PSCI methodology.

01Purpose and scope

This document specifies the methodology, composition, calculation, source licensing, and governance of two indices published by Pallet Solutions:

• Pallet Solutions Cost Index (PSCI) -- a weekly composite of cost-input variables that drive pallet pricing.
• Pallet Solutions Pressure Index (PSPI) -- a monthly composite of leading indicators that signal forward pallet demand.

Posture

PSCI and PSPI are reference indices. They are not investment products, not contracts, and not financial benchmarks regulated by IOSCO. Pallet Solutions is the publisher; the underlying federal data is in the public domain. Any party may construct an equivalent index using the same public sources. We disclose the formula and the source licensing so any user may reproduce or audit our calculation.

Architectural commitment

Production code paths fetch directly from BLS, EIA, Census, and BEA public APIs (api.bls.gov, api.eia.gov, api.census.gov, bea.gov). FRED is research-only and never appears in production calculations or citations. See PS Source Spec, Rule 0 (point of ingest governs) for the legal basis.

02Limitations and recommended use

PSCI is the only publicly-available, reproducible composite cost-input index in the pallet industry. Because no comparable benchmark exists, this section documents what PSCI measures, what it does not measure, and how procurement teams should and should not use it. If a downstream user is going to defend a sourcing decision against an internal challenge, they should be able to cite this section.

What PSCI measures

A weighted geometric mean of five federal public-domain series tracking upstream cost pressure on pallet manufacturer cost basis: Wood Pallet PPI, Softwood Lumber PPI, EIA Diesel by PADD, warehouse worker wages, and Paper Containers PPI. A 1% increase in PSCI indicates approximately 1% upward pressure on the cost inputs that determine pallet manufacturer pricing, holding all other factors constant.

What PSCI does not measure

PSCI is not, and is not intended to be:

• A market-clearing price. PSCI does not represent the price any specific buyer pays any specific vendor in any specific market.
• A substitute for a vendor quote. Sourcing decisions should be made off vendor quotes, not off PSCI movement.
• A price prediction. PSCI movement is a backward-looking measure of cost-input changes already published by federal agencies. It is not a forecast.
• A hedge instrument. PSCI is a reference index, not a tradeable benchmark, and is not regulated under IOSCO Principles for Financial Benchmarks.
• A regional benchmark. Diesel is regionalized by PADD; the other four components are national. Buyers in markets with structural cost differences from national averages should expect divergence between PSCI movement and their observed costs.

Known sources of divergence between PSCI and observed pallet prices

Procurement teams will observe pallet prices that diverge from PSCI movement. The most common reasons:

• Demand-side factors. PSCI is cost-input only. Regional pallet demand shifts (a new e-commerce DC opens; a manufacturer closes; seasonal retail surge) will move pallet prices independently of PSCI.
• Regional supply imbalances. Pallet recyclers operate within roughly 150 miles. A regional shortage of recovered pallets will move local pricing without moving PSCI.
• Vendor-specific pricing strategy. Some vendors price to capacity utilization, not cost basis. Their quotes can move opposite to PSCI in a given period.
• Contract terms. Annual fixed-price contracts will not reprice when PSCI moves; spot-market quotes will. Buyers comparing the same vendor's contract price to PSCI movement should expect lag and dampening.
• Quality and certification differentials. Heat-treated, ISPM-15, and food-grade pallets carry cost adders not captured by the broader PPI series. PSCI is a generic-pallet cost-input proxy.

Recommended use cases

PSCI is appropriate for:

• Budget setting. Year-over-year PSCI movement is a defensible input to next-year pallet-spend forecasting at the procurement-team level.
• Procurement timing signals. Significant PSCI movement (e.g., greater than 5% week-over-week or 10% month-over-month) is a signal to revisit vendor agreements ahead of cycle.
• Quote evaluation. When a vendor proposes a price increase, PSCI movement over the same period gives the buyer a reference for how much of the increase is cost-input-driven versus other factors.
• Internal escalation. Procurement teams defending a vendor change to finance can cite PSCI movement as a public, audit-defensible reason.
• Cross-period comparison. PSCI provides a consistent reference for comparing pallet sourcing periods (Q1 2025 vs. Q1 2026) without having to reconcile across changing vendor mixes.

Anti-use cases

PSCI is not appropriate for:

• Pricing pallet quotes off the index. Buyers should quote vendors, not PSCI.
• Settling disputes. PSCI is not a market price; it cannot be the basis for resolving a vendor-buyer pricing dispute.
• Forward contracting. PSCI is not a tradeable instrument and has no associated derivatives.
• Cross-industry comparison. PSCI is a pallet-industry composite. It is not directly comparable to single-component indices (Random Lengths lumber, CME futures), commodity composites (Bloomberg Commodity Index, S&P GSCI), or trucking indices (Cass, DAT).

Empirical validation status

PSCI v1.3 weights and composition were chosen from input-cost shares for NAICS 321920 (wood pallet manufacturing) as documented in BLS PPI methodology and published industry-cost-structure analyses. The weights are reasoned approximations, not the output of a regression analysis against observed pallet pricing.

