How many recall cases and public health alerts will USDA FSIS post between 9 June and 11 August 2026, per the FSIS recalls archive?

• My most likely prediction for the number of distinct FSIS recall cases and public health alerts is 13.50, with a 50% chance of falling between 11.50 and 16.50.
• Historical data from 2022–2024 establishes a consistent baseline of roughly 5 to 8 distinct safety events per month, translating to approximately 13–14 events for a 64-day window.
• The 2026 regulatory environment is notably active, with Q1 2026 recording the third-highest volume of recalled product in two decades (37.1 million pounds).
• An ongoing Salmonella contamination chain linked to California Dairies Inc. continues to trigger rolling notices, though many are likely to be excluded as expansions or updates.
• Strict resolution criteria favor ‘distinct’ new cases, excluding follow-ups and modifications, which serves as a significant limiting factor on the final count.

Key figures

Historical context

The USDA Food Safety and Inspection Service (FSIS) maintains a public archive of all recalls and public health alerts (PHAs). Historically, activity is relatively consistent but can be punctuated by large-scale outbreaks. In 2022, FSIS recorded 45 recalls and 23 PHAs (68 total). In 2023, this increased to 65 recalls and 24 PHAs (89 total). By 2024, the numbers moderated slightly to 56 recalls and 20 PHAs (76 total). Common triggers include product contamination (38%), unreported allergens (30%), and misbranding (29%). Major past events, such as the 2024 Boar’s Head listeriosis outbreak (7 million pounds) and the 2026 California Dairies Salmonella incident, demonstrate how a single source of contamination can ripple through the supply chain for months, though many subsequent notices are categorized as expansions rather than new cases. Seasonally, production often ramps up in the summer months, which can lead to a slight increase in reporting during the June–August window.

Tailwinds

• Heightened regulatory focus on ‘Country of Origin’ labeling and ingredient transparency by the current administration, leading to more misbranding alerts.
• High inventory levels at major processors like Tyson Foods ($5.48B in March 2026) and Hormel, increasing the total ‘product at risk’ in the supply chain.
• The ongoing rolling nature of the California Dairies Salmonella incident, which may still yield entirely new distinct cases in previously unaffected product categories.
• Seasonal summer production peaks that historically correlate with higher volumes and increased potential for process failures.

Headwinds

• Strict resolution criteria that exclude expansions, reclassifications, and follow-up recalls, which can significantly deflate the final ‘distinct’ count even during active outbreaks.
• Historical data from 2022 and 2024 suggesting lower annual totals (68 and 76 respectively), which could imply a lower summer baseline.
• Improved industry testing and ‘voluntary’ internal holds by major processors like Tyson and Hormel that may prevent issues from reaching the level of a public FSIS posting.

Detailed reasoning

My prediction is based on a synthesis of historical annual base rates and current 2026 industry trends, adjusted for the specific ‘distinct entry’ criteria defined in the question.

Historical analysis shows that the USDA FSIS typically manages between 68 and 89 safety events (recalls and Public Health Alerts) per year. Specifically, 2022 saw 68 events, 2023 saw 89, and 2024 saw 76. On average, this equates to roughly 78 distinct events annually, or approximately 6.5 per month. The forecast window from June 9 to August 11, 2026, spans 64 days, representing roughly 17.5% of the calendar year. Applying the historical average to this timeframe yields a baseline expectation of 13.6 distinct events.

However, the 2026 regulatory environment has shown signs of increased activity and severity. First-quarter data from 2026 reported 37.1 million pounds of recalled product, the third-highest quarterly total in two decades. Furthermore, a major Salmonella contamination event involving California Dairies Inc. powdered milk has triggered a ‘rolling’ series of downstream recalls and alerts throughout May and early June 2026. While many of these downstream actions might be classified as ‘expansions’ or ‘updates’—which are excluded by the resolution criteria—the sheer scale of the incident increases the likelihood of discovering distinct, unrelated safety failures during the subsequent ‘heightened surveillance’ period.

The ‘functional rule’ in the resolution criteria is a critical downward pressure on the final count. It specifies that any entry that acts as a follow-up, reclassification, or expansion does not count as a new distinct entry. This means that even if the FSIS archive shows 20 new postings in the window, if five are expansions of existing recalls and three are formal recalls replacing previous Public Health Alerts, only 12 would count.

Consequently, I have weighted the historical baseline (approx. 13-14 cases) most heavily (80%). I also factored in a 20% probability of an elevated regime (18-20 cases) to account for the active contamination environment and the possibility of new, distinct pathogen outbreaks or misbranding clusters. The most likely value of 13.5 reflects the intersection of these two scenarios, staying close to the historical mean while acknowledging the upward volatility observed in early 2026. The 50% HDI of 11.5 to 16.5 accounts for the typical Poisson-like variance expected in food safety administrative data.

Key uncertainties

• The ‘Functional Rule’ interpretation: The exact number of postings that FSIS characterizes as ‘updates’ versus ‘distinct cases’ can be subjective and varies by incident.
• Emergence of a new, unrelated pathogen outbreak (e.g., E. coli or Listeria) during the summer window could cause a significant spike.
• Precision of FSIS archive posting dates: Events occurring late in the window might not be posted until after August 11, falling outside the resolution period.

Conclusion

• The forecast identifies 13.5 as the most likely number of distinct FSIS postings, primarily driven by a historical baseline of 5-8 events per month.
• While 2026 has seen record-breaking product volumes (37.1 million pounds in Q1), the strict resolution criteria excluding follow-up alerts and recall expansions will likely keep the distinct case count within historical norms.
• Decision-makers should monitor the resolution of the California Dairies Salmonella chain; if it transitions from rolling updates to entirely new distinct incidents, the count could push toward the upper bound of 19.5.

