The Environment Plan Census 2026
Every Environment Plan decision on the NOPSEMA public register, 2012 to 2026, measured as a census. Not a survey, not a sample, not an estimate. Every statistic carries the population it was computed over and its n. Every rate names its denominator. Where the public record cannot answer a question, the report says so in place, in the body, rather than filling the space with something plausible.
Somebody in your organisation is building a schedule around an approval date right now.
A rig is booked, a vessel is chartered, a seasonal window is picked. The number underneath that schedule is usually borrowed: from the last plan the company filed, from a colleague's recollection, or from a regulator's published compliance figure.
The register can answer the question directly. This report asks it of the whole record at once. What a licensed report owes a buyer is not a number. It is a number that survives questioning.
Four measurements, each with its population and its n.
The assessment clock has re-based, and it has not come back down
Grouped by decision year, the median wall-clock assessment for accepted plans ran 111 to 126 days across 2017 to 2020, with annual n between 30 and 42. It reached 474 days in 2023 (n=24) and settled at 275 days in 2024 (n=37) and 292 in 2025 (n=29): roughly two and a half times the earlier baseline. The same re-basing appears independently in region, in project lifecycle stage, in activity type and in submission type.
The tail, not the median, is the operational exposure
The 90th percentile reached 743 days in 2023 and 844 days in 2025. In both of those years, one accepted plan in ten took longer than two years. A schedule built on the median accepts a one-in-ten chance of the assessment outrunning it by more than a year.
The exit risk is cancellation, not refusal
Refusal is rare: 13 plans of 713 on the register (1.8%), or 13 of the 618 that reached a decision (2.1%). Cancellation is not: 60 plans, 8.4% of the register, over four times the refusal rate. A risk register that models refusal but not withdrawal is modelling the smaller of the two exits.
Reported and measured are two different clocks
Across thirteen financial years from FY2012-13, NOPSEMA's published on-time figure never leaves a band of 95% to 100%. Over the same years the register's median wall-clock duration runs from 66 days to 474. Both series are public and neither is wrong. They measure different intervals, and only one of them is the interval a campaign window runs on. The report disputes no NOPSEMA figure and alleges no error by anybody.
The metrics, group by group.
Below is what the report measures and at what resolution. A complete, machine-enforced inventory of every measurement travels with the quote, so the person approving the spend can settle “am I buying one finding padded out” without reading the report. Ask for it and it comes.
| Measured | Detail |
|---|---|
| Environment Plans on the register | 713, after excluding 7 flagged phantom records from 720 raw |
| Date coverage | Submissions 2012-01-01 to 2026-07-17; decisions 2012-01-27 to 2026-06-24 |
| Field completeness | 710 carry a submission date, 625 carry a decision date |
| Submitting entities and operator groups | 109 entities resolving to 59 corporate groups, 18 with ten or more filings |
| Activity types | All 15 values of the register's own classification |
| Regulatory alerts collected | 217, spanning 2011-12-14 to 2026-05-18 |
| Measured | Detail |
|---|---|
| Filings by submission year | 15 years, each split into accepted, other decided, and still under assessment, every row summing to its filing count |
| Duration percentiles by decision year | p25, median, p75, p90 and maximum for each of 15 decision years, with n on every row |
| The same series at three resolutions | Annual, two-year (7 periods) and three-year (5 periods), each with 95% process intervals, published together rather than choosing the one that reads best |
| Duration by region and era | 8 regions, 2017-2020 against 2023-2025, with n on both sides and thin cells marked |
| Duration by activity type and era | 12 activity types on the same two-era basis |
| Duration by lifecycle stage and era | All 5 stages of NOPSEMA's own Lifecycle Classification field |
| Duration by submission type and era | New plans against revisions, 494 and 219 on the register |
| The three most recent complete years | 2023, 2024 and 2025 with their intervals, which is what rules out the recovery reading |
| Measured | Detail |
|---|---|
| Outcome distribution | All 8 register states with n and share, not collapsed to a binary |
| Refusal and cancellation | Both refusal denominators, the cancellation count and rate, and the concentration by activity |
| The two live assessment states | Plans with the titleholder against plans with NOPSEMA at the snapshot date |
| Outcome volumes by region | 9 region rows against 7 outcome columns, on the full register basis |
| Activity types in full | Filings, accepted population, all-time median duration and cancellations for each of 15 types |
