383
essential for the determination of product-level performance metrics and Key
Performance Indicators (Chap. 12).
Establishment scale data is not detailed enough for some important applications.
What we really need for FEW systems research is process scale data. The process
scale is the natural scale for calculations of efficiency, inputs, and outputs because
processes by definition (as opposed to establishments) transform inputs into outputs. Most establishments host dozens or even thousands of FEW processes including human, technological, and environmental processes. Working at the establishment
scale confounds these processes, and prevents us from precisely linking
establishment- level inputs with the multiple products and services that are the outputs of a typical establishment. For example, many electrical power plants have
several different types of generators using different processes, in a single facility,
and refineries and farms likewise produce multiple outputs using multiple processes
at the same establishment. Human processes tend to be associated with establishments in n:1 cardinality relationships and to be smaller than establishments.
Processes can be of any size and are not usually associated with establishments.
The most actionable process information concerns major businesses and utilities
and infrastructure processes because these are stakeholders in the FEW system that
concentrate a large volume of FEW decisions in a small number of establishments.
Fortunately, these large operators are among the most sophisticated agents in the
FEW system, and they are often interested in FEW systems research. Industry and
utility consortia like The Sustainability Consortium are forming around data sharing, and these efforts are among the most promising developments for establishment
scale primary data collection and sharing. Notably, product sustainability labeling
efforts are process-level efforts.
The process-scale data is a frontier of data science and sensor technology that
links to industrial engineering, microeconomics, business, behavioral, social science, policy, and systems management. The internet of things, combined with new
business models and regulatory requirements, will move toward process-level data
collection in FEW systems. Process scale data shares data privacy and quality and
computational complexity issues with establishment scale data. Process scale data
may be simpler and more precise than establishment data because processes’ inputs
and outputs tend to be more sharply defined—especially for the outputs that are
products—than for establishments. Many businesses can clearly identify the inputs
and output of a process or a product, and also for the business as a whole including
all its establishments, but do not collect the same data at the level of individual
facilities.
The methods for obtaining data across these scales and resolutions tend to
broadly fit two classes: top-down and bottom-up. Top Down methods involve estimation or approximate measurement from afar (often, from space!) and intrinsically
enforce a “mass balance” so that all is included in the aggregate measure. Typically,
top-down methods are very limited in their resolution, validity is difficult to confirm, and the validity of top-down data degrades further via disaggregation error as
resolution grows finer, but privacy and completeness are not problems. Bottom Up
methods involve direct measurement at the process scale, and subsequent aggregation
14 Data
essential for the determination of product-level performance metrics and Key
Performance Indicators (Chap. 12).
Establishment scale data is not detailed enough for some important applications.
What we really need for FEW systems research is process scale data. The process
scale is the natural scale for calculations of efficiency, inputs, and outputs because
processes by definition (as opposed to establishments) transform inputs into outputs. Most establishments host dozens or even thousands of FEW processes including human, technological, and environmental processes. Working at the establishment
scale confounds these processes, and prevents us from precisely linking
establishment- level inputs with the multiple products and services that are the outputs of a typical establishment. For example, many electrical power plants have
several different types of generators using different processes, in a single facility,
and refineries and farms likewise produce multiple outputs using multiple processes
at the same establishment. Human processes tend to be associated with establishments in n:1 cardinality relationships and to be smaller than establishments.
Processes can be of any size and are not usually associated with establishments.
The most actionable process information concerns major businesses and utilities
and infrastructure processes because these are stakeholders in the FEW system that
concentrate a large volume of FEW decisions in a small number of establishments.
Fortunately, these large operators are among the most sophisticated agents in the
FEW system, and they are often interested in FEW systems research. Industry and
utility consortia like The Sustainability Consortium are forming around data sharing, and these efforts are among the most promising developments for establishment
scale primary data collection and sharing. Notably, product sustainability labeling
efforts are process-level efforts.
The process-scale data is a frontier of data science and sensor technology that
links to industrial engineering, microeconomics, business, behavioral, social science, policy, and systems management. The internet of things, combined with new
business models and regulatory requirements, will move toward process-level data
collection in FEW systems. Process scale data shares data privacy and quality and
computational complexity issues with establishment scale data. Process scale data
may be simpler and more precise than establishment data because processes’ inputs
and outputs tend to be more sharply defined—especially for the outputs that are
products—than for establishments. Many businesses can clearly identify the inputs
and output of a process or a product, and also for the business as a whole including
all its establishments, but do not collect the same data at the level of individual
facilities.
The methods for obtaining data across these scales and resolutions tend to
broadly fit two classes: top-down and bottom-up. Top Down methods involve estimation or approximate measurement from afar (often, from space!) and intrinsically
enforce a “mass balance” so that all is included in the aggregate measure. Typically,
top-down methods are very limited in their resolution, validity is difficult to confirm, and the validity of top-down data degrades further via disaggregation error as
resolution grows finer, but privacy and completeness are not problems. Bottom Up
methods involve direct measurement at the process scale, and subsequent aggregation
14 Data