The risk assessment that fell off the agenda

On 22 January 2026, a landslide at the Mount Maunganui Beachside Holiday Park killed six people. We think of them first, and of their families.

In August, the external review by retired High Court judge Paul Davison KC concluded that the hazard was known, that an effective and inexpensive means of managing it had twice been recommended, and that the Council had every opportunity to put it in place. RNZ summarised his finding this way: information identifying a risk to life was received, filed, discussed, forwarded and then left.

This post follows one piece of that information down a path the review was not asked to follow: the public agenda of the board that governs Mount Maunganui. It is a short path. A geotechnical report that rated rockfall risk at the campground as High was printed in one agenda. By the next agenda that actually became a meeting, it had gone, and nothing in the papers said so.

We are not writing about who was at fault. The review does that, and so will other processes. We are writing about something narrower and more useful: what was on the page, what wasn’t, and which small, practical controls would have put the report back in front of elected members. Every one of them is something a council, a board or an individual councillor can adopt tomorrow.

Three agendas, one report

The body is Ngā Poutiriao ō Mauao, the joint board that administers Mount Maunganui alongside the Mauao Trust. Its agendas are public. We read three.

Tauranga City Council has said publicly that the 6 August meeting “did not proceed because of concerns about quorum requirements”, that a link to the agenda had been circulated before it was cancelled, and that none of the councillors on the board read the HD Geo report. It has also said it has not been able to establish why the report was not referred to on 28 November. A council spokesperson told RNZ they were not sure why the August meeting was never rescheduled.

Two details from the November pack matter for what follows. The board’s memorandum of understanding says a quorum is four members, including at least two appointed by the Trustees and at least two by Council, and that the board meets “no less than two (2) times a year” (Nov agenda, p.25). So the board met twice that year, and a single cancelled meeting left a gap of nearly six months with no obvious mechanism to catch what fell into it.

What was in the August pack

If you had opened the August agenda and gone looking, this is what the HD Geo report said:

• It rated rockfall risk to life at the campground High (likelihood Possible, consequence Major), alongside High ratings for sections of several tracks. It rated landslide risk to life at the campground Moderate (Table 4, p.48).
• It said the risk assessment found “sections of the campground at high risk of rockfall that could impact TCC campground assets and users such as cabins and campers” (p.46).
• It recommended an updated quantitative risk assessment, giving as reasons that there had been “no detailed landslide or rockfall risk assessment for the last 20 years”, that there had been at least 101 recorded landslide or rockfall events on Mount Maunganui since 1999, and that permanent cabins sit within the rockfall run-out zone (pp.51-52).
• It said a Trigger Action Response Plan “could be developed relating to heavy rainfall or seismic events”, listed under passive risk management, with the example of temporary track closures (p.50). Davison later described this as the second consultant report to recommend a trigger plan for the campground (review, para 351).

Two fair points about how this reads.

First, the report arrived as part of an Operational Report. The paragraph that introduced it describes HD Geo as the contractor “now completing the Mauao rock stability monitoring” and says the first report “included background research and a wide overview to ensure a quality transition” (p.23). That reads as a supplier handover. A reader skimming paragraph 13 would not learn that pages 46 to 52 of the attachment contained a High rating.

Second, the report was written mainly about tracks. On landslides it reduced the likelihood rating “in relation to track users/risk to life because it is less likely that there will be track users present in heavy rainfall” (p.46). That reasoning is sensible for people walking a track. It is less obviously sensible for people staying in a campground, who are present in heavy rain, and earlier reports quoted in the review had warned that sleeping occupants in fragile accommodation face a higher risk than walkers. Nobody could have been expected to spot that from a skim. A reader could have asked who the assessment was about.

The agenda item itself asked the board to receive the report. Nothing in the pack asked it to do anything about the recommendations.

What was missing from November

We ran the August and November agendas through Aide, one at a time, as any councillor’s briefing would be run. The August brief flagged the High ratings in “What’s at risk”, and flagged in “What’s unmeasured” that no timeline, budget or decision to commission the recommended assessment was before the board.

The November agenda said nothing about rockfall, HD Geo or the campground. The absence of what was on the previous agenda does not announce itself. The one adjacent line was the tracker entry “Complete”. A status like that closes a line. It says the monitoring programme exists, not what its first report found or whether anyone acted on it.

The Aide November brief — because it cross-checks previous briefs — noticed that something that was on the previous agenda had disappeared without being noted as actioned or resolved.

Governance is the job

The Local Government Act is clear: elected members carry democratic accountability for the decision-making of the council. They’re obliged to perform the powers, authorities, and duties of their office — in the best interests of the region or district — to the best of their skill and judgement.