The question asks for the number of distinct new FSIS recalls plus public-health alerts posted between 2026-06-09 and 2026-08-11, counting only distinct incidents and excluding expansions, reclassifications, retail-list updates, corrections, and PHA→recall follow-ons. Forecasters agree the best anchors are same-window analogues: reconstructed 2019–2024 windows average about 9 (median ≈8.5, sd ≈3), with 2024 verified at 8 and the exact 2025 window reconstructed at 9, so a high-single-digit center is appropriate. Annual totals imply higher raw posting rates (which scale to ~12–16 per 64 days), but teams downweight those because seasonality, clustering, and follow-ups inflate archive volume relative to distinct-case counts. Pre-window 2026 indicators show moderate activity but many updates/retractions and follow-ons, reinforcing the view that some headline items will not count. Modelers recommend an overdispersed Poisson-style distribution centered near 9 with plausible roughly 5–14 as a central credible range (p10≈5, p50≈9, p90≈13) and a modest right tail for an independent outbreak or multiple routine compliance cases. There is broad agreement on methodology and central estimate; no team argued for a large structural shift upward absent clear outbreak signals entering the window.

Forecast and Reasoning

Forecast: Probability distribution:

• 10.00% chance of value below 6.079709
• 20.00% chance of value below 8.082019
• 40.00% chance of value below 11.227254
• 60.00% chance of value below 14.233619
• 80.00% chance of value below 18.306282
• 90.00% chance of value below 22.424201

Forecast Reasoning

We are forecasting the number of distinct FSIS recall cases plus distinct FSIS public health alerts with posting dates from 9 June 2026 through 11 August 2026 inclusive (a 64-day window, about 17.5% of a year). Recent-year anchors cited by the team for annual volumes of recalls+PHAs are mostly in the mid-60s to low-80s per year, which prorates to roughly 11–14 distinct items over 64 days. Some analysts also noted 2026-to-date cadence appearing somewhat lower (roughly 3–5 distinct items per month in early 2026), which, if it persists, would pull the expectation down a bit.

The resolution criteria’s “functional rule” (exclude expansions/updates/reclassifications; treat a later recall following a prior alert as a non-new entry) means raw archive posting counts can overstate distinct-case counts during clustered incidents. That pushes the distribution slightly downward and adds uncertainty (because clusters can either inflate raw postings via updates or compress distinct counts when multiple postings map to one incident).

Synthesizing these considerations, I center the distribution near 13 distinct entries for the 64-day window, with a moderately wide spread (overdispersed relative to a simple Poisson) to reflect burstiness/outbreak clustering and the possibility of an unusually quiet summer period.

A typical amateur approach would extrapolate from a single year or a recent month and give a narrow point estimate (e.g., “about 12”) without accounting for (a) the distinctness rules that exclude follow-ups/updates and (b) the real overdispersion/burstiness in recalls/alerts. My forecast explicitly rate-anchors on multiple recent-year annual totals, scales to the exact 64-day window, then widens the tails to reflect clustering and the distinct-case filtering rule. Confidence is moderate: the base rate is fairly stable, but short windows can swing notably due to a few outbreaks or an unusually quiet period.

Forecast rationale (numeric):

— Iteration 1 — Across the forecasts, the main reasoning pattern is to anchor the estimate in recent FSIS activity and then adjust for count uncertainty and seasonal clustering risk.

• Baseline from recent and historical rates: The forecasts use Jan–May 2026 activity and longer-run FSIS norms to project the 64-day window. This yields a central expectation in the high single digits to around ten postings.
• Strict counting rules matter: A major adjustment is that only distinct recall cases and public health alerts should count, while updates, follow-ups, and related reposts are excluded. This keeps the projected total from being inflated by multi-part incident reporting.
• Overdispersed count process: The reasoning treats postings as a Poisson-like but overdispersed event count, reflecting that FSIS recalls can come in clusters rather than arriving smoothly over time.
• Seasonality and outbreak risk: Several rationales note that summer months can bring elevated foodborne-risk activity (e.g., Listeria or E. coli), which could push counts above baseline if a cluster occurs.
• Quiet-period downside risk: The lower end of the range reflects the possibility of an unusually quiet window, consistent with occasional low months in the recent archive.
• Wide uncertainty bands: All forecasts preserve substantial tail risk because of archive/counting ambiguity, clustering, and limited access to the underlying FSIS archive.

Overall consensus: the most likely outcome is a moderate number of postings, roughly 8–10, with meaningful upside from outbreak clusters and some downside if the period is unusually quiet.

— Iteration 2 — Overall, the forecasts use a rate-based extrapolation from recent FSIS activity, then adjust for the 64-day window, seasonality, and uncertainty about what counts as a distinct posting.

Main reasoning patterns

• Recent pace as the primary anchor:
  The strongest reference point is April–May 2026 activity, which appears to run around 3–4 distinct recalls/alerts per month. Extending that to the full window implies a central expectation in the mid-single digits to high single digits.

• Historical average as a counterweight:
  One line of reasoning gives more weight to longer-run FSIS frequency, which is higher than the recent spring pace. That approach pushes the expected count upward, suggesting that the recent months may understate the true underlying rate.

• Summer seasonality may modestly raise activity:
  The June–August period is treated as somewhat prone to elevated foodborne illness and outbreak-related activity, supporting a slight increase over a simple monthly extrapolation.

• Event clustering and right-skewed risk:
  The distributions are described as right-skewed, reflecting the possibility that one contamination episode could generate several recall or alert postings in a short span. This creates meaningful upside tail risk even when the central estimate is moderate.

• Follow-up and classification uncertainty:
  There is some ambiguity about whether follow-up notices or related postings are counted separately, which can materially affect the tally.

