Using the framework up to this point has identified many kinds of data relevant to assessing public value. The next step is to choose the specific variables or points of observation for collecting the assessment data. This is as much a practical problem as an analytical one. Many variables may be relevant for each combination of stakeholder and value type. A few variables that would be relevant to this example are shown in Table 3.

This table illustrates two important aspects of its role in the framework and assessment process. First, it is highly unlikely that applying the framework to any IT investment would lead to variables in all, or even most of the cells of this matrix. This matrix can be thought of as an heuristic device, prompting questions about what might be useful and available variables for each row and column without expecting to fill them. The second is that identifying a specific variable relies on combining information about stakeholder interests, the value type, the impact mechanisms, and the context. This is a complex and demanding process. This section offers additional guidance on choosing the best public return variables for a given assessment. In addition, the other ROI methods described below include many variables and additional methods that can be helpful in that task. However, the more the choice of variables can be tailored to the specific public value context, the more valid and persuasive the assessment is likely to be.

An actual public value assessment should be based, of course, on the best available information. But actual assessments take place in practical settings of limited resources and access to data, plus being part of the additional work needed for internal returns and costs. The priority setting described in the risk analysis section can narrow the field to only the most important public value outcomes. The section below describes additional strategies for choosing the appropriate variables and analysis methods.

The choices of variables and analysis methods for the empirical parts of a public value assessment should be considered together. In terms of basic measurement and analytical methodology, what constitutes an appropriate analysis depends, in part, on the types of data and variables involved.

The scope of this framework, as applied to a particular IT project, could encompass a very wide range