How many formal “Referral Decisions” will the Australian Government publish on the EPBC Act Public Portal between June 2, 2026, and August 10, 2026?

• My most likely prediction for formal Referral Decisions is 13.50, with a 50% chance of falling between 11.50 and 33.50.
• The forecast window covers a 72-day period characterized by the most significant administrative transition in the history of the EPBC Act: the handover from DCCEEW to the new National Environmental Protection Agency (NEPA) on July 1, 2026.
• Historical data from 2022-23 suggests a baseline of approximately 42 decisions for a window of this length, but recent portal activity in early 2026 shows a sharp decline to approximately 5 decisions over a similar timeframe.
• The $500 million in funding for environmental law implementation and the goal of streamlining approvals provide potential for a surge in output, but this is likely to be delayed by the practical challenges of the agency transition.
• The 50% highest density interval reflects the high uncertainty of whether the new agency will immediately clear backlogs or suffer from initial ‘teething’ delays during the first six weeks of operation.

Key figures

Historical context

• In the 2022-23 fiscal year, the DCCEEW reported 214 referral decisions, which included 136 ‘Controlled Actions’ and 63 ‘Not a Controlled Action’.
• A 2020 audit by the Australian National Audit Office (ANAO) criticized the previous administration of these assessments as ‘ineffective’ and noted frequent procedural errors and delays.
• Historically, the median approval time for projects under the EPBC Act increased from 48 weeks (20 years ago) to 118 weeks by late 2025, prompting the current legislative reforms.
• On November 27, 2025, the Australian government reached a legislative agreement with the Greens to establish the NEPA and move toward a ‘one-stop shop’ for environmental approvals.
• Recent portal data from early 2026 (March to May) showed five specific decisions, such as the Snowy 2.0 Main Works Modification (May 19) and the Murchs Corner Battery Storage (May 6).

Tailwinds

• Reform Mandate: The 2025 reforms specifically aim to ‘fast-track’ approvals for energy, housing, and critical minerals to boost economic productivity.
• New Funding: Over $500 million has been allocated to implement these laws, including nearly $106 million for AI tools to assist in assessment navigation.
• Backlog Clearing: There is a possibility of a surge in publications as the outgoing DCCEEW tries to finalize pending referrals before the July 1 transition.
• Bilateral Agreements: New ‘one-stop shop’ agreements, such as the MOU signed with Western Australia, may reduce the workload on federal staff, allowing for faster processing of remaining referrals.

Headwinds

• Administrative Friction: The transition from DCCEEW to NEPA on July 1, 2026, is likely to create temporary bureaucratic delays and ‘teething issues’ with new systems.
• Recent Portal Slump: Only 5 decisions were recorded between March 4 and May 19, 2026, suggesting a significant downward trend in publication frequency compared to 2023.
• Staffing Uncertainty: As of February 2026, the CEO of the new NEPA had not yet been appointed, which could lead to a lack of decisive direction during the first weeks of the agency’s operation.
• Regulatory Complexity: The introduction of new National Environmental Standards may require extra time for officials to ensure initial decisions under the new regime are legally robust.

Detailed reasoning

My analysis of the projected number of referral decisions is based on a synthesis of historical throughput, recent observed trends on the public portal, and the major administrative overhaul occurring mid-window.

Historically, the Department of Climate Change, Energy, the Environment and Water (DCCEEW) has maintained a steady output. In the 2022-23 fiscal year, the department published 214 referral decisions. Pro-rating this across the 72-day window (June 2 to August 12) yields a baseline expectation of approximately 42 decisions. However, more recent data from early 2026 suggests a significant downturn in activity. From March 4 to May 19, 2026 (a 76-day period), only five referral decisions were documented on the public portal. This recent rate of approximately 2 decisions per month is drastically lower than the historical average and serves as a strong signal of either administrative stagnation or a shift in how decisions are being processed and published during the transition period.

The forecast is heavily influenced by the July 1, 2026, commencement of the National Environmental Protection Agency (NEPA). This transition is the result of the National Environmental Protection Agency Act 2025 and is intended to streamline approvals for critical sectors like housing and energy. I have considered two primary scenarios for this transition:

1. The ‘Transition Friction’ Scenario: Administrative handovers of this magnitude often result in short-term delays as personnel are reassigned, new systems (including a $105.9 million AI tool) are implemented, and new leadership (a CEO yet to be named as of early 2026) takes office. This scenario favors the lower end of the forecast.
2. The ‘Backlog Clearing’ Scenario: Conversely, there is often a push to finalize ‘legacy’ decisions before a new agency takes over, or the new agency may prioritize a rapid burst of decisions to demonstrate its efficiency and justify its $500 million budget allocation.

I weighted the ‘Transition Friction’ scenario more heavily because the recent portal data (only 5 decisions in 76 days) suggests that a slowdown is already underway. While the legislative reforms aim to accelerate project approvals, the immediate impact during the first 43 days of the new agency’s existence (July 1 to August 12) is likely to be tempered by the practicalities of institutional change.

The predicted most likely value of 13.5 reflects a compromise between the extremely low recent run rate (which would suggest ~5 decisions) and the historical baseline (~42 decisions). The 50% HDI of 11.5 to 33.5 covers the range from a continued slump to a moderate return toward historical averages as the new agency begins to find its footing.

Key uncertainties

• Technical Portal Stability: It is unclear if the migration of data from DCCEEW to the new NEPA portal will cause temporary publication blackouts or delays in documents appearing online.
• Definition of ‘Decision’: The forecast depends strictly on the publication of formal documents; any delay between the actual decision and the portal upload will affect the final count.
• Leadership Vacancy: If the CEO recruitment for NEPA is further delayed past July 1, the agency may operate in a ‘caretaker’ mode with lower decision output.
• State Cooperation: Stalled negotiations with key states like Queensland regarding bilateral agreements could keep the federal workload higher than anticipated.

Conclusion

• The transition to NEPA on July 1, 2026, is the most critical variable. Decision-makers should watch for the announcement of a new CEO and any ‘transition reports’ issued in the first weeks of July to gauge if administrative friction is higher than expected.
• If the current slump in portal activity (only 5 decisions in ~76 days) is due to technical lag rather than a true drop in activity, the final count will likely skew toward the upper end of the 25–45 range.
• The 20-business-day statutory timeframe means that the volume of decisions in June is already largely determined by referrals submitted in May. Any sudden spike in late-July or August would indicate a successful ‘clearing of the decks’ by the new agency.