Areas of consensus

• The most likely outcome is not extremely low, but also not a large spike under normal conditions.
• A mid-range outcome is favored, with modest variation around it.
• The forecast should allow for both quiet periods and clustered outbreak-driven bursts.

Main disagreement

• The key divergence is between a recent-activity approach that implies roughly 6–9 events and a historical-rate approach that implies something closer to around 14 over the window.
• The difference mostly comes from how much weight to place on the latest two months versus longer-run FSIS norms, and whether recent aggregator data may be missing smaller postings.

— Iteration 3 — Across the forecasts, the main reasoning pattern is to anchor on recent FSIS activity in spring 2026, then scale that rate to the 64-day window from 9 June to 11 August 2026. All three models treat the expected count as a moderate, not high, number of distinct recall cases/public health alerts, with recent monthly posting rates implying a low-to-mid single-digit monthly pace.

Key factors driving the estimates

• Recent trend weighting: March–May 2026 activity is the primary baseline.
• Window-length scaling: Estimates are converted from a monthly pace into a 64-day total.
• Deduplication / archive rules: Models adjust downward for the fact that the archive counts distinct cases, not every update or retraction.
• Historical comparison: Broader FSIS history is used as a sanity check, generally suggesting somewhat higher counts than the recent spring trend alone.
• Clustering risk: A few models emphasize that food safety events can cluster, so one incident can spawn multiple recall entries or a burst of postings.

Areas of consensus

• The most likely outcome is in the high single digits to low teens.
• A reasonable central estimate is around 8–13 cases, with the center of mass near 10.
• The distribution is right-skewed: quiet periods are possible, but the main upside risk comes from outbreak clusters rather than gradual drift upward.

Main differences in emphasis

• One forecast leans more heavily on the recent low spring rate, pulling the estimate toward the lower end of the range.
• Another balances recent data against higher historical averages, leading to a somewhat higher midpoint.
• All three allow for a substantial upper tail, but differ on how much probability to assign to very high counts.

Overall synthesis

The collective view is that FSIS is expected to post a moderate number of distinct recall/public health alert entries in the target period, probably around 10, with most plausible outcomes in the upper single digits to low teens and a smaller chance of an unusually busy stretch driven by clustered incidents.

Summary of Forecaster Reasoning

1. Historical Base Rates vs. 2026 Slowdown Forecasters universally agree that the 64-day resolution window (approximately 2.1 months) must be evaluated against both multi-year historical averages and the specific run-rate of 2026. Historically, a similar timeframe yields a higher baseline of expected cases. However, 2026 has demonstrated a distinct structural slowdown in the frequency of enforcement actions. Forecasters attribute this sluggish pace to industry consolidation—which results in fewer but much larger “mega-recalls”—as well as potential shifts in FSIS strategies toward proactive assessments and federal staffing shortages.

2. The Strict “Distinct Case” Deduplication Rule The most critical factor cited across the rationales is the strict resolution criteria for a “distinct” case. Forecasters heavily emphasize the “functional rule,” which dictates that updates, expansions, reclassifications, and foreign-language translations do not count as new entries. Notably, if an ongoing Public Health Alert (PHA) upgrades to a formal recall, it is treated as an update. Forecasters agree this rule will severely suppress the final count by filtering out the noise of cascading supply-chain alerts.

3. Current Outbreaks and Upward Regulatory Pressures Despite the overall slowdown, forecasters identify several upward pressures. The FSIS recently implemented stricter testing regulations, including expanded Listeria and allergen verification programs, which naturally elevate baseline enforcement actions. Furthermore, forecasters note ongoing, widespread contamination investigations—most prominently a massive FDA-level Salmonella issue in dairy/milk powder that affects FSIS-regulated meat and poultry products. While this presents upside risk, many forecasters argue that the strict deduplication rules mean subsequent downstream alerts will likely be grouped as updates rather than distinct new cases.

4. Seasonality and Administrative Lag Summer traditionally sees a spike in bacterial illnesses (Salmonella, E. coli) tied to warmer temperatures and outdoor grilling. While some forecasters expect this seasonal trend to pull the 2026 run-rate slightly upward during the window, others caution that epidemiological traceback and administrative lags usually delay the corresponding regulatory recalls until late August or September.

5. Clustering and Distribution Strategy Because food safety violations are inherently “bursty” and subject to multi-state supply chain failures, forecasters favor distributions that account for systemic variance. While the most likely outcome balances the sluggish 2026 baseline with a mild summer bump, distributions generally reflect right-tail risks to account for the possibility of sudden, independent clusters of distinct manufacturer recalls.

Forecast — Post 414

Modal: 10-15 (34.0%) • partial_shift • 8m23s

Interpretation

Reading: strict

The resolution criteria are unusually detailed and prescriptive, specifying precise deduplication logic (functional override of unique identifiers, multi-language collapsing, follow-up exclusions) and a named data source read at a specific post-resolution snapshot date. This narrows interpretation considerably and demands a literal, rule-by-rule application rather than a casual count of archive entries.

Time bounds

• observation_window: Postings with a Posting Date from 2026-06-09 to 2026-08-11 (inclusive), as the archive appears on or after 2026-08-12
• resolution_date: 2026-08-11

Edge cases identified

1. A new recall number could be issued for what is functionally an expansion of a prior case — resolver must apply the functional rule and exclude it even if the number is technically unique.
2. A public health alert and a subsequent formal recall may cover the same incident — only the first posting counts, not the follow-up.
3. Entries posted before June 9 but later expanded or updated within the window do not count toward the total (the update is excluded; the original is outside the window).
4. Multiple language versions (e.g., English and Spanish entries for the same case) count as one distinct entry.
5. An entry whose posting date falls on June 9 or August 11 is included; the boundary dates are inclusive.
6. FSIS occasionally back-dates or retroactively adds entries — resolution is based on archive state on or after August 12, 2026, so any late postings visible by that date with dates in the window would count.
7. An alert or recall that covers multiple products or companies in a single posting counts as one distinct entry.
8. If FSIS redesigns or takes down the archive before resolution, the question may be difficult to resolve — though this is a data-source-availability edge case.