| Activity crossed with region | Every cell reaching the reporting floor, with the share decided before 2021 printed beside it so the era confound is visible rather than hidden |
| Operators with ten or more filings | Filing and cancellation counts for all 18 groups, covering 598 of the 713 plans |
| Lifecycle stage | Volume, cancellation count and rate, and all-time median for each of 5 stages |
| The decommissioning cohort | All 111 plans: activity-type composition, full outcome distribution, region spread and filer concentration |
| Measured | Detail |
|---|---|
| Document anatomy | Median documents served and stored text per plan, on the 2017 onwards cohort, with the earlier summary-era cohort named and excluded |
| Acceptance criteria failed | 315 failed-criterion rows across 58 returned plans, by criterion, from NOPSEMA's published decision notices for 117 plans |
| What a return costs in time | Observed medians for the full return loop, on plans returned once and across all measurable returns, each stated as a floor |
| Measured | Detail |
|---|---|
| Decisions | This report's decided-plan count against NOPSEMA's published assessments completed, 2017 to 2023: an exact match in every year |
| Submissions | The same seven years on the submission basis, with the definitional gap shown and explained rather than smoothed over |
| Acceptances, independent source | NOPSEMA's own Environment Plan acceptance counts from the endorsed Program appendix of each annual report, 9 financial years, agreeing in 7; refusals agree in all 9 |
| Published timeliness | NOPSEMA's own assessment-timeliness figure for all 14 financial years, on both the Environment Plan basis and the wider all-key-documents basis, each year labelled with which it rests on |
| The register on the same axis | Duration by financial year of decision for 15 financial years, so the two series sit on one chart |
| Measured | Detail |
|---|---|
| The open book | All 30 plans under assessment at the snapshot date, with elapsed time already served and the definition of an open plan set out and defended |
| Where they stand | The 30 plans placed in 5 bands against the report's own published medians and p90 |
| The standing queue | The measured arrival rate and mean time in system, and what the queueing identity implies against the observed book |
| Three scenarios | The three duration regimes this register has actually recorded, each with its n, median, p90 and implied queue, and the decision dates each implies for a plan filed on the snapshot date |
| How much longer the open plans have to run | Conditional remaining time for 26 of the 30 open plans, by band, estimated by survival analysis over a 176-plan submission cohort with the open plans censored rather than dropped, and the four longest-running left unestimated because too few plans remain at risk |
No scenario is ranked and none carries a probability
The report contains no forecast, no market size and no growth rate. Every figure in the forward chapter is either a fact about a plan already on the register or arithmetic on a measured quantity, carried forward under a condition stated in the same sentence.
Sixteen chapters.
- Executive summary — the register at a glance
- Scope, definitions, and how to read this report
- The register 2012 to 2026 — volume, operators, activity mix
- The assessment clock
- Outcomes
- Activity types
- Operator activity
- Document anatomy
- Lifecycle stage
- Resubmission and the acceptance criteria
- Reconciliation against the regulator
- Reported and measured
- The forward view
- What this data cannot tell you
- Methodology
- Data appendix — every table behind every figure
By the person who has to use it.
The percentile table is the deliverable, and one finding in it is counter-intuitive enough to be worth the licence on its own: for a plan already under assessment, the time still to run does not shrink as the wait lengthens. Across the open plans the report can estimate, the median remaining time sits in a narrow band whatever the plan has already served, so time already spent is not evidence that a decision is near. The report prints p25, median, p75, p90 and the maximum for every decision year, so a planner can pick the percentile that matches their own risk tolerance rather than being handed one.
Two numbers change a standing risk register. Cancellation runs at over four times the refusal rate, and it concentrates by activity rather than spreading evenly. The report also names, in a claims register of its own, every statement it does not support. That is the section an assurance function reads first.
Decommissioning is the lifecycle stage that re-based hardest of any on the register, at 4.9 times its earlier median. NOPSEMA classifies 111 plans as decommissioning and they file under seven different activity types, so the stage and the activity label answer different questions. The report prints the full composition, the cohort's outcomes, its region spread and its filers.