Question: how many formal EPBC Act referral-stage decisions will be published on the EPBC Public Portal between 12:00 AM AEST 2 June 2026 and 11:59 PM AEST 12 August 2026? All forecasters agree the portal screenshot showing a running count of 0 is just a starting condition and not the final count, and they use official annual throughput as the best quantitative anchor (214 decisions in 2022–23 and 237 in 2020–21 → ~18–20 decisions/month). Converting those annual totals into a 72-day window yields a central estimate around 43–45 referral decisions (a median near 44), with overdispersed uncertainty (typical team quantiles: p10 ~36, p25 ~40, p75 ~49, p90 ~53 and ranges cited roughly 30–57). All teams highlight qualitative modifiers that could push counts down or make publication lumpy: visible upstream queues and statutory 20-business-day decision clocks support continued flow, while validation delays, suspension/clock-stops on specific projects, and the July 1, 2026 institutional transition introduce downside risk and timing noise. Practical note: resolve and monitor by directly filtering the EPBC Public Portal by Publish Date (2/06/2026–12/08/2026 unless platform clarifies the August 10 vs 12 ambiguity) for notices named “Decision whether action needs approval” and related decision types, since indexed project pages can be stale. Overall, treat ~44 decisions as a base-rate anchor with substantial uncertainty from operational timing rather than a precise point prediction.

Forecast and Reasoning

Forecast: Probability distribution:

• 1.00% chance of value below 30.0
• 5.00% chance of value below 36.0
• 10.00% chance of value below 40.0
• 20.00% chance of value below 45.0
• 30.00% chance of value below 49.0
• 40.00% chance of value below 52.0
• 50.00% chance of value below 56.0
• 60.00% chance of value below 59.0
• 70.00% chance of value below 63.0
• 80.00% chance of value below 68.0
• 90.00% chance of value below 76.0
• 95.00% chance of value below 83.0
• 99.00% chance of value below 95.0

Forecast Reasoning

We are forecasting the integer count of distinct EPBC Act Public Portal publications that are specifically at the “Referral Decision” stage (controlled action / not controlled action / particular manner / clearly unacceptable / s74A refusal), with publication dates between 2 Jun 2026 and 12 Aug 2026 AEST (72 days per the resolution criteria).

Across analysts, the most evidence-grounded anchor is the recent annual throughput: multiple forecasters independently cite recent DCCEEW annual figures on the order of ~250–280 referral decisions/year (e.g., ~276 decisions/year). Scaling that baseline to a 72‑day window (about 19.7% of a year) implies a central expectation around 50–56 decisions. Forecaster 3’s independent cross-check using recent 2026 weekly “EPBC Act Part 7–9 decisions published” gazette counts (typically 3–7 per week, ~5/week average) maps to ~51–52 over ~10.3 weeks, consistent with the annual-rate approach.

Main uncertainty drivers are (a) short-window operational volatility/batching, (b) any seasonal/end-of-financial-year effects around late June, and (c) possible administrative disruption or acceleration from policy/administrative changes (Forecaster 4’s NEPA-transition spike argument is plausible but insufficiently evidenced here, so I treat it as an upside risk rather than the base case). Net: I keep the median close to the annual-rate baseline but allow a moderately heavy right tail into the 80s/low-90s, while keeping the left tail tighter than the most pessimistic forecasts because getting far below ~40 over 72 days would require an unusually large and sustained slowdown.

This yields a median of 56 decisions, with the bulk of probability between the mid‑40s and high‑60s and a small (but non-zero) chance of exceeding the 90.5 upper bound.

An amateur forecast is likely to (i) confuse “referral decision” with other portal milestones (public comment/assessment/approval), and/or (ii) pick a midpoint of the provided bounds without anchoring to observed throughput. This synthesis instead triangulates on two semi-independent rate signals the analysts used (recent annual decision totals and recent weekly gazette volumes), then adjusts uncertainty for short-window operational variance and plausible but not well-validated disruption/spike scenarios. My forecast is less tail-heavy on the low side than some individual forecasts (reflecting aggregation over 72 days) while preserving a realistic upside tail (including a small open-above chance).

Forecast rationale (numeric):

— Iteration 1 — Across the forecasts, the main approach is to anchor on historical EPBC referral-decision throughput and then adjust for a short, somewhat uncertain publication window from June 2 to August 10, 2026.

Core reasoning pattern

• Base-rate extrapolation: Each forecast starts from an estimate of annual referral decisions and scales it to the ~72-day period.
• Adjustment for current administrative conditions: All models factor in some combination of:
  • EPBC portal disruption / read-only status / update notices
  • Potential backlog or batch publication effects
  • Uncertainty from Tranche 1 reforms that may slow processing or alter how decisions are labeled/published
• Wide uncertainty bands: Because of possible delays, batching, or publication changes, the forecasts keep a broad distribution rather than a tight point estimate.

Areas of consensus

• The expected number of published formal “Referral Decisions” is moderate rather than extreme.
• Most of the probability mass is centered in the low-to-mid dozens, with medians in the low-to-mid 40s for two forecasts and around the high 50s in the more throughput-heavy estimate.
• All models agree that portal/reform uncertainty is the main source of forecast error, not the underlying referral process alone.

Main differences

• Underlying annual rate assumption: One forecast uses a higher historical baseline (~295/year), producing a higher central estimate (~58), while the others use lower baselines (~220/year or 5–8/week), yielding medians around 43–45.
• Weight given to backlog vs. disruption: Some models lean toward backlog catch-up increasing published counts, while others emphasize that portal issues and reform-related friction could reduce visible decisions in the period.
• Tail risk treatment: All allow for both low-output and high-output scenarios, but they differ in how much probability they assign to each extreme.

Overall synthesis

The collective view is that the count will likely be in the several-dozen range, with a central expectation around the mid-40s to high-50s, and substantial uncertainty driven by portal functionality, publication timing, and reform-related administrative changes.

— Iteration 2 — Across the forecasts, the main reasoning is that the June 2–August 10 window is short enough to estimate from recent throughput rather than long-term trends, so the expected number of published EPBC “Referral Decisions” is scaled from an implied annual or monthly cadence. Most models arrive at a central expectation in the low-to-mid 40s up to low 50s, with the range largely driven by how often the portal typically posts decisions and whether the period is treated as closer to 70 or 72 days.

Key factors used across the rationales:

• Historical publication rate / baseline throughput:
  The forecasts anchor on an implied annual flow of roughly a few hundred decisions, which converts to about 40–60 decisions over the target window.

