits case that depends on systems being joined up, and who has checked whether that is achievable in the timeframe you have assumed?

Four things interdependence does to a merger

It relocates the problem rather than removing it. The reform is aimed at duplication and inconsistency. Territorial and regional authorities share responsibility for spatial planning, district plans, regional plans, policy statements and environmental regulation: two organisations doing adjacent work, sometimes at odds.

Merging them removes the interface between the two organisations. It does not remove the work, the disagreement or the trade-off. It moves all of that inside one entity, where it becomes an internal argument between two directorates instead of an external argument between two councils. That can be an improvement; internal arguments are usually faster and cheaper than external ones. But it is not elimination, and a benefits case that treats it as elimination is overstating itself. The tension between environmental regulation and development consenting is structural, because the two objectives genuinely conflict. No organisation chart resolves that.

It creates coupling where there was tolerance. Separate councils fail separately. That is inefficient, and it is also a form of resilience. Four organisations with four systems means a failure in one is contained.

Join them, and you get shared efficiency and shared exposure. One consenting platform means one outage stops consenting everywhere. One payroll means one error affects everyone. One set of data standards means one bad decision about data structure propagates across a region.

This is a real trade, worth making deliberately rather than discovering by accident. The systems literature is unanimous that tight coupling plus complexity is where accidents come from, and Charles Perrow's work on normal accidents is the classic statement of it, holding up well outside its original nuclear and aviation setting. Efficiency and resilience are not the same objective, and optimising hard for one will cost you the other. Which raises the question nobody is asking in the Head Start conversations. What are you willing to give up in resilience to get the efficiency you are promising?

The interfaces you cannot merge become more important, not less. Even a large unitary authority sits inside a system it does not control. It has to work with the new resource management system and the digital planning backend that comes with it. It sits alongside the new water entities, which are already consolidating the biggest genuine savings opportunity, and doing it without requiring amalgamation at all. It works with iwi and hapū, with a mega-ministry that is itself mid-merger, and with a Commerce Commission that may end up assessing whether rates deliver value.

Every one of those relationships is an interface. None of them merges. All of them get harder to manage while your own organisation is internally unsettled, and several are being redesigned at the same time. Councils are being asked to reorganise while the planning system changes, water services move out, and their monitoring ministry is assembled from four other agencies: multiple interdependent reforms, on overlapping timelines, each assuming the others land as planned.

The knowledge that runs the place is not in the systems. Every council contains a small number of people who know why the rating anomaly on that street exists, which stream floods when it rains hard from the north, which consent conditions are actually enforced, and who to ring at the regional council to unstick something.

None of that is written down. It transfers in corridors, and it does not survive restructuring, because the people carrying it are frequently the ones who take the package or read the room and leave early.

The public service side of this is running on a similar logic, with roughly 8,700 fewer core roles targeted by 2029 and mergers expected to deliver a large share of it. Attrition is explicitly part of the plan, and attrition is not neutral. It selects for the people with options, which correlates with the people who are good, which correlates with the people holding the undocumented knowledge that makes the system work.

A question worth asking before the restructure, not after. Who are the five people whose departure would quietly break something, and does anyone senior actually know their names?

What a systems-aware proposal does differently

Map the dependencies before drawing the boundary. Not all of them. The ten that matter. Where does information have to flow for a consent to issue, for rates to be struck, for a flood response to work? Draw the boundary so it cuts the fewest of those, rather than drawing it on a map and discovering the cuts later.

Sequence around services, not around organisations. Consolidation tends to succeed when it starts with a small number of high-volume journeys and builds outward, rather than attempting everything at once. Pick the two or three things that touch the most people, join those properly, and let the rest wait. Big-bang integration is how you end up with a six-year NewCore.

Name the tolerances you are giving up. Write down, explicitly, what you are choosing to couple and what you are keeping separate on purpose. Redundancy that looks like waste on a spreadsheet is sometimes the reason the lights stay on.

Assume the interfaces are the work. if your transition plan has a workstream for structure and a workstream for systems, but no workstream for the relationships with the entities you cannot merge, the plan is incomplete.

Fund the boring middle. The costs land in years one to three. The benefits, if they come, land in years four to eight. Any plan without a credible answer for how the organisation functions during the trough is not a plan. It is an aspiration with a timeline attached.

Capture the undocumented knowledge while the people are still there. This is cheap, unglamorous, and almost never done, because it feels like admitting the restructure will cost something. It will.

The through-line

Part one argued that structure does not fix judgement. This part argues that structure does not fix connection either. A merger produces a single legal entity on a stated date. That is the easy part, and it is the part the deadline measures. What it does not produce is a single organisation. That has to be built, deliberately, through the connections between systems, processes, data and people, and it takes years rather than months.

The organisations that manage this well will be the ones that understood, early, that they were not simplifying anything. They were relocating complexity, and complexity that has been moved rather than removed has a way of surfacing somewhere less convenient. One question to take into the next transition meeting. If the merger completes exactly on time and nothing else changes, what will still be broken on day two?

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Axis of Action works with organisations at the point where strategy meets delivery. Transformation advisory is the core of the practice: shaping programmes so they hold together from mandate to delivery, testing whether the intended change is actually being built, and helping leaders make the calls that keep it on course. Alongside this sits programme assurance, independent review, strategic alignment, governance services, and executive coaching for individuals and leaders. See the full range of services