The acceptance-criteria analysis shows that returned plans do not come back with one targeted problem: the same short list of criteria recurs across the cohort. The practical consequence is stated plainly, and it is not the one most readers expect.
Every figure can be carried into a board paper with its population printed beside it, its denominator named, its method stated and its limits declared before anyone has to discover them. Licence holders may ask for identified figures to be re-derived against the frozen snapshot they were computed from, while the edition is current.
Questions about any figure, its population or its derivation are answered directly. That is the point of publishing the populations in the first place.
Four absences, and they are the reason the rest can be relied on.
No market size, no forecast, no growth rate
The register records filings and decisions, not spend. Any market figure would have to be imported from undisclosed estimates or invented.
No cross-operator duration league table
That comparison is confounded by filing era: an operator's all-time median is substantially a measurement of when it filed. One internally drafted headline of exactly that shape failed an era-controlled test and was withdrawn before publication. The report explains the failure mode rather than printing the table.
No causal claim about the regulator, in either direction
The 2023 movement in the clock coincides with regulatory change, resourcing changes, litigation and shifts in activity mix. The register cannot separate them, so the report measures the effect and does not assert a cause. The symmetric claim, that submissions got worse, is equally absent.
No quoted Environment Plan text
Plan documents remain their titleholders' copyright and none is reproduced anywhere in the report.
How the numbers are held to account.
- One source, named so any reader can go to it. Every figure is computed by ProveINTEL from the NOPSEMA public register of Environment Plans at info.nopsema.gov.au. No third-party dataset, no survey, and no privately supplied operator data enters any figure. ProveINTEL is independent of, and not endorsed by, NOPSEMA, and nothing in the report is a statement by, or attributable to, the regulator.
- Every number has a population and an n. Where two denominators are defensible, both are printed with a note on which question each answers.
- A stated reporting floor, applied visibly. No median or rate computed on fewer than 20 records is presented as a finding. Cells below the floor are printed with their n and marked in place, rather than dropped or quietly included.
- Limits collected in a chapter of their own. Fifteen named limits, each stating what is missing, why it matters, and what data would lift it. Forty-six in-place markers record the specific questions the public record cannot answer, at the point in the text where a reader would otherwise expect an answer.
- Two independent reconciliations against the regulator's own published counts, on two different publications and two different time axes.
- A published quality statement against two external frameworks, assessing this report on seven fixed quality dimensions and assessing the register as an administrative data source on four practice areas. Where the honest grade is no assurance, no assurance is printed. Neither standards body has reviewed this report and no endorsement is claimed or implied.
- A frozen, identifier-pinned snapshot. Every figure is re-derivable against it, and a future edition's snapshot can be diffed against this one record by record, so a figure that moves between editions is attributable rather than mysterious. A published corrections policy states in advance how an error will be handled: labelled as a correction, dated, quantified, with the superseded version retained.
- It pre-registers its own scoring. The report names three of its own measurements, pins exactly what will settle each, and commits to publishing the result in a later edition whether or not the answer flatters it. That is not a forecast; it is a public undertaking to come back and mark our own work.
A$10,000, single-organisation licence. Published August 2026.
The unit is a licence, not a seat and not a per-report fee. Invoiced on release; no subscription is required. Quotation is permitted with attribution and with populations named as printed. Order a copy and we will send a quote, the full contents list and the order form, which carries the licence terms and is the instrument that binds them. Nothing is charged before you have all three in front of you, and there is no card field anywhere on this site.
A free edition is available now for an email address, and it is the register-level summary carrying the headline series. This licensed edition is the depth it does not go to.
ProveINTEL, The Environment Plan Census 2026. Every figure traceable to the NOPSEMA public register (info.nopsema.gov.au). ProveINTEL is not affiliated with, or endorsed by, NOPSEMA.
Buy a copy, or read the free edition first.
Ordering sends you a quote, the full contents list and the order form that carries the licence terms. It does not order anything and nothing is charged. If you would rather see the method before the price, the free edition proves it by doing it.
Or write to services@klarite.com.au.