• Business-day and statutory-process logic:
  One rationale explicitly uses the idea of about one formal decision per business day, while another points to the 20-business-day statutory deadline as a reason decisions should continue at a steady pace.

• Calendar effects and seasonality:
  The June end-of-financial-year period is treated as a possible source of a modest increase in decision volume.

• Portal and administrative uncertainty:
  All forecasts allow for the possibility of batch uploads, backlog clearing, portal restructuring, staffing changes, or definitional ambiguity, which could push the count materially above or below the central estimate.

• Right-skewed uncertainty:
  The distributions are generally asymmetric, with heavier upper tails to reflect surge/batch scenarios, while still leaving room for lower-than-expected output if publication slows or stalls.

Areas of consensus:

• The period should likely produce dozens, not single digits or hundreds, of decisions.
• The most plausible center is around 40–55.
• There is meaningful uncertainty, so the forecast should be broadly distributed, not tightly concentrated.

Main disagreement:

• The main difference is how high the central rate should be: one view assumes a lower baseline and lands in the low 40s, while others infer a somewhat stronger cadence and land around 52–53.

— Iteration 3 — Overall, the forecasts converge on a count in the dozens, not single digits or hundreds, with a central expectation in the mid-40s to mid-50s over the roughly 70–72 day window.

Main reasoning patterns

• Historical throughput is the anchor. Each estimate is built from past EPBC referral decision volumes, translated into a daily or weekly rate and then scaled to the date window.
• Publication timing matters. The forecasts treat the portal as a publication stream rather than a pure decision-maker, so they allow for delays, batching, or backlog clearance to shift the observed count.
• Policy and operational uncertainty widen the range. Ongoing environmental-law reform, possible processing slowdowns, or administrative disruption could reduce output; conversely, a backlog-clearing push could raise it.

Areas of consensus

• The likely outcome is around 40–60 formal referral decisions.
• All forecasters allow for substantial uncertainty, with wide intervals reflecting volatile publication timing and possible changes in processing behavior.
• All see the risk of both lower-than-normal counts from disruption and higher-than-normal counts from accelerated publication.

Main differences

• One forecast is somewhat higher, centering around the mid-50s, based on a stronger daily throughput assumption.
• The others are slightly lower, centering in the high 40s or low 40s, using annualized or weekly historical baselines and more caution around reform-related slowdowns.
• Despite this spread, the estimates remain broadly aligned: a moderate volume of published referral decisions is expected, with a plausible range spanning the 20s to 60s, and some tail risk beyond that.

Summary of Rationales

Historical Base Rates and Timeline The rationales uniformly base their initial analysis on the approximately 72-day resolution window. Looking at historical data from the Department of Climate Change, Energy, the Environment and Water (DCCEEW), the baseline volume of referral decisions has steadily increased year-over-year. While historical averages provide a standard baseline for a period of this length, all rationales agree that mid-2026 is a highly anomalous regulatory period that will deviate from standard trends.

Upward Pressures: Grandfathering and Backlog Clearing Forecasters identify several strong drivers expected to push the number of decisions above the historical baseline:

• The “Rush to Refer”: Because stricter “Nature Positive” environmental standards and a new regulatory agency take effect on July 1, 2026, developers heavily front-loaded project submissions in early 2026 to “grandfather” them under older rules. Given standard statutory processing lags, this massive wave of applications will mature into formal decisions exactly during the June–August window.
• Pre-Transition “Desk-Clearing”: Aligning with the end of the financial year (June 30), the DCCEEW is highly incentivized to clear existing administrative backlogs before handing authority over to the new regulatory body.
• Government Interventions: Recent budget allocations, digital/AI triage upgrades, and dedicated “strike teams” are actively boosting the department’s capacity to process pending housing and renewable energy projects.

Downward Pressures: Transition Friction and Bottlenecks Offsetting the anticipated surge are severe administrative headwinds, primarily tied to the July 1 structural transition:

• Administrative Disruption: The mid-window launch of the independent National Environmental Protection Agency (EPA/NEPA) requires handing over regulatory powers, migrating personnel, and updating IT portals. Forecasters widely expect this massive machinery-of-government change to cause temporary bureaucratic friction and workflow slowdowns throughout July and early August.
• “Stop the Clock” Provisions: Because proponents rushed their applications to beat the July 1 deadline, many submissions are likely of lower quality. Forecasters anticipate the department will heavily utilize legal extensions and “stop the clock” provisions to request more information, thereby pausing statutory deadlines and pushing many final decisions beyond the forecast window.

Overall Outlook The rationales synthesize these opposing forces into a front-loaded dynamic: an elevated spike in decisions throughout June as the department clears the grandfathered backlog, followed by a noticeable slump in July and August due to transition-related administrative friction. Ultimately, the shared expectation is a total decision volume that exceeds the historical baseline, though extreme spikes will be constrained by the department’s processing limits and transitional bottlenecks.

Forecast — Post 401

Modal: 30-40 (30.0%) • partial_shift • 6m1s

Interpretation

Reading: strict

The resolution criteria are highly specific, enumerating exact qualifying document types, an explicit exclusion list, a named official data source, and precise AEST timestamp boundaries. The strict reading is further reinforced by the instruction that it counts only the ‘Referral Decision’ stage and explicitly excludes other portal document categories, leaving little room for interpretive latitude.

Time bounds

• observation_window: Referral Decision documents with publication date strictly between 12:00 AM AEST 2026-06-02 and 11:59 PM AEST 2026-08-10 (note: resolution criteria text says August 12 in the time constraint section, which differs from the question title’s August 10 — see edge cases)
• resolution_date: 2026-07-12

Edge cases identified

1. Date discrepancy: The question title and answer range header state ‘August 10, 2026’ as the end date, but the Time Constraint section of the resolution criteria states ‘11:59 PM AEST on August 12, 2026’ — a 2-day difference that could affect the count by a few decisions.
2. Resolution date (July 12) is well before the stated observation window end (August 10/12), meaning the question may be evaluated on partial data — only decisions published up to the resolution date would be observable at resolution time.
3. What constitutes a ‘distinct project’ if a project receives multiple Referral Decisions (e.g., a revised or reconsidered decision): counted once or multiple times?
4. The portal may not provide a directly filterable date range; the resolver may need to manually scan or use portal search filters, introducing potential for human counting error.
5. Whether projects that received a Section 74A refusal (a procedural refusal before formal referral acceptance) qualify as ‘Referral Decisions’ under the criteria — the resolution text explicitly includes them, but edge cases around withdrawn or lapsed referrals may arise.
6. Time zone boundary: decisions published close to midnight AEST on June 2 or August 10/12 may fall inside or outside the window depending on exact timestamp interpretation.

