# Price Assessments ## How is the approach chosen for each series? {#how-is-the-approach-chosen-for-each-series} Every price series follows exactly one of the five assessment approaches — compiled, trade-based, formula-based, manufacturing cost-based, or freight-based. The choice reflects the data available for that commodity in that market: trade-based assessments rest on international trade records reported by countries, compiled assessments on publicly available price series, formula-based and freight-based assessments derive a price from related assessments where direct market data is thin or absent, and manufacturing cost-based assessments build it up from the cost of process inputs. Once set, the approach is fixed for the series — it does not shift from month to month — and it is disclosed at the series level: the data legend and the assessment catalog state which approach produces each published series. No single approach takes precedence over the others; each is the appropriate instrument for the data situation it addresses, and all five feed the same normalization stage described in [How are assessments normalized?](#how-are-assessments-normalized). ## Compiled Prices {#compiled-prices} ### What are compiled prices? {#what-are-compiled-prices} Compiled prices are assessments calculated from publicly available price data. Raw data consist of price series—value per quantity—published over a month. These raw data undergo a formatting step that adapts them to the Intratec system. After insertion, statistical validation detects anomalies and possible outliers and removes them. The resulting price series are then typically calculated as a monthly average of closing prices for daily or weekly published price data. Once generated, they are directed to the normalization step alongside all other price assessments. ### How are compiled prices calculated? {#how-are-compiled-prices-calculated} The calculation follows a consistent pipeline: - **Source data:** Raw price series from public sources, collected over the reporting month in daily or weekly intervals. - **Formatting:** Raw data are transformed to match Intratec's data structure and units. - **Validation:** Statistical analysis identifies and removes anomalies and outliers from the formatted data. - **Averaging:** Closing prices are averaged across the month to produce a single monthly figure. - **Normalization:** The monthly average is passed to the normalization step, the standard final stage for all price assessments. Because normalization is shared across every approach, a compiled price arrives in the same form as a trade-based or formula-based assessment, directly comparable within its series. ## Trade-Based Prices {#trade-based-prices} ### What are trade-based prices? {#what-are-trade-based-prices} Trade-based prices are assessments derived from international trade information reported by countries. Because each reporting country records what it traded, at what volume, and at what declared value, this data can serve as a direct source of commodity price signals — provided the underlying records are sufficiently homogeneous to represent a single market assessment. Two price types are produced from this data: - **Unit value** — the total traded amount divided by total quantity for a given country and month, taken directly from customs authority reports with no statistical treatment. - **Transaction price** — a filtered version that removes low-volume records and applies an outlier-detection algorithm to improve accuracy. ### How is data homogeneity handled? {#how-is-data-homogeneity-handled} Data homogeneity is the criterion used to determine whether a set of trade records all refer to the same assessment — the same commodity, similar specifications, and comparable trade conditions. A completely homogeneous dataset would show no variation; in practice, some variance within a single commodity specification is expected and acceptable. To establish homogeneity, Intratec applies a clustering algorithm that groups similar trade transactions based on identified statistical modes and data dispersion. Clustering serves two purposes: it discerns underlying market patterns and reduces the impact of anomalies; and it produces a unique price assessment per cluster, avoiding the "basket" effect that arises when heterogeneous trades are averaged together without separation. If no homogeneity is found in a raw dataset, the data are unqualified for price assessment calculation and are not used. ### How is a unit value calculated? {#how-is-a-unit-value-calculated} A unit value is the total traded value divided by total quantity, taken directly from customs data with no statistical treatment — see [What are trade-based prices?](#what-are-trade-based-prices) for how it compares to the filtered transaction price. ### How is a transaction price calculated? {#how-is-a-transaction-price-calculated} A transaction price applies two sequential filters to improve data quality before arriving at a price assessment. **Volume minimum filter.** Because trade data are usually not provided at the individual-transaction level, minimum volume thresholds are set as a multiple of a typical trade size: - For internationally traded commodities, the minimum is often two or three times the standard cargo size, which varies by region and commodity. - For domestic or small-quantity trades, the minimum is a multiple of the most common transportation unit for the region and commodity. Records below the applicable minimum are excluded before any further processing. **Gross outlier removal.