Five controls that would have caught this

Agenda packs precede every meeting, so the logical place for controls is in the agenda itself. Each of these suggested controls is a standing line on an agenda or a column in a table.

1. A “promised and not delivered” line. In June, the agenda said the finalised report was “planned to be tabled”. Any item an agenda promises for a meeting and does not deliver should appear in the next one under a heading such as Promised, not yet received. Here: the report would have been listed at the next meeting, with a date.
2. A carry-over rule for cancelled meetings. The August agenda existed, was circulated and was never considered. Every item in a cancelled meeting’s agenda should be re-listed, in full or by reference, under Carried from [date] at the next meeting that sits. Here: November would have opened with “carried from 6 August: Operational Report, including Attachment 1”.
3. One standing item for risks to life. Any staff report that rates a risk to life High or above gets one line in a standing register on every agenda: what the risk is, who owns the next step, and by when. Here: the campground rockfall rating would have been on the register from August until someone closed it.
4. “Complete” has to say what happened to the findings. A status tracker should separate “work done” from “findings dealt with”. Here: F9 would have read “Monitoring set up. First report rated campground rockfall High; follow-up assessment not commissioned.”
5. “Receive” needs an owner. When a report recommends a next step and the agenda asks the board only to receive it, the paper should name who owns the recommendation after the meeting. The review makes a related point about staff: ownership of a risk should not be discharged by “forwarding a document or referring the matter to a meeting” without express written approval (para 441). A meeting is a place where a recommendation can be seen, not a place where it is automatically owned. Here: “Recommended: updated quantitative risk assessment. Owner: [role]. Report back: [date].”

None of these requires anyone to read more. They move the burden from memory to structure.

The review makes the same argument about trigger plans: a good one “does not depend on anyone forming, unaided and in the moment, a correct judgement”, and instead “fixes, in advance and by reference to defined criteria”, what must happen (para 395). An agenda can be built the same way.

What a councillor can do right now

If your council hasn’t adopted any of the above, you can still ask, in writing or at the table:

• Is there anything promised for the last meeting that isn’t in this pack?
• Did the last meeting happen? If not, what was on its agenda and where is it now?
• Which items in this pack are risks to life, to safety or to major spending, and who owns the next step on each?
• When a tracker says “Complete”, what did the work find and what was done about it?
• Who is the assessment actually about? Track users, residents, staff, campers? Is that the group most exposed?

Those questions are not accusations, and they are hard to answer with a shrug.

Aide cuts hours of agenda analysis down to minutes

The sheer volume of information that comes with bureaucracy means mistakes are virtually inevitable. Items slip off the agenda. Despite the best intentions of council staff, that’s reality.

Aide was built to catch these exact administrative gaps. When councillors are wading through thousands of agenda pages, human memory cannot reasonably be the only safeguard against dropped reports.

Aide analyses the deluge of council information — agendas, reports, financial statements, projections, outcomes — and distils it into what’s tracking, what’s at risk, what’s unmeasured, and what you are really being asked to endorse, with page and item references so you can check and dig deeper.

When we built the first version of Aide, it wouldn’t have caught a critical report falling off an agenda. So we changed it.

That’s just one of the design choices we’ve made to help councillors and council staff operate in immensely complex situations where decisions and omissions have real life consequences.

Aide isn’t a silver bullet — it only sees the public packs it is fed and cannot tell you why an item vanished — but it guarantees that an unexplained absence will not go unnoticed.

Aide brings the attention that systemic gaps obscure, giving councillors the visibility they need to ask the right questions before a meeting is closed.

For how this differs from asking a general chatbot, or from relying on a staff briefing alone, see Aide vs ChatGPT and Aide vs staff briefings. Briefly: a staff paper shows what staff chose to put in front of you, and a chatbot has never seen last August’s pack.

Sources

• Ngā Poutiriao ō Mauao and Mauao Trust joint meeting agendas: 12 June 2025; 6 August 2025; 28 November 2025 (headed “5 November”); 28 November minutes. Page numbers are the printed page numbers in each agenda.
• Tauranga City Council, Statement from Tauranga City Council, 21 August 2026.
• RNZ, Report on Mt Maunganui landslide risk unread: Tauranga City Council, 21 August 2026.
• Hon Paul Davison KSO KC, External Review into the Mount Maunganui Beachside Holiday Park Landslide on 22 January 2026 (August 2026), paras 351, 395, 441-449. Council’s FAQ on the review.
• Local Government Act 2002, as at 23 September 2026: ss 10, 39, 41, 42; Schedule 7 cl 14. Section 10 was replaced on 19 September 2026 by the Local Government (System Improvements) Amendment Act 2026.
• Aide briefs of the 6 August 2025 and 28 November 2025 agendas.