Research (3/3 variants, shared evidence pool)

Total evidence registered (shared pool): ?

Research Brief

Evidence confidence: medium

Scenario 1: Moderate rate — 30 to 50 referral decisions in window [high evidence]

Conditions favoring

Historical run-rate of ~346 combined referral+approval decisions per FY 2024-25 (per source 11) implies roughly 17–20 referral decisions per month if referral decisions constitute ~60–70% of that total; over ~2.3 months, this yields 39–46. EPBC numbering showing hundreds of referrals per year (per sources 2, 3) and the renewable energy sector alone producing ~40 referrals/year in 3 states (per source 12) supports a nationally robust referral rate. Standard operating conditions with no major administrative disruption.

Conditions against

The 346 figure includes approval decisions, inflating the estimate if referral decisions are a smaller share. Annual report data (per sources 8, 9) shows only 17–20 referral decisions required formal statements of reasons per year, suggesting possible lower absolute count if most decisions are routine NCA decisions not separately tallied. Transition to new NEPA from 1 July 2026 could slow throughput.

Scenario 2: Lower rate — fewer than 30 referral decisions in window [low evidence]

Conditions favoring

The annual report statements-of-reasons counts (17–20 per year under sources 8, 9) would imply only ~3–4 per month or ~7–9 per 2.3-month window if taken as a proxy for all referral decisions (though this interpretation is likely too narrow). Regulatory transition to NEPA from 1 July 2026 (per sources 6, 7) could slow decision throughput. The resolution date of July 12 means the question may be assessed on only ~6 weeks of actual data rather than the full window.

Conditions against

The statements-of-reasons data almost certainly undercounts total referral decisions, as routine NCA decisions may not require a formal statement of reasons. Portal data (sources 2, 3) shows active and ongoing decision-making in 2026. Long-run average of 285 referrals received per year (source 17) plus the typically short decision timeframe makes a count below 20 over 10 weeks implausible absent major disruption.

Scenario 3: Elevated rate — more than 50 referral decisions in window [medium evidence]

Conditions favoring

If proponents surge referral submissions ahead of the 1 July 2026 NEPA transition (per sources 6, 7) to secure decisions under the familiar EPBC Act framework, the rate of referral decisions in June–August 2026 could temporarily spike above the historical average. The renewable energy sector alone produced ~40 referrals/year in 3 states (source 12); nationally across all sectors, the rate could be considerably higher.

Conditions against

There is no direct evidence of a confirmed pre-transition lodgment surge. Administrative capacity is finite and decision timeframes are statutory (typically 20 business days after referral), meaning even a surge in lodgments would not immediately translate into a surge in decisions within the same window. The 2023-24 and 2024-25 annual figures do not indicate an accelerating trend beyond the established range.

Key uncertainties

• The exact share of total ‘referral and approval decisions’ (346 in FY 2024-25) that are specifically referral decisions versus approval decisions is not directly known from the evidence.
• Whether the regulatory transition to a National Environmental Protection Agency from 1 July 2026 will cause a surge in pre-transition lodgments, administrative slowdown, or neutral effect on decision throughput during the window.
• The resolution date of 12 July 2026 is before the observation window end (10/12 August 2026), meaning the question will be assessed on approximately 6 weeks of data rather than the full 10-week window — the implied count at resolution will be lower than the final window count.
• Whether the DCCEEW Quarterly Assessment Performance Report or annual report will have published updated data by the resolution date that could directly confirm per-period referral decision counts.
• Whether routine NCA decisions (which may constitute the majority of referral decisions) are systematically published on the portal on the same schedule as controlled action decisions, or whether there are lags.

Forecast variants (3/3)

Variant 0: inside_view (inside_view_v1)

Inside View Causal Analysis

Causal chain: Current state: EPBC Act referral decisions are being issued actively in 2026 (EPBC 2026/10431 decided 17 March 2026, EPBC 2026/10437 decided 16 April 2026 — source 2). The portal numbering in the 10400-10470+ range for early 2026 referrals confirms dozens of referrals filed monthly.

Base rate anchor:

• FY 2024-25: 346 combined “referral and approval decisions” (source 11) = ~29/month combined
• Long-run average: ~240-285 referrals received/year (sources 1, 17) = ~20-24/month received
• Renewable energy alone in 3 states: ~40/year (source 12), suggesting nationally across all sectors the referral decision rate is materially higher

Key rate estimation: The 346 combined referral+approval decisions (source 11) is the strongest quantitative anchor. The brief’s Poisson projection (high evidence strength) assumes referral decisions are ~60-70% of that combined total, yielding ~17-20 referral decisions/month, and ~39-46 decisions over the ~2.3-month window.

Scenario weighting by evidence strength:

1. Moderate rate (30-50 decisions): high evidence strength — driven by the Poisson projection centered at 39-46 decisions, supported by source 11’s 346 combined decisions/year, source 12’s renewable energy sector data, and active 2026 portal activity.
2. Elevated rate (50+ decisions): medium evidence strength — NEPA transition from 1 July 2026 (sources 6, 7) could cause pre-transition surge, but no confirmed evidence of surge lodgments and statutory decision timelines (~20 business days) limit near-term translation.
3. Lower rate (<30 decisions): low evidence strength — annual report statements-of-reasons data (17-20/year in sources 8, 9) is explicitly noted as a subset, not total count; routine NCA decisions are not in this figure.

Distribution construction:

• The high-evidence moderate scenario (30-50) should dominate, particularly bins 3 (30-40) and 4 (40-50)
• The medium-evidence elevated scenario adds mass to bins 5-6 (50-70)
• The low-evidence lower scenario adds modest mass to bins 1-2 (10-30)
• Tail bins (0-10 and 70+) get minimal mass

Note: The resolution date is July 12 — before the August 10 window end — but the question asks for the full June 2 to August 10 window count, so I’m forecasting for the complete window.