Research (3/3 variants, shared evidence pool)

Total evidence registered (shared pool): ?

Research Brief

Evidence confidence: medium

Scenario 1: Near-historical-baseline: ~10–15 distinct cases posted in the window [high evidence]

Conditions favoring

Historical 4-year average of ~1.33 distinct cases/week × 9 weeks ≈ 12; Jan–May 2026 monthly data (3–5/month) is consistent with this rate; FSIS reorganization has not yet caused measurable disruption; arrival-process projection mean is ~12 (source 10).

Conditions against

Strict deduplication rules may reduce effective count below raw posting numbers; reorganization could begin to affect workflows mid-window as physical relocations proceed; individual months can deviate substantially from the average.

Scenario 2: Suppressed output: ~5–9 distinct cases due to reorganization disruption [medium evidence]

Conditions favoring

FSIS reorganization involves relocation of ~200 support/administrative staff mid-window (source 16, 17); Senators and consumer groups warned of slowed public communications (sources 18, 19); physical relocations expected to begin during the June–August 2026 timeframe; workforce RIFs had already reduced USDA-FSIS staff by ~8% by March 2025 (source 11); April 2026 showed a dip to ~3 postings (source 14).

Conditions against

FSIS official statement says reorganization does not affect food safety activities (source 13); May 2026 posting cadence returned to baseline ~5/month (source 15); recall and PHA postings are driven by contamination events, not primarily by administrative capacity; physical relocation had not yet occurred as of early June 2026 (source 17).

Scenario 3: Elevated output: ~16–25 distinct cases due to outbreak cluster or disease season [medium evidence]

Conditions favoring

2023 showed 89 combined cases (~7.4/month), demonstrating that elevated periods are historically plausible; arrival-process projection p75=16, p90=24 (source 10); summer months can see increased food safety incidents; one high-profile outbreak can trigger multiple related recalls.

Conditions against

No current evidence of an emerging outbreak cluster as of early June 2026; 2025 and 2024 showed notably lower annual rates (66 and 53 respectively); reorganization may suppress rather than elevate the posting rate; the strict deduplication criteria would exclude multi-case expansions.

Scenario 4: Very low output: ≤4 distinct cases due to severe reorganization disruption or data gap [low evidence]

Conditions favoring

If the physical relocation causes acute workflow disruption to recall communication chains mid-window; if the June 9 – August 11 window overlaps with the most disruptive phase of staff moves; arrival-process p5=1 (source 10) shows the low tail is non-negligible; April 2026 already showed a dip (source 14).

Conditions against

No historical period in 2022–2025 shows a 9-week stretch with ≤4 distinct cases; FSIS explicitly stated food safety activities continue uninterrupted (source 13); May 2026 returned to normal cadence (source 15); food contamination events are driven by external factors, not agency capacity alone.

Key uncertainties

• Whether the FSIS reorganization’s physical relocation of ~200 staff will materially disrupt recall and public health alert communications workflows during the June–August 2026 window
• Whether any major food contamination outbreak or cluster emerges during the window, which would drive the count toward the high end of the historical range
• How strictly the resolution criteria’s deduplication rules (excluding expansions, updates, reclassifications, multi-language duplicates) will reduce the raw posting count relative to what auxiliary trackers would report
• The actual timeline for FSIS staff physical relocation — if it occurs mid-window, the second half of the window could see different posting rates than the first
• Whether the 2026 annual trajectory (Jan–May: ~19 raw postings, ~1 retraction) continues at its current pace or reverts to a higher or lower mean

Forecast variants (3/3)

Variant 0: inside_view (inside_view_v1)

Inside View Causal Chain Analysis

Current state → mechanism → outcome:

The key causal mechanism here is: FSIS receives contamination/mislabeling reports from inspectors/industry → internal review → posting to archive. The rate is fundamentally driven by food safety incident frequency (external), not administrative capacity (internal).

Base rate anchor (high evidence): Per sources 2–5 (authoritative FSIS annual summaries), the 4-year average is ~69 distinct cases/year ≈ 1.33/week. The arrival-process projection (source 10) for 9 weeks (June 9 – Aug 11) gives mean≈12, median≈10, p25≈5, p75≈16, p95≈30. This is tagged relative_evidence_strength: high and I treat it as the dominant quantitative anchor.

2026 in-year trajectory: Jan–May 2026 (auxiliary tracker, source 7): 1+5+5+3+5 = 19 raw postings over 5 months (~3.8/month or ~0.88/week). This is slightly below the 4-year average (~5.75/month) but consistent with the 2024 trough year (4.4/month). Under strict deduplication (excluding updates, expansions, multi-language), the effective rate is likely somewhat lower than raw counts.

Reorganization disruption risk (medium evidence): FSIS explicitly stated the reorganization “does not affect frontline personnel or food safety activities” (source 13, authoritative). May 2026 returned to baseline ~5/month (source 15) despite the April reorganization announcement. Physical relocation had not yet occurred as of early June (source 17). Senate/CFA warnings (sources 18, 19) are forward-looking concerns, not observed disruptions. The relative_evidence_strength for the suppression scenario is “medium” — I give it some weight but not dominant.

Strict deduplication effect: The resolution criteria exclude expansions, multi-language versions, and PHA→recall upgrades. This likely reduces raw counts by ~10-20%. If raw rate is ~4/month, distinct rate under strict criteria is ~3.5/month.