As Pallet Solutions accumulates buyer-side transaction data through OMS managed-programs operations, future methodology versions may publish:

• Empirical correlation between PSCI and observed pallet pricing across managed-programs corridors
• Weight refinements informed by observed price elasticity to each input
• Regional sub-indices for buyers with concentrated regional footprints

First-mover disclosure

PSCI is the only publicly-available, reproducible composite cost-input index in the pallet industry. Pallet Solutions is aware that publishing an index in a category that has not previously had one carries first-mover risk: the methodology will be scrutinized by procurement audiences, by competitors, by industry trade press, and by source-data publishers. Pallet Solutions is committed to:

• Transparent methodology publication (this document)
• Reproducibility from public data (Rule 0)
• Versioning with advance notice on changes
• Restatement protocol on source-data revisions, not silent corrections
• Honest disclosure of limitations and validation status (this section)

03Source licensing posture

Procurement audiences and competitors will examine our sources. We make our licensing posture explicit upfront.

Public-domain federal data

The majority of PSCI and PSPI components are sourced directly from US federal agencies (BLS, EIA, Census, BEA). These are public-domain datasets. Pallet Solutions ingests, transforms, and republishes these data in derivative composite form with full source attribution. No license fees apply. No third-party permission is required. Citation includes the agency name and series identifier on every published value.

Public-records aggregation

PSPI's WARN component is built by aggregating state Department of Labor public-records databases. WARN notices are public records under federal and state open-records statutes. Pallet Solutions cites the state agency on every WARN-derived figure.

Licensed third-party sources (cite-and-link only)

Several authoritative industry sources are referenced by name in this methodology -- DAT, Cass, ATA, ISM, Random Lengths, Fastmarkets RISI, CBRE, JLL, Cushman, Bloomberg. Pallet Solutions does not ingest, store, redistribute, or recreate the data series of any of these sources. Where we mention them, we link to their free public-facing pages and identify them as the source for buyers seeking that specific data.

04PSCI -- Cost Index composition

PSCI measures the weighted movement of cost inputs that determine pallet manufacturer prices. A 1% increase in PSCI corresponds to approximately 1% upward pressure on pallet prices, holding other variables constant.

All PSCI components are public-domain federal data. Weights sum to 100%. Weights are fixed at publication. Rebalancing requires a versioned methodology release with 90-day advance notice.

05PSCI calculation

Each component is normalized to a base period (January 2024 = 100). Weekly component values are computed by:

• For monthly series (BLS, Census), the most recent published value is held constant for all weeks in the month following the release date.
• For weekly series (EIA diesel), the released value is used for the corresponding week.
• Component normalization: ComponentIndex_t = (ComponentValue_t / ComponentValue_base) × 100

PSCI is calculated as the weighted geometric mean of normalized components:

Geometric mean is used over arithmetic mean because component series are themselves indices (compounding inputs), and geometric weighting prevents disproportionate influence from any single high-velocity component.

06PSCI publication discipline

• Released every Tuesday at 9:00 AM Eastern, reflecting the prior week's data.
• Published values are final at release; restatement only for source-data revisions per the restatement protocol below.
• If any component source has not released its scheduled update by Tuesday 6:00 AM, PSCI publication is suppressed for that week with a note explaining which component is delayed.

Display and attribution

• Headline value: PSCI_t to two decimal places.
• Week-over-week change: percentage to one decimal place.
• Year-over-year change: percentage to one decimal place.
• Component contribution: each component's contribution to weekly change shown in expandable detail.
• Every component value links directly to its publishing agency's source page (BLS, EIA, Census).

07Federal data outage handling

If a federal source goes silent (government shutdown, agency outage, API deprecation), the dependent index publishes "indeterminate" rather than imputing.

Beyond the window, the dependent component is suppressed and the composite index publishes with reduced component coverage, disclosed in the publication footer.

08Restatement protocol

Restatements of previously-published values appear as separate publications, not silent corrections, with the following format:

Restatements are normal data hygiene, not error. Census-dependent components are expected to restate within 90 days of initial publication due to standard Census revision cycles. If the publication pipeline fails to compute a value by 8:55am ET, the publication shows "indeterminate" rather than delaying or imputing.

09Backfill disclosure (look-ahead bias)

Backfilled values for periods prior to PSCI's first live publication are computed by applying the v1.2 formula and weights to historical federal data. Backfilled values are clearly marked as "computed retrospectively" and carry an explicit caveat: weights were chosen in 2026 with knowledge of the period being backfilled, which introduces look-ahead bias. Backfilled values are useful for directional comparison but are not equivalent to live publications. Live values are date-stamped and immutable except via the explicit restatement protocol above.

10Disclosure footer

Every published PSCI/PSPI value, in any surface, carries the following disclosure footer:

11Release calendar

• PSCI: every Tuesday, 9:00 AM ET. Tuesday Read newsletter sends 10:00 AM ET.
• PSPI: second Friday of each month, 9:00 AM ET.
• BLS PPI release: typically 11th-14th of each month (consult BLS release calendar).
• EIA weekly diesel: Mondays, 5:00 PM ET.
• Census Construction Spending / Housing Starts / Retail Trade: per Census release schedule (30-45 day lag).