Approximate distribution:

• <10 (bin_0): 1% — nearly impossible given established run-rate
• 10-20 (bin_1): 4% — only if severe NEPA transition disruption
• 20-30 (bin_2): 12% — low-evidence scenario’s upper range
• 30-40 (bin_3): 28% — center of high-evidence moderate scenario
• 40-50 (bin_4): 28% — upper range of high-evidence moderate scenario
• 50-60 (bin_5): 15% — lower range of medium-evidence elevated scenario
• 60-70 (bin_6): 7% — upper range of elevated scenario
• 70-80 (bin_7): 3% — tail of elevated scenario
• 80-90 (bin_8): 1% — very unlikely

• 90 (bin_9): 1% — very unlikely

Variant 1: outside_view (outside_view_v1)

Reference class: monthly EPBC referral-decision issuance over recent years. Base-rate derivation: FY 2024–25 shows 346 combined “referral + approval” decisions (≈29/month) (source 11); assuming referral decisions are ~60–70% of that combined flow (brief’s arrival-model assumption, evidence summary), implies ~17–20 referral decisions/month. Adjustment: the window June 2–Aug 10, 2026 spans ≈2.3 months, so the simple outside-view base-rate projects ~39–46 referral decisions in the window (brief evidence summary and Poisson arrival-model projection). I make only modest adjustments around that base rate to reflect uncertainty from the 1 July 2026 NEPA transition (sources 6,7) and partial-observation/resolution-date timing noted in the brief (which increase tail risk), and to respect the brief’s scenario weighting: the “Moderate rate — 30 to 50” scenario is explicitly tagged high evidence, the “Elevated >50” scenario medium evidence, and the “Lower <30” scenario low evidence. Weighting rule applied: give materially more mass to the high-evidence scenario (bins covering 30–50) and smaller tails to medium/low scenarios. Distribution explanation: most mass is placed on bins 30–40 (bin_3) and 40–50 (bin_4) because the outside-view base-rate (≈39–46) centers there (bin_3 0.46, bin_4 0.35 = 81% combined). A modest 10% mass is placed in 20–30 (bin_2) to reflect the low-evidence possibility noted in the brief (sources 8,9) that formal-statement counts undercount total referrals and that administrative disruption could reduce throughput. A smaller 6% tail is placed in 50–60 (bin_5) for the medium-evidence elevated-surge possibility tied to pre-transition lodgment surges (sources 6,7, 12). A 3% small probability is assigned to 10–20 (bin_1) to reflect remote but possible severe slowdown or interpretation/measurement edge-cases. Remaining bins (>=60) receive negligible probability given historical annual rates and the briefing evidence (sources 1,3,11,12). Confidence in this distribution: medium — it follows the brief’s outside-view base-rate (sources 11,17) and respects the brief’s scenario evidence-strength tags (high for the 30–50 central scenario, medium for >50, low for <30).

Variant 2: contrarian_v2 (contrarian_v2)

Adversarial reading

The implicit consensus in the brief is a moderate-rate arrival process (~17–20 referral decisions/month) extrapolated over the nominal 10-week window, landing in the 30–50 bin. The bulk of the quantitative evidence (346 combined decisions in FY24-25, portal activity in 2026, renewable-energy subset rates) supports that underlying rate under normal operations.

Counter-evidence found (or absence)

The brief explicitly flags that the July 12 resolution date truncates observation to ~6 weeks of the 10-week window (key_uncertainties and edge_cases sections). This is not a minor caveat; it directly reduces the observable count by roughly 40 % relative to a full-window projection. The NEPA transition and statements-of-reasons under-count are also noted but are secondary to the truncation effect.

Distribution implications

Because the high-evidence moderate-rate scenario must be scaled to the truncated observation period, the 20–30 bin receives the largest share. The 10–20 bin gains mass to reflect downside risk from transition disruption or slower NCA publication; the 30–40 bin retains only residual probability. Bins above 40 receive negligible weight, as even the elevated-rate scenario is implausible once truncated.

Consolidate

Strategy: logit_mean • Drafts: 3/3

Reviewers (3/3)

anchoring → shift_higher (low magnitude, medium confidence) anchoring_v1

The draft rationale’s “contrarian_v2” variant explicitly flags that the July 12 resolution date is ~6 weeks before the August 10 window end, arguing this truncates observable data by ~40%. However, the question asks for the full June 2–August 10 window count, not just what has been published by the resolution date. The forecaster appears to have let the contrarian variant partially anchor the distribution lower (shifting mass into bin_2 at 23%) rather than forecasting the full-window outcome. Simultaneously, the moderate-rate scenario’s core Poisson projection of 39–46 decisions is itself derived by assuming referral decisions are “roughly 60–70% of the combined total” of 346 — an assumption with no direct empirical support cited, creating an upward anchoring risk relative to the statement-of-reasons counts (17–20/year) which may be a more reliable (if partial) floor. The modal bin (bin_3, 30–40) is placed at the pessimistic end of the high-evidence scenario range (30–50) rather than near the implied Poisson mean of ~42, suggesting mild downward anchoring from the contrarian variant’s truncation argument influencing mass allocation even though the question is about the full window.

Flagged concerns

• Resolution-date truncation misapplied to full-window forecast: The contrarian_v2 variant argues the July 12 resolution date truncates observation to ~6 weeks, reducing the count by ~40%. But the question asks for the full June 2–August 10 count — the resolution date doesn’t shorten the measurement window; it only means the question is resolved before the full window ends. The draft absorbs this truncation logic, inflating 20–30 (bin_2) to 23% and creating a p50 well below the Poisson mean of ~42.
• 60-70% referral-share assumption used as Poisson mean anchor: The brief’s Poisson projection rests on assuming referral decisions are ~60–70% of the 346 combined ‘referral and approval decisions.’ This ratio is asserted without supporting data and is the key lever producing the ~39–46 modal estimate. If the true referral share is lower (e.g., 50%), the mean drops to ~33; if higher (e.g., 80%), it rises to ~53. The forecaster treats the 60–70% midpoint as modal rather than uncertain, locking in a center that may be optimistic or pessimistic.
• Peak mass at pessimistic end of high-evidence scenario: The high-evidence scenario spans 30–50 decisions with a Poisson mean of ~42. The draft places peak mass in 30–40 (bin_3) at 37% rather than near the conditional mean (~42, which falls in bin_4). This is a mild downward anchor, likely caused by the contrarian truncation argument bleeding into the moderate-scenario weighting.

ceiling → shift_lower (medium magnitude, medium confidence) ceiling_v1

The critical structural constraint the draft underweights is the resolution date truncation: the question resolves on July 12, 2026 — only ~40 days (approximately 6 weeks) after the June 2 start date, yet the observation window nominally runs to August 10/12 (10+ weeks). As the brief’s edge case explicitly states, “the question may be evaluated on partial data — only decisions published up to the resolution date would be observable at resolution time.” At ~17–20 referral decisions/month, 6 weeks yields approximately 25–30 decisions observable at resolution, not the 39–46 projected for the full window. The draft acknowledges this (“Note: The resolution date is July 12…I’m forecasting for the complete window”) but then applies a full-window distribution anyway — this is a structural miscalibration. If the resolver counts only through July 12, bins 2 (20-30) and 3 (30-40) should dominate, with bin 3 (30-40) being the ceiling of the plausible range at resolution time. The contrarian variant correctly identified this but was averaged away by the logit-mean aggregation, leaving the draft over-weighted toward the 30-50 range rather than 20-40.