Distribution construction:

• Primary scenario (high evidence): ~10–15 distinct cases → bins 3 (10-15) and 4 (15-20) should carry most mass
• Suppressed scenario (medium evidence): ~5–9 cases → bins 2 (5-10) carries meaningful weight
• Elevated scenario (medium evidence): ~16–25 → bins 4 and 5 carry some weight
• Very low scenario (low evidence): ≤4 → bins 1 gets small weight

The arrival projection mean is ~12, median ~10. I center the distribution around bins 3-4 (10-20), with the 2026 YTD being slightly below historical average, pulling the mean modestly lower. Given the strict deduplication will reduce raw counts somewhat, I center around 9-12.

Final allocation:

• Below 0 (bin_0): ~0 (impossible)
• 0-5 (bin_1): ~5% (very low tail, low-evidence disruption scenario + p5 of arrival process)
• 5-10 (bin_2): ~25% (suppressed scenario + lower 2026 rate + deduplication effect)
• 10-15 (bin_3): ~35% (high-evidence baseline scenario centered at median ~10-12)
• 15-20 (bin_4): ~20% (upper baseline, partial elevated scenario)
• 20-25 (bin_5): ~9% (elevated scenario, p75-p90 territory)
• 25-30 (bin_6): ~4% (elevated scenario upper tail)
• 30-35 (bin_7): ~1% (extreme tail)
• 35-40 (bin_8): ~1% (extreme tail)
• Above 40 (bin_9): ~0%

The inside view here strongly supports the 10-15 range as modal, with meaningful spread given year-to-year volatility in FSIS posting rates.

Variant 1: outside_view (outside_view_v1)

Reference class: historical USDA FSIS combined recalls + public-health-alert postings (annual totals 2022–2025: 68, 89, 53, 66 → four-year average ≈ 69/year ≈ 1.33 distinct cases/week) (research brief facts; arrival-projection basis). Base-rate implication: a 9-week window (2026-06-09 to 2026-08-11) anchored on that base rate implies a mean near 12 distinct items (1.33 × 9) with wide historical dispersion (brief arrival-process projection: mean≈12, median≈10, p25≈5, p75≈16, p90≈24, p95≈30) (research brief source 10). Adjustment: I widen the distribution modestly to reflect (a) the brief’s strict deduplication rules that will reduce raw posting counts (deduplication/exclusion rules per resolution criteria/interpretation), and (b) moderate uncertainty about short-term operational effects from the April 2026 FSIS reorganization (FSIS statement says frontline unaffected, but Senators/advocates warned of potential communications slowdowns; relocation had not yet occurred as of early June) (research brief facts 13, 16–19). Because the brief labels the near-historical baseline scenario as the strongest evidence (relative_evidence_strength: high), that scenario receives the largest single share of mass; medium-strength scenarios (suppressed and elevated) get materially less mass, and the low-evidence very-low-output tail gets a small allocation. Reading: strict — I apply the resolution criteria literally (exclude expansions/updates, collapse multi-language duplicates, treat follow-up recall for same incident as not new). How the probability mass maps to bins (and why): - 10–15 (bin_3): 47% — centers on the base-rate arrival-projection mean/median (≈10–12) and is weighted by the brief’s high-evidence “near-historical-baseline” scenario (research brief scenario #1, relative_evidence_strength=high). - 5–10 (bin_2): 20% — captures the brief’s medium-evidence “suppressed” scenario (≈5–9) plus part of the lower tail from the arrival-process projection (p25≈5) (research brief scenario #2, relative_evidence_strength=medium). - 15–20 (bin_4): 11% and 20–25 (bin_5): 10% — together 21% for the medium-evidence “elevated” outbreak-tail (p75–p90 region) supported moderately by historical high-year variability (2023 = 89) and the arrival projection’s upper tail (research brief scenario #3, relative_evidence_strength=medium). - 0–5 (bin_1): 10% — the low-evidence possibility of severe suppression / unusually few distinct items (arrival p5 ≈ 1 and brief scenario #4, relative_evidence_strength=low). - 25–30 (bin_6): 2% — a small tail consistent with the arrival projection’s p95≈30. - bins 0,7–9: 0% — negative counts impossible; extremely large counts (>30–40) are historically implausible in a 9-week window given the 2022–2025 record. Calibration notes (outside view): I anchored on the historic rate (four-year average ≈69/year) and adjusted only modestly for known risks (reorganization uncertainty, strict deduplication) so that the final distribution is concentrated near 10–15 but allows credible upper- and lower-tail outcomes per the brief’s computed arrival-process spread (research brief evidence and scenario strength tags).

Variant 2: contrarian_v2 (contrarian_v2)

Adversarial reading

The implicit consensus from the brief is that the count will land near the historical arrival-process mean of ~12 (high-evidence scenario: 10–15 distinct cases). Authoritative annual summaries (sources 2, 3, 5), Jan–May 2026 auxiliary data (source 7), and the computed projection (source 10) all converge on a weekly rate of ~1.33 that produces this central band over the 9-week window.

Counter-evidence found (or absence)

Searched for counter-evidence; nothing substantive found; consensus appears robust. The only prospective counter-factor (FSIS reorganization and staff relocation) is tagged medium/low evidence strength, has not yet begun (source 17), is explicitly stated by FSIS not to affect food-safety activities (source 13), and produced no observable suppression through May 2026 (source 15). Strict deduplication rules are already incorporated into the “distinct cases” framing of the high-evidence scenario, so they do not shift mass away from the 10–15 band.

Distribution implications

Because the adversarial test found no material counter-evidence, the distribution receives modestly higher mass on 10–15 (bin_3) than a naïve average of all four scenarios would suggest, and lower mass on the low tail (bin_1) than the low-evidence “very low output” scenario alone would warrant. Bin_2 still receives substantial weight to reflect possible modest downward adjustment from strict deduplication, but the overall shape stays centered on the high-evidence historical baseline.