Full per-release schedule is published in the operations runbook and pinned to each Tuesday Read footer. TBD public-facing release calendar surface.

12AI Forecast Layer governance (optional overlay)

The AI Forecast Layer is an optional forecasting overlay applied to PSCI cost inputs. It produces probabilistic 30-day forward projections with confidence intervals. The Forecast Layer is not a separate index, and the underlying PSCI v1.3 index is unaffected by this overlay. It is published alongside PSCI as a procurement timing signal, not as a substitute for vendor quotes and not as a price prediction.

Model

The overlay is powered by TimesFM 2.5, an open-source time-series foundation model published by Google Research. Approximately 200M parameters, decoder-only architecture, zero-shot forecasting, Apache 2.0 license. Pallet Solutions does not train, fine-tune, or modify the model -- it uses published checkpoints unchanged. The model version is pinned in the methodology footer on every published forecast value.

Deployment

The forecast pipeline runs via BigQuery ML's AI.FORECAST function. Each of the five PSCI cost-input series is projected 30 days forward. The series-level projections are then composed via the published PSCI weights to produce a forecast PSCI value with a confidence band. Per-PADD diesel forecasts are produced by the same overlay applied to per-PADD EIA series. Cached weekly in Pallet Solutions' read-only Supabase cache; the public API serves cache reads only and does not run BigQuery on the request path.

Reproducibility

Any party with BigQuery access can run the same AI.FORECAST query against the same federal data with the pinned model version and obtain the same forecast range, within minor model nondeterminism (typically less than 0.1% on horizon-end values). The reproducibility check is the integrity check, identical to the principle that governs PSCI v1.3 itself.

Confidence intervals

Every published forecast value carries a standard 80% confidence interval. The CI represents model uncertainty, not measurement uncertainty. Procurement teams reading a forecast value should treat the CI as the truth-telling mechanism: when the band is wide, the model is uncertain. Forecasts published without CIs are non-compliant with this methodology and should be flagged.

What the AI Forecast Layer does not do

• It is not a price prediction. The overlay projects cost-input INDEX movement, not pallet prices.
• It does not modify PSCI. PSCI v1.3 is methodology-locked. The overlay is a tool applied to PSCI inputs; it never alters the published PSCI value or its underlying composition.
• It is not a tradable benchmark. Same posture as PSCI -- a reference forecast, not a regulated derivative.
• It is not a substitute for vendor quotes. The forecast informs procurement timing decisions; it does not price contracts.
• It is not PSPI. PSPI is a separate forward-looking demand-pressure index (Section 07-08). The AI Forecast Layer projects cost inputs (the same federal series PSCI uses); PSPI tracks demand-side leading indicators (housing starts, ISM PMI, container imports, etc.). They are different objects.

Look-ahead bias disclosure

Forecasting models trained on past data may show artificially-strong "historical accuracy" on training-window dates. Pallet Solutions does NOT publish historical-accuracy claims using training-window dates. Forward-going accuracy is established by tracking real-time forecast values against subsequent observed federal data; statistics will be published after 26+ weeks of prospective forecasts have accumulated.

Validation roadmap

The AI Forecast Layer's accuracy is currently theoretical pending empirical validation. Two phases are planned:

• Phase 1 (immediate): log every published forecast in a separate Supabase table. After 26 weeks, compute MAPE (mean absolute percentage error) of forecast vs. realized federal indicator values. Publish in a future methodology version disclosure.
• Phase 2 (gated on Phase 1 results): stratify forecast accuracy by market regime (stable / moving / shock). Publish honest stratified accuracy statistics; the overlay is expected to be most useful for routine planning and least useful for shock response.

Versioning

The current version is AI Forecast Layer v1.0 with TimesFM 2.5 pinned. Changes to model version, confidence interval methodology, or pipeline architecture trigger a methodology version bump and 90-day advance notice, identical to PSCI versioning protocol.

PSPI -- Pallet Solutions Pressure Index

PSPI is a separate index from PSCI -- it tracks forward-looking pallet demand pressure (60-90 day lead), not cost inputs. Buyers focused on cost-input methodology can stop reading at Section 12. PSPI documentation continues here for buyers tracking demand-side pressure signals.

APSPI -- Pressure Index composition

PSPI measures forward pressure on pallet demand. Rising values indicate tightening conditions ahead (60-90 day lead). Falling values indicate softening demand. PSPI complements PSCI: PSCI tells you what costs are doing now, PSPI tells you what's coming.

Five of six PSPI components are public-domain federal data. The sixth (ISM PMI) is referenced via its publicly-released headline value. Weights sum to 100%. WARN component is sign-inverted (rising WARN volume contributes to falling PSPI).

BPSPI calculation and publication

PSPI is calculated as a weighted arithmetic mean of normalized component changes (year-over-year):

• Released the second Friday of each month at 9:00 AM Eastern, reflecting prior-month data.
• Lag accommodates Census release schedule.
• Direction interpretation: > +0.5 = tightening pressure; -0.5 to +0.5 = neutral; < -0.5 = softening.
• Each release includes a one-paragraph plain-English read on what the value implies for the next 60-90 days.