Flagged concerns

• Resolution-date truncation ceiling: The observation window is June 2 – August 10/12, but resolution occurs July 12 (~40 days in). At ~17–20 referral decisions/month base rate, only ~25–30 decisions are observable by resolution date. The draft explicitly acknowledges this but forecasts for the full window anyway, artificially inflating mass in bins 3-5 (30-60). Bins above 40 are implausible at resolution time under baseline rates unless the resolver uses the full window despite the early resolution date.
• Upper tail ceiling from base rate arithmetic: Even accepting the full 10-week window, the brief’s own Poisson projection centers at 39-46 decisions (source 11: 346/year × 60-70% referral share ÷ 12 × 2.3 months). At ±2σ Poisson uncertainty (λ≈42, σ≈6.5), the 95th percentile sits near ~55. Bins 6-9 (60-90+) collectively receive ~1.7% in the draft but the structural arithmetic implies they should receive even less — the draft is already low here but the rationale doesn’t articulate this ceiling explicitly.
• NEPA transition effect on structural ceiling: The July 1 NEPA transition (sources 6, 7) falls within the window. If the transition causes even a modest administrative slowdown of 2-3 weeks, the effective decision rate post-July 1 could drop, further constraining the upper tail. This is unquantified in the draft but directionally supports a lower ceiling than the base rate implies.

math → shift_lower (medium magnitude, medium confidence) math_v1

The draft distribution assigns 59.2% of probability mass to bins 3–4 (30–50), which aligns with the brief’s high-evidence “Moderate rate” scenario (30–50 decisions), yet the rationale cites a Poisson projection of “~39–46 decisions” as the central estimate. However, the question’s resolution date (July 12) falls ~40 days before the observation window end (August 10), meaning only ~6 weeks of data will be observable at resolution time, not the full 10-week window. The draft’s modal bins (30–50) are calibrated to the full window, not the truncated observation period. This creates an internal inconsistency: if the true window-end count is 39–46, but only ~60% of that window is observable by July 12, the expected observable count should be ~23–28, yet the draft places its mode at 30–40. The distribution should be shifted downward to reflect truncation, or the rationale should explicitly justify why the July 12 truncation is immaterial.

Flagged concerns

• Truncation mismatch: The rationale derives an expected count of ~39–46 for the full June 2 – August 10 window, but the resolution date (July 12) observes only ~6 weeks (40/70 days, ~57%) of that window. The modal bins (30–50, 59.2% combined) should reflect a ~57% reduction from the full-window projection, implying a central mode around 22–26, not 30–50. The draft does not explain how this reconciliation is handled.
• Probability sum check: Summing the draft prediction_map values: 0.0100 + 0.0872 + 0.2334 + 0.3657 + 0.2260 + 0.0574 + 0.0112 + 0.0039 + 0.0027 + 0.0027 = 1.0002, which is within the ±0.01 tolerance (rounds to 1.000). This is acceptable.
• Scenario weighting inconsistency: The draft’s scenario weights claim 81% on bins 3–4 (via 0.46 + 0.35 notation in the gpt-5-mini rationale), yet the published prediction_map shows bin_3 = 0.3657 and bin_4 = 0.2260, summing to 0.5917 (59.2%), not 81%. The internal rationale text does not match the final distribution.
• Window vs. resolution-date slippage: The brief’s edge_cases and key_uncertainties sections explicitly flag that ‘The resolution date of 12 July 2026 is before the stated observation window end (10/12 August 2026), meaning the question will be assessed on approximately 6 weeks of data rather than the full 10-week window.’ The draft’s Poisson projection (39–46) is calibrated to the full window but the distribution is not discounted for partial observability.

Reconcile: partial_shift

Reviews consumed: 3

The three reviewers split on direction. The anchoring reviewer argued for a mild shift higher, pointing out the question asks for the FULL June 2–August 10 window count, so the modal bin should sit closer to the Poisson mean (~42) in bin_4. The ceiling and math reviewers both pushed lower, arguing the July 12 resolution date truncates observable data to ~6 weeks. I side with the anchoring reviewer on this structural question: the question explicitly defines the window as June 2 to August 10/12, and resolution mechanics typically resolve once the window completes (or use whatever the portal shows by resolution time for the full window) — the metric measured is the full-window count, not a partial count. The truncation argument from ceiling and math reviewers conflates “when the question resolves” with “what is measured,” so I reject the medium-magnitude downward shift. However, I accept the anchoring reviewer’s secondary point that the draft’s modal mass sits slightly pessimistic relative to the Poisson mean (~42), so I shift a small amount of mass from bin_3 to bin_4 and trim the lower tail modestly. The math reviewer’s sum check passed; their scenario-weighting inconsistency note reflects the logit-mean aggregation, not a bug. Net result: partial shift, modestly higher mode, retaining reasonable upper tail for NEPA-transition surge possibility.

Rejected critiques

• [ceiling] Resolution-date truncation should compress the distribution toward 20-30 (bin_2) since only ~6 weeks of data observable by July 12.

  • Why rejected: The question’s metric is defined as the count of decisions with publication dates in the full June 2 – August 10/12 window. The resolution date controls WHEN the market settles, but the resolver still measures the full-window count (which will be observable shortly after August 10 even if the market formally resolves July 12, or the resolver will await the full window). The anchoring reviewer correctly identified this misapplication.

• [math] Modal bins should be shifted to ~22–26 due to ~57% truncation factor from July 12 resolution.

  • Why rejected: Same reason as ceiling reviewer: the resolution criteria specify a fixed-window count (June 2 – August 10/12), not a partial-observation count at resolution date. Discounting by truncation would mis-measure the actual resolved quantity.

• [math] Internal rationale text (81% on bins 3-4) doesn’t match published distribution (59% on bins 3-4).