Consolidate

Strategy: logit_mean • Drafts: 3/3

Reviewers (3/3)

anchoring → shift_lower (low magnitude, medium confidence) anchoring_v1

The draft’s p50 sits squarely in the 10–15 bin, which corresponds to the arrival-projection mean (~12), not its median (~10). The median of the Poisson/NB process is explicitly stated as ~10 in the brief (source 10: “mean≈12, median≈10”), so the true modal expectation is in the 5–10 or lower-10–15 range, not pinned at the mean. Concretely, the 2026 YTD auxiliary data (Jan–May: 1+5+5+3+5 = 19 raw over 5 months ≈ 3.8/month vs. the 4-year average of 5.75/month) suggests the current-year run rate is roughly 34% below the historical average. Applying this lower YTD rate (~0.88/week × 9 weeks ≈ 8) rather than the historical rate shifts the modal expectation toward 5–10. Additionally, strict deduplication rules (excluding expansions, multi-language entries, PHA→recall upgrades) will further suppress the effective count below raw posting numbers. The draft places 40.6% in 10–15 (bin_3) but only 25.8% in 5–10 (bin_2), even though the combination of a below-average 2026 YTD rate and strict deduplication rules both push the conditional mean closer to 7–10, not 10–13. The 10–15 bin is optimistically anchored on the historical multi-year mean rather than the more relevant current-year trajectory.

Flagged concerns

• Modal anchor optimism: The draft centers on the 4-year historical mean (~12/9-week window) rather than the 2026 YTD rate (~3.8/month raw ≈ 8.6 projected, or ~7-8 after strict deduplication). Using the historical mean rather than the current-year-adjusted rate places p50 ~1-2 units too high.
• Deduplication discount underweighted: Strict resolution criteria exclude expansions, multi-language duplicates, and PHA→recall reclassifications. The brief notes this ‘likely reduces raw counts by ~10-20%’ but the draft’s rationale only qualitatively adjusts for this, concentrating mass in 10-15 (bin_3) rather than shifting it toward 5-10 (bin_2).
• Below-average 2026 run rate not fully incorporated: Jan–May 2026 shows ~19 raw postings over 5 months (~3.8/month), versus the 4-year average of ~5.75/month. This 34% below-average pace should shift the distribution leftward toward bin_2 more than the draft reflects. The current forecast gives bin_2 only 25.8% vs. bin_3’s 40.6%, but the current-year data suggests the split should be closer to equal or even favor bin_2.

ceiling → shift_lower (low magnitude, medium confidence) ceiling_v1

The structural ceiling here is set by the bin definitions themselves (max bin is “Above 40”) and the historical annual totals. The brief states that the highest annual total in the 2022–2025 period was 89 (2023), implying a maximum weekly rate of ~1.71/week. Over 9 weeks this yields a structural plausible ceiling of ~15 at the 4-year mean and ~27 at the 2023 peak rate (1.71 × 9 ≈ 15.4 at mean, but 89/52 × 9 = 15.4 using 2023 rate). More critically, the brief’s arrival-process projection explicitly states p95≈30, which means only ~5% of outcomes historically fall above 30. The draft appropriately concentrates mass in bins 2–4, but the strict deduplication constraint introduces a structural downward adjustment that the draft under-weights: the resolution criteria exclude expansions, updates, multi-language duplicates, and PHA→recall upgrades. The brief’s Jan–May 2026 raw rate is ~3.8/month, already below the 4-year average of 5.75/month, and after strict deduplication this could be closer to ~3–3.5/month (~0.75–0.8/week), implying a 9-week window mean of ~7–8, not 12. The draft’s 5–10 (bin_2) allocation of ~25.8% may actually be the structural mode, not 10–15 (bin_3) at ~40.6%, given the combination of (1) below-baseline 2026 YTD rate and (2) strict deduplication shaving additional counts.

Flagged concerns

• Strict deduplication downward bias: The resolution criteria impose strict deduplication (exclude expansions, updates, multi-language, PHA→recall upgrades). The brief’s Jan–May 2026 raw rate is ~3.8/month (below the 4-year average of 5.75/month). After applying 10–20% deduplication reduction, the effective distinct rate is ~3–3.5/month (~0.75–0.8/week), implying a 9-week window mean of ~7–8, not ~12. The draft centers mass on 10–15 (bin_3) at 40.6%, but structurally the adjusted mean falls in 5–10 (bin_2). Mass should shift toward bin_2.
• Annual ceiling constraint on upper tail: The historical maximum annual total is 89 (2023), implying a ceiling weekly rate of ~1.71/week. Over 9 weeks this yields ~15 at even the peak historical rate. The draft already correctly assigns minimal mass (≤3%) to bins 6–9, consistent with this ceiling. No major concern here, but bins 6–9 are structurally implausible and their combined ~3% allocation is appropriate.

math → redistribute (medium magnitude, medium confidence) math_v1

The draft’s probability distribution exhibits an internal inconsistency: the rationale explicitly centers the modal scenario at bins 3–4 (10–20) and describes “10–15 (bin_3): 35%” and “15–20 (bin_4): 20%” as the primary allocation, yet the submitted prediction_map assigns labeled “10-15” (bin_3) 40.6% and “15-20” (bin_4) only 14.9%, inverting the stated weighting. The arrival-process projection (mean≈12, median≈10, p25≈5, p75≈16) supports a distribution peaked at 10–15, but the actual numeric allocation places 7.5% of mass below 5, 25.8% in the 5–10 band, and only 40.6% in the 10–15 band—shifting probability leftward relative to the articulated scenario weights. This mismatch suggests the numbers do not faithfully encode the qualitative reasoning.