** Trade prices that significantly deviate from the median of all transactions are excluded using the Median Absolute Deviation — a dispersion measure more resilient to outliers than the standard deviation. The outlier-detection algorithm is applied through a simplified model to remove inaccurate data. The result is a price assessment grounded in the bulk of representative transactions, with anomalous or atypical records filtered out at both the volume and price levels. ### What are the limitations of unit values? {#what-are-the-limitations-of-unit-values} A unit value is the simplest trade-based figure — a direct ratio of reported value to quantity, with no statistical treatment. That simplicity is also its weakness: against the transaction prices and compiled prices published alongside them, a raw unit value carries three limitations for commodity price analysis. - **Aggregation bias.** Customs records group heterogeneous products under broad classifications. A unit value blends those distinct commodities into a single average, so the figure can drift away from the price of any specific grade actually traded. - **No outlier removal.** Because no statistical treatment is applied, a unit value does not exclude implausible or erroneous transactions. Transaction prices, by contrast, filter low-volume records and apply outlier detection to keep abnormal values from distorting the assessment. - **Lack of market specificity.** A unit value averages across market destinations, price types, and specifications at once. Compiled prices instead differentiate market destination — wholesale, retail, industrial — price type — spot versus long-term contract — and commodity specification, segmentation a unit value cannot reproduce. For these reasons a unit value serves as a transparent, unadjusted reference, while transaction prices and compiled prices give the more robust basis for tracking a specific commodity market. ## Formula-Based Prices {#formula-based-prices} ### What are formula-based prices? {#what-are-formula-based-prices} Formula-based prices are assessments calculated as a function of the prices of related commodities, economic indicators, currency rates, and similar variables. They are used when direct market data is unavailable or sparse. The methodology identifies which commodities and factors drive the assessment, establishes a mathematical relationship through regression analysis, and then applies current market data to generate the forecast price. Intratec explicitly identifies which assessments in each solution are formula-based. ### How is a formula-based price model built? {#how-is-a-formula-based-price-model-built} The process follows three steps: - **Identify relationships.** Market research determines which commodities and economic factors are most likely to correlate with the assessment (typically raw materials or primary derivatives). - **Fit the regression model.** A regression model is calibrated by finding parameters that align the model's predictions with historical assessment data. - **Calculate the assessment.** The fitted model is applied to current market prices and indicator values to generate the forecast price. The result maintains the currency and quantity units of the assessment being produced, making it directly comparable with other prices in the same series. ## Manufacturing Cost-Based Prices {#manufacturing-cost-based-prices} ### What are manufacturing cost-based prices? {#what-are-manufacturing-cost-based-prices} Manufacturing cost-based prices estimate the operating cost of producing a commodity based on process inputs. These prices reflect the cash outlay required to run a plant, including raw materials (net of by-product credits), utilities, labor, maintenance, operating charges, plant overhead, local taxes, and insurance. Price = function of KPIs, price of raw materials, utilities, labor, and an industrial-cost index. :::text-center ![Manufacturing cost build-up: gross raw materials cost less by-product credits and net utilities form the variable cost; operating labor through plant overhead, taxes, and insurance form the fixed cost; together they total the operating cost.](/static/images/manufacturing_cost.svg) ::: ### Which cost components are included? {#which-cost-components-are-included} Manufacturing cost-based prices incorporate all significant operating expenses: - **Raw materials and by-product credits** — The cost of feedstocks and other inputs, net of revenue from saleable by-products or waste streams recovered during the process. - **Utilities** — Electricity, steam, fuel, water, and other energy sources required for production. - **Labor** — Direct and indirect operating staff. - **Maintenance** — Upkeep, repairs, and replacement of process equipment and infrastructure. - **Operating charges** — Catalysts, consumables, and other process-dependent costs. - **Plant overhead** — Facilities management, administration, and shared services allocated to production. - **Local taxes and insurance** — Statutory and operational taxes, plus coverage for equipment and liability. ### Where do the input prices come from? {#where-do-the-input-prices-come-from} The raw material and utility inputs feeding the cost model are themselves Intratec-assessed figures, drawn from three sources: - **Raw materials and by-product credits** — Intratec's own commodity price portfolio, the Primary Commodity Prices assessments described throughout this methodology. - **Energy references** — Intratec Energy Prices & Markets, which assesses feedstock and fuel prices such as naphtha, natural gas, and power. - **Industrial utility costs** — Intratec Industry Economics & Competitiveness, which publishes industrial utility cost estimates. Because every raw material, by-product, and utility cost feeding the model traces back to its own published assessment, a manufacturing cost-based price stays auditable end-to-end. The same input series also carry