  • Why rejected: The discrepancy reflects logit-mean aggregation across three draft variants (one of which was the contrarian shifted-low variant). The aggregated distribution is the legitimate consolidation; the gpt-5-mini text was only one variant’s rationale.

Final Distribution

Evidence Sources (19)

1. Decisions on referred actions under the EPBC Act - DCCEEW — Since the EPBC Act commenced in July 2000, over 6,000 referrals have been received (as of late 2025), of which more than 1,000 were controlled actions. This implies approximately 240-280 referrals per year on a 24-year average, but recent rates may differ.
2. EPBC Act Public Portal — Referral Decisions 2026 — The EPBC Act Public Portal shows referral decisions being issued regularly in 2026 — examples include EPBC 2026/10431 (Supernode North BESS, Queensland, decided 17 March 2026) and EPBC 2026/10437 (Robertstown BESS, SA, decided 16 April 2026), both classified as “Not a Controlled Action”.
3. EPBC Act Public Portal — Project Referral Summaries — The EPBC Act Public Portal uses EPBC numbers like 2026/10472 (Brisbane Stadium project, “Referral Decision Made” status), 2025/10258 (Google Subsea Fibre Cable, “Referral Decision Made”), indicating the numbering system includes year prefix and 5-digit sequential number — suggesting hundreds of referrals per year.
4. Quarterly assessment performance report - DCCEEW — The DCCEEW quarterly performance report (latest published: July-September 2025 quarter, updated February 2026) tracks EPBC Act referral decisions by industry type, but actual count data is embedded in a Power BI dashboard rather than a plain text table, making it hard to extract directly.
5. Quarterly assessment performance report - DCCEEW — The DCCEEW quarterly assessment performance report for July to September 2025 was last updated 25 February 2026 and covers data through Q3 2025. The report includes a Power BI dashboard showing EPBC Act referral decisions by industry type — but actual numbers are embedded in the dashboard and not visible in plain text.
6. Stronger environmental protection and restoration - DCCEEW — Australia’s Environment Protection Reform Bill 2025 would establish a new National Environmental Protection Agency from 1 July 2026, and all remaining reforms are to commence on or before 1 December 2026 — meaning the EPBC Act and its referral process will still be operating during the June 2 to August 10, 2026 window.
7. Key amendments to the Environment Protection Reform Act 2025 - Norton Rose Fulbright — The Environment Protection Reform Act 2025 and associated legislation received Royal Assent on 1 December 2025. The bulk of substantive reforms are scheduled to commence in 2026, and a National Environmental Protection Agency is planned from 1 July 2026 — creating potential transition uncertainty for the June-August 2026 EPBC referral window.
8. DCCEEW Annual Report 2023-24 (PDF) — The DCCEEW 2023-24 annual report mentions: 16 referral decisions under section 75 (controlled action or NCA decisions), plus 4 referral decisions under section 75 and/or another section — suggesting approximately 20 referral decisions in that annual report snippet, for FY 2023-24. Also: ‘clearly unacceptable’ decisions under section 74B are listed separately.
9. DCCEEW Annual Report 2024-25 - EPBC Act section on transparency.gov.au — The DCCEEW 2024-25 annual report (transparency.gov.au) references 12 referral decisions under section 75 and 5 referral decisions under section 75 and/or other sections, in the context of statements of reasons provided for referral decisions in 2024-25.
10. Environment protections have gone backwards: a case study in EPBC Act reform — According to a Tandfonline journal article on EPBC Act reforms (2025), over 2,500 referral decisions have been captured in the EPBC Act portal in total, with just over 50 of these subject to reconsideration requests — confirming that the portal is the definitive source for counting referral decisions.
11. Outcome 2 – Environment - DCCEEW Annual Report 2024-25, Transparency Portal — In financial year 2024-25 (July 2024 to June 2025), DCCEEW had 346 total “referral and approval decisions” due, of which 305 were on time (88%) and 41 were late — this is “referral and approval decisions” combined, not just referral decisions alone. This implies roughly 346/12 ≈ 29 total (referral + approval) decisions per month combined.
12. HSF/CEIG: Delivering Major Clean Energy Projects – Review of the EPBC Act for Renewable Energy Projects in QLD, NSW and VIC — HSF/CEIG analysis of EPBC Act referrals (as of November 2024) for renewable energy projects (solar, wind, transmission) in NSW, QLD, and VIC found: 20 referrals in 2018, 15 in 2019, 23 in 2020, 29 in 2021, 40 in 2022, 40 in 2023, and 39 in 2024. This is a subset (renewable energy only, 3 states) showing an increase from ~20/year to ~40/year in that sector.
13. DCCEEW Annual Report 2024-25 - Transparency Portal — From the DCCEEW 2024-25 Annual Report (Table 21) referenced on transparency.gov.au, there are tables of “Decisions on EPBC Act referrals by jurisdiction in 2024-25” and “by activity category in 2024-25” — indicating annual totals for referral decisions are tracked officially by year.
14. The latest on EPBC clean energy project approvals - Lexology — According to a Lexology (Herbert Smith Freehills) article from November 2025 analyzing the EPBC Public Portal, 13 of 19 controlled actions referred in 2025 (for renewable energy) are in New South Wales, and the analysis covers referrals 2018-2025. This suggests 19 controlled action decisions for renewable energy alone in 2025 (partial year as of November 2025).
15. Key administration statistics – 3rd Party Appeals and the EPBC Act (Australia Institute, 2015) — As of August 2015 (Australia Institute briefing), approximately 5,500 projects had been referred since EPBC Act commencement in July 2000 — approximately 366 referrals per year over 15 years. Of these, around 1,500 were assessed as controlled actions (27% ratio). This is the long-run average referral intake rate, not the decision rate.
16. Environment protections have gone backwards - Tandfonline (Bell-James et al.) — According to a 2025 academic paper (Bell-James et al.) analyzing EPBC portal data, almost half of EPBC referrals are deemed “not controlled actions” and a notable portion are “NCA – particular manner” (NCA PM); the paper analyzed “all referral decisions from 2011 to 2020” as a dataset. This confirms that ‘referral decisions’ (as distinct from ‘approval decisions’) are the primary output of the referral stage.
17. DCCEEW Annual Report 2023-24 — The DCCEEW 2023-24 annual report states that in 2023-24, 285 referrals were received (cumulative total: 7,610 since Act commencement), and 43 referrals were withdrawn (cumulative 807). This provides a recent annual referral intake rate.
18. EPBC Act Public Portal – Referral summary for North West Shelf Project Extension (EPBC 2018/8335) — On the EPBC Act Public Portal, a reconsidered project retains the same EPBC number (e.g., EPBC 2018/8335 for the North West Shelf Project Extension) and is listed as a single project entry. The reconsideration decision appears as an additional decision event under the same project record, not as a separate project. The portal page for EPBC 2018/8335 shows project status updated to “Approval Decision made” (reflecting the lifecycle progression through original referral decision, reconsideration, and approval), not a new EPBC number.
19. Quarterly assessment performance report - DCCEEW — The DCCEEW Quarterly Assessment Performance Report page (last updated 25 February 2026) still showed the report period as “July to September 2025” (Q1 FY 2025-26), indicating no Q2 (Oct–Dec 2025) data was yet publicly available as of late February 2026.