Flagged concerns

• Probability allocation inconsistency: Rationale text (inside_view final allocation) specifies bin_1: ~5%, bin_2: ~25%, bin_3: ~35%, bin_4: ~20%, but the prediction_map encodes 0-5 (bin_1): 0.0751, 5-10 (bin_2): 0.2578, 10-15 (bin_3): 0.4056, 15-20 (bin_4): 0.1487. The pattern is close but not exact; bin_2 and bin_3 are swapped in relative emphasis (rationale emphasizes bin_3 as the high-evidence center, but prediction_map makes bin_3 only marginally heavier than bin_2).
• Scenario weighting mismatch: The high-evidence baseline scenario (10–15 distinct cases) is stated to carry dominant weight (35% in the text), yet the prediction_map’s bin_3 receives 40.6%, which is only 4.6 percentage points higher than 25.8% (bin_2). Given that the high-evidence scenario explicitly centers on 10–15, the ratio of mass in the 10–15 range to the 5–10 range should more clearly reflect the asymmetry. The current distribution is too flat between bins 2 and 3.
• Upper-tail allocation: Rationale allocates 20–25 (bin_5): ~9%, but prediction_map encodes 0.0782 (7.8%). Rationale allocates 25–30 (bin_6): ~4%, but prediction_map encodes 0.0257 (2.6%). The discrepancy is small but cumulative; the upper tail (bins 4–6) totals ~32.5% in text but only 23.9% in prediction_map, pulling the distribution’s right tail slightly lower than intended.

Reconcile: partial_shift

Reviews consumed: 3

Both anchoring and ceiling reviewers independently flagged a “shift_lower” concern with similar logic: the 2026 YTD raw rate (~3.8/month, ~34% below the 4-year average) combined with strict deduplication rules implies an adjusted weekly rate closer to ~0.8/week, giving a 9-week expected mean of ~7–8 rather than ~12. Both proposed forecasts that roughly equalize bins 2 (5-10) and 3 (10-15), with bin_2 slightly leading. I accept this as a partial shift because two independent reviewers converged on the same direction with the same evidentiary basis (YTD pace + deduplication). The math reviewer flagged internal inconsistency but pushed the opposite direction (more mass into bin_3, less into bin_2)—I partially reject that because their concern is about mechanical encoding fidelity, not about which forecast is better calibrated given the 2026 YTD signal that the other two reviewers correctly emphasized. The anchoring/ceiling concerns are about external accuracy, which dominates. I shift modest mass from bin_3 → bin_2 and slightly thin the upper tail, ending close to the average of the anchoring and ceiling reviewer forecasts.

Rejected critiques

• [math] Internal inconsistency between inside_view rationale text (which described bin_3 as carrying ~35% and bin_2 ~25%) and the consolidated prediction_map encoding; suggested redistributing more mass into bin_3 and the upper tail.
  • Why rejected: The math reviewer’s concern is about faithfulness of the inside_view variant’s encoding, not about calibration. The consolidated draft is a logit_mean across three variants, so it doesn’t need to mechanically match any single variant’s stated allocation. More importantly, the two other reviewers (anchoring, ceiling) independently argued the distribution should shift LOWER (toward bin_2), not higher (toward bin_3). Following the math reviewer’s redistribution would move the forecast in the direction opposite to the substantive accuracy concerns. I prioritize the external-calibration signal from the two converging reviewers over the internal-consistency signal.

Final Distribution

Evidence Sources (20)

1. Auto-prepended resolver source — Auto-prepended source-query: https://www.fsis.usda.gov/recalls (fetched 2026-06-03T20:04:23.580148Z). value=nil. Excerpt: Recalls & Public Health Alerts | Food Safety and Inspection Service

ALERT: FSIS Issues Public Health Alert for… See more details

Recalls & Public Health Alerts

PHA-05092026-01 Public Health Alert Product Contamination

FSIS Issues Public Health Alert for Headcheese Deli Meat Products That May Be Contaminated With Listeria

Crawford Sausage Co., Inc.

Active

Sat, 05/09/2026 - Current

Indiana, Illinois

Editor’s Note – May 14, 2026: Whole genome sequencing results show that headcheese samples collected by FSIS and produced at Crawford Sausage Co. (Est. 21406) tested positive for the o