forward-looking estimates, so a manufacturing cost-based price can be projected using them. ## Freight-Based Prices {#freight-based-prices} ### What are freight-based prices? {#what-are-freight-based-prices} Freight-based prices are assessments calculated from another assessment by adding or subtracting the cost of moving the commodity between two delivery points. They let a price established at one location stand in for the same commodity priced at a different point in the supply chain, when direct trade data for that second point is thin or absent. Two types are produced: - **Netforward** — the value of the product arriving at the destination terminal, obtained by adding maritime freight and insurance to the origin price. - **Netback** — the value of the product returned to the loading terminal, obtained by subtracting maritime freight and insurance from the destination price. :::text-center ![Netforward builds a delivered (CIF) price by adding maritime freight and insurance to an FOB price; netback works in reverse, recovering an FOB price by subtracting insurance and freight from a CIF price.](/static/images/netback_netforward.png) ::: ### How do netback and netforward differ? {#how-do-netback-and-netforward-differ} They move in opposite directions along the same freight leg: - A **netforward** assessment starts from a price at the loading terminal (an FOB basis) and adds maritime freight and insurance to express the delivered cost at the destination terminal (a CIF basis). - A **netback** assessment starts from a delivered price at the destination terminal (a CIF basis) and subtracts insurance and maritime freight to recover the value at the loading terminal (an FOB basis). The result is reported in the currency and quantity unit of the assessment being produced, so a freight-based price is directly comparable with the other prices in its series. ### Which freight rate is applied? {#which-freight-rate-is-applied} The freight rate reflects the transport form most commonly used for each assessment, even where other modes are sometimes used for shipping. Freight rates and insurance are themselves assessed values, entering the calculation alongside the source assessment: *Price = function of other assessments, freight rates, and insurance.* ## Normalization {#normalization} ### How are assessments normalized? {#how-are-assessments-normalized} :::text-center ![Whichever of the five price types produces an assessment, a shared normalization stage restates it on one common, comparable basis — the same currency, unit, and timeline as every other assessment in the program.](/static/images/normalization_flow.svg) ::: Normalization adjusts gathered market information so every assessment reflects the same base standard, rather than merely averaging trades from a diverse market—an average that would not represent the assessment value accurately. It runs concurrently and iteratively with modeling, not as a strictly sequential pre-modeling step, and corrects four dimensions: - **Location differences** — corrected using freight rates. - **Data quality** — corrected using factors, indexes, and other sources. - **Quantity and specification disparities** — corrected using averages and outlier detection. - **Assessment timing offsets** — the time translation between assessment curves of the same commodity, corrected using mathematical formulas. This shared stage is covered in full in [Comparability](../data-to-assessments/comparability.md). At the end of the stage, values are rounded: historical and preliminary data to three significant figures, forecast data to two. ## Preliminary Prices and Forecasts {#preliminary-prices-and-forecasts} ### What happens when data is insufficient? {#what-happens-when-data-is-insufficient} When official data are unavailable, inadequate, or insufficient for a representative assessment—or when a data-driven approach is not suitable—analysts estimate values based on a wide range of factual market information. This estimation process follows a rigorous methodology: analysts meticulously collect and verify all market data used and propose mathematical models grounded in observed market behavior. Information used in these estimates may include other published assessments, producer or export indices, labor costs, completed transactions, spread and exchange trades, fundamental supply-and-demand data, and other relevant factors. The resulting estimates are treated as preliminary data marked with "(P)" in publications, ensuring transparency about their methodological basis. ### What are preliminary prices? {#what-are-preliminary-prices} Preliminary prices are assessments derived from proprietary mathematical models developed to address delays in official statistics, which are typically released with lags of 1 to 3 months. The models provide the best available price estimates for periods up to the current month, ensuring timely market information without waiting for official data releases. Because preliminary prices are model-derived rather than observed transactions, their accuracy is monitored and they are updated monthly as official data becomes available. ### How are price forecasts produced? {#how-are-price-forecasts-produced} In Primary Commodity Prices, short-term price forecasts for selected commodities are derived from mathematical models using a combination of forecast data: prices of related commodities, economic indices, and industry indicators. These models reflect the most probable scenario for the upcoming six months as of the date they are generated. Forecasts are updated monthly because market parameters can shift significantly within short periods. Like preliminary prices, their accuracy is monitored internally and the models are recalibrated as newer data arrives.