wooley_rhino • run 9fe8ff6e • $3.26 • 6m1s

Extracted Percentile Values: {0.1: 8, 1.0: 18, 5.0: 35, 10.0: 40, 15.0: 43, 20.0: 45, 25.0: 47, 30.0: 49, 35.0: 50, 40.0: 52, 45.0: 53, 46.0: 53, 47.0: 54, 48.0: 54, 49.0: 54, 50.0: 55, 51.0: 55, 52.0: 55, 53.0: 56, 54.0: 56, 55.0: 56, 60.0: 58, 65.0: 60, 70.0: 62, 75.0: 64, 80.0: 67, 85.0: 70, 90.0: 74, 95.0: 79, 99.0: 87, 99.9: 94}

Model’s Answer: The most important anchor is the latest official throughput figure: DCCEEW’s 2024–25 annual report says there were 276 referral decisions in the full financial year, which implies about 23 per month. A roughly 70-day window therefore points to a baseline around 53 decisions. That is the strongest single datapoint because it is recent, official, and directly about referral decisions rather than broader EPBC workflow.

I then adjust only modestly. There is no strong evidence of a major secular surge or collapse in referral decisions. The July 1, 2026 reform commencement could create some turbulence, but the most plausible effect is timing noise of a few decisions rather than a regime change. The portal/publication layer adds some extra variance because the question resolves on publication dates, not just decision dates, and portal updates can cause bunching. There is also a small ambiguity in the prompt about the exact end date, which is another reason to keep the distribution a bit wider.

So my center is in the mid-50s, with a fairly broad middle reflecting normal monthly variation plus publication/batching noise, and very fat but low-probability tails for unusual portal migration, backlog clearing, or process disruption. I think a low-40s result is quite plausible, a high-60s result is also plausible, below 30 would likely require a real slowdown/disruption, and above 80 would likely require an unusually strong surge or batching/republication effect.

The assumptions that matter most are these. First, that 2026 underlying referral-decision throughput remains roughly similar to 2024–25; I’d put about a 25% chance this is materially wrong. Second, that publication timing on the portal remains fairly close to normal administrative practice rather than being distorted by upgrades or data migration; about a 12% chance this is materially wrong. Third, that the eventual resolver uses a counting interpretation close to “unique projects with qualifying referral decision documents published in the window,” without some quirks around duplicate documents or the date-window inconsistency changing the total by several counts; about a 10% chance this matters materially.

Low-outcome path: a reform-transition slowdown, staffing bottleneck, or portal maintenance period that delays publication into later August. I’d give a roughly 12% chance to outcomes under 40. High-outcome path: batching near end of June/start of July, faster processing, or republication/cleanup effects that inflate the number of qualifying publications. I’d give roughly a 13% chance to outcomes above 70.

Percentile .1: 8 Percentile 1: 18 Percentile 5: 35 Percentile 10: 40 Percentile 15: 43 Percentile 20: 45 Percentile 25: 47 Percentile 30: 49 Percentile 35: 50 Percentile 40: 52 Percentile 45: 53 Percentile 46: 53 Percentile 47: 54 Percentile 48: 54 Percentile 49: 54 Percentile 50: 55 Percentile 51: 55 Percentile 52: 55 Percentile 53: 56 Percentile 54: 56 Percentile 55: 56 Percentile 60: 58 Percentile 65: 60 Percentile 70: 62 Percentile 75: 64 Percentile 80: 67 Percentile 85: 70 Percentile 90: 74 Percentile 95: 79 Percentile 99: 87 Percentile 99.9: 94

Annual referral throughput, statutory publication steps, and fixed public-comment windows constrain achievable counts over any 72-day span.

An institutional timetable change on 1 July 2026 and routine end-of-financial-year activity further compress the scheduling envelope.

Historical annual throughput implies a lower anchor near the low forties for this window.

Contemporary intake serial growth and weekly public-notice batches point toward a midrange near fifty, with pronounced week-to-week volatility.

If the 1 July institutional transition triggers pre-transition clearing, counts will shift toward the high fifties; if it causes a short operational pause, counts will slide toward the low forties.

Concentrated end-of-financial-year processing and periodic batching can move several decisions across the June–August boundary, changing the tally by a few cases.

Primary uncertainties are publication-date noise, portal status labels that mimic final decisions, and occasional late public notifications.

Balancing the lower annual anchor, the recent weekly-notice analogue, intake growth, and institutional timing yields a compressed judgment of 52 formal Referral Decision publications for June 2–August 12, 2026, with modest sensitivity to boundary and operational noise.

Forecast (median): 42.55 formal “Referral Decisions” published on the EPBC Act Public Portal between June 2, 2026 and August 10, 2026.

• Most of the work is rate-driven: The forecast is anchored to historical volumes of EPBC referral-stage decisions, then scaled to the ~10-week window length.
• Recent data is mixed, not clean: There are some signals suggesting a lower recent pace (relative to older averages), but the most relevant recent figure combines multiple decision types, so the referral-only portion is uncertain.
• Timeline helps, but publication can slip: Referral decisions follow a statutory decision clock (about 20 business days), which supports a fairly steady flow—but “stop-clock” pauses and administrative batching can push publication in or out of the window.
• Window is long enough for stability, short enough for uncertainty: Over ~71 days you’d typically expect counts in the low-to-mid 40s, but the lack of live portal run-rate data keeps tails fairly wide (roughly into the 20s on the low side and the 50s on the high side).
• Main risk is the referral share (and current throughput): If the number of referral decisions is materially lower than implied by the combined annual figure, the outcome could fall into the low 30s or high 20s; if higher, it rises toward the 50s.