2. Summary of Recall and PHA Cases in Calendar Year 2025 — In Calendar Year 2025 (full year), FSIS posted 42 total recalls and 24 total Public Health Alerts (PHAs), for a combined total of 66 distinct cases. Source: FSIS official annual summary, last updated Feb 18, 2026.
3. untitled — FSIS CY2024: 34 recalls + 19 PHAs = 53 total distinct cases for full year 2024 (per FSIS annual summary).
4. untitled — FSIS 2023 annual total: 65 recalls plus 24 PHAs, equaling 89 combined entries.
5. untitled — FSIS CY2022: 45 recalls + 23 PHAs = 68 combined distinct cases (per FSIS official annual summary).
6. untitled — FSIS CY2020: 31 total recalls (no PHA count listed in the 2020 summary since the combined PHA tracking apparently started later). Source: FSIS official 2020 annual summary.
7. USDA Recall Update Monthly Reports 2026 - Powder & Bulk Solids — USDA FSIS recalls+PHAs by month in 2026 (per Powder & Bulk Solids auxiliary tracker): Jan=1, Feb=5, Mar=5, Apr=3 (1 retracted), May=5 (including 1 updated from April). Note: these counts may include updates/expansions that would not count as distinct cases under the resolution criteria.
8. untitled — The FSIS recalls archive (authoritative resolver source) showed as of early June 2026, recent active postings with identifiers: PHA-05092026-01 (May 9, 2026 - Listeria), PHA-05042026-01 (May 4, 2026 - allergen), PHA-04302026-01 (Apr 30, 2026), PHA-04092026-01 (Apr 9, 2026), PHA-03232026-01 (Mar 23, 2026), PHA-03132026-01 (Mar 13, 2026), 001-2026 (Jan 16, 2026 recall). The window for this question starts June 9, 2026, so the floor is 0 so far.
9. untitled — Based on 2022-2025 FSIS data, annual combined recalls+PHAs ranged from 53 (2024) to 89 (2023), with 2022=68 and 2025=66. Average of last 4 years = (89+53+68+66)/4 = 69 per year, implying ~5.75/month or ~1.33/week.
10. untitled — Arrival forecast for June 9 – Aug 11, 2026 (9 weeks, ~63 days) using historical base rate of 1.33 new distinct FSIS cases/week (from 4-year average of ~69/year) with overdispersion=0.5: mean=12.0, median=10, p5=1, p25=5, p75=16, p90=24, p95=30. The wide distribution reflects high year-to-year variance (range 53-89/year).
11. Federal Workforce Data Reveal Impact of Trump Admin RIFs on USDA Food Safety Expertise — Between January and March 2025, USDA-FSIS lost 8% of its total workforce due to Trump administration RIFs (Reductions in Force). USDA also announced a reorganization of FSIS in April 2026, though the reorganization was stated to not impact FSIS inspection operations directly.
12. USDA FSIS Recalls & Public Health Alerts — Authoritative resolution source is the USDA FSIS ‘Recalls & Public Health Alerts’ archive at https://www.fsis.usda.gov/recalls (resolver-named source).
13. Reorganization - Food Safety and Inspection Service - USDA — The FSIS April 2026 reorganization (announced April 23, 2026) explicitly stated it “does not affect frontline personnel or food safety activities, which will continue without interruption,” and the reorganization was focused on administrative, technical, and support operations — not inspection activities.
14. USDA Recall Update for April 2026 - Powder & Bulk Solids — In April 2026 — the same month the FSIS reorganization was announced — FSIS issued only 3 public health alerts (one of which was later retracted), down from 5 in both February and March 2026. Powder & Bulk Solids noted this “reflects a slight dip” but attributed it to normal variation, not the reorganization.
15. USDA Recall Update for May 2026 - Powder & Bulk Solids — In May 2026, FSIS issued 5 total recalls and public health alerts (1 recall + 4 public health alerts, including 1 updated from April), which Powder & Bulk Solids noted was “on par with March and February” — i.e., normal cadence was restored the month after the reorganization announcement.
16. USDA Announces Food Safety and Inspection Service Reorganization, Establishes National Food Safety Center in Iowa — The FSIS reorganization involved relocating about two-thirds of its National Capital Region workforce (~200 employees) to Iowa, Georgia, and Colorado — but the reorganization primarily affects administrative, IT, financial, and training support functions, not the recall/public health communications chain directly.
17. Union Says Relocation of USDA Workers Will Undermine Agency’s Work - Quality Assurance & Food Safety — As of May 2026, FSIS had only sent relocation notices to employees — actual physical relocation had not yet taken place — meaning any disruption to recall-communications workflows was not expected to materialize until the relocation is physically completed (timeline unclear, but FSIS stated it “expects to complete its [reorganization] changes” over time).
18. Slotkin, Klobuchar Raise Alarm about Food Safety and Inspection Service Reorganization — 20 U.S. Senators (including Slotkin and Klobuchar, Senate Agriculture Committee members) wrote on May 29, 2026, warning that the FSIS reorganization could “slow traceback investigations and public communication during multistate outbreaks” and “delay the identification and containment of outbreaks” — but this was a forward-looking concern about future capacity, not documentation of any actual disruption to recall posting already having occurred.
19. USDA shuffling staff out of Washington D.C. - Food Safety News — The Consumer Federation of America stated in May 2026 that the FSIS reorganization “will likely result in a diluted version of the public services currently provided” and “in the short-term, it will likely make the agency much smaller” — a qualitative warning about future capacity, but no documented reduction in recall posting frequency had yet been observed.
20. USDA Recall Update for May 2026 - Powder & Bulk Solids — No data exists as of early June 2026 for FSIS recall/public health alert counts in June, July, or August 2026; the most recent monthly data point available is May 2026 (5 alerts), which showed no post-reorganization suppression relative to the Feb–March 2026 baseline of 5 per month.

wooley_rhino • run face9397 • $3.0 • 8m23s

Inspection capacity, supply-chain scale, and regulatory reporting norms set a slow-moving ceiling and floor on case flow.

Post-2020 institutional shifts and steady growth in public-health-alert usage have lowered the calendar baseline relative to pre-2020 years.

Early-2026 posting activity is subdued, yielding a central expectation concentrated near ten distinct postings in the 9 June–11 August window.

Seasonal production and distribution rhythms raise exposure pathways in summer but have not recently translated into consistent spikes.

A major multi-site contamination or supplier-wide labeling failure would act as a pivot and substantially raise counts.

Sustained low bacterial detection, conservative PHA-to-recall conversion, or dominant single-facility problems keep counts clustered at the lower end.

Classification ambiguity—alerts that convert to recalls and administrative expansions—creates one-to-two case counting sensitivity.

Small-sample variability in the post-2020 era and the possibility of a clustered event produce asymmetry in outcomes and a non-negligible right tail.

Forecast: median = 10.3979

• FSIS posting activity over past years has been fairly repeatable, so for a fixed two-month window it typically lands around the high single digits to low teens.
• The “recent baseline” for 2026 looks softer than 2024–2025, which pulls the expectation down somewhat.
• Staffing/capacity changes are expected to suppress detection and posting rates, but it’s not guaranteed to fully reduce the number of distinct cases.
• There are already active investigations near the start of the window; those are more likely to generate follow-on updates than brand-new distinct entries.
• The main risk is clustering: one larger outbreak or ingredient-linked chain could create several distinct postings quickly, pushing the total upward.