Printable version

Data Quality Objectives

The seven-step DQO process addresses the planning cycle for Triad projects.

The Environmental Protection Agency has developed and refined a framework for planning data collection known as the Data Quality Objective (DQO) process. The DQO process addresses the planning cycle for a project from problem statement through the data collection design to make decisions about whether or not regulatory thresholds are exceeded. Planning for what will be done if the thresholds are exceeded (e.g., design of a cleanup or site reuse strategy) is not considered within the DQO process. The DQO process includes seven steps that are as follows:

1. State the Problem. The initial step of the DQO process includes three components. The first identifies team members, including the decision-makers. The second describes the problem. Partly this is stating the problem in a clear, uncomplicated manner. The second component also includes the development of a CSM equivalent to what is required by the Triad. The third component defines the personnel, budget, and schedule constraints that constrain possible decision options.



2. Identify the Decision. The second step of the DQO process includes four components. The first identifies the principal study question that must be addressed. The second defines alternative actions that might be taken depending on the outcome of the decision. The third combines the study question with alternative actions to form a decision statement. The fourth organizes multiple decisions and decision statements, if required to address the problem.



3. Identify Inputs to the Decision. With the decision statement defined, the third step of the DQO process identifies the inputs to the decision. The third step has four components. The first identifies the kind of information that will be needed. The second identifies the various sources of information that might be used. The third defines decision values or "action levels" that will be the criterion for choosing among alternative actions. The last identifies the sampling and analysis methods that can potentially meet data requirements.



4. Define the Study Boundaries. The fourth step of the DQO process defines the boundaries of the study. This includes defining the target population of interest, specifying the spatial boundaries within which data will be collected, determining the time frame or temporal boundaries for data collection, identifying practical constraints on collecting data, and determining the smallest subpopulation, area, volume, or time (i.e., the decision unit) for which separate decisions must be made.



5. Develop a Decision Rule. The fifth step of the DQO process converts the decision statement into a decision rule, with the decision rule based on the expected inputs to the decision. This involves three separate components. The first selects an appropriate population parameter (e.g., mean, median, etc.) that will be estimated based on the data collected for each decision unit. The second verifies that the decision value/action level will be clearly identifiable given the selected parameter and the data sources that will be used (e.g., detection limits are below action level). The third formulates if-then statements that can actually guide decision-making.



6. Specify Tolerable Limits on Decision Errors. The sixth step of the DQO process recognizes, as does the Triad, that environmental decisions are uncertain. The purpose of the sixth step is to define how much uncertainty can be tolerated when making the decision of interest. The DQO process takes a classical statistical approach to uncertainty, incorporating the concepts of a null hypothesis, the gray region, and false rejection (of the null hypothesis, Type I) and false acceptance (of the null hypothesis, Type II) errors. These concepts are combined with an analysis of the consequences of making an incorrect decision to determine acceptable probabilities of making decision errors.



7. Optimize the Design for Obtaining Data. The final step of the DQO process takes the results of the first six steps, and uses them to select a sampling program design that achieves the desired goals at least cost. This includes finalizing analytical methods and determining sample numbers and techniques.

As conceptualized, the DQO process presents a structure for data collection planning that can be consistent with the Triad approach. In practice, DQOs are frequently defined and implemented to provide either a classical statistical basis for sampling program design (e.g., assumptions of homogeneity, statement of a null hypothesis, specification of Type I and II error rates, selection of a statistical test, etc.). However there is nothing about the DQO process that constrains it to quantitative statistical analysis alone. Historically the DQO process has not been used for developing the types of dynamic work strategies envisioned by the Triad, but there is nothing in the structure of systematic planning using DQOs that prevents this.

In contrast, systematic planning under the Triad approach encompasses activities that extend beyond data collection to determine compliance with some action level or cleanup goal. The Triad anticipates the information needed to develop a CSM that can help evaluate site reuse options, guide remedial design (as needed), and develop a long-term monitoring strategy (as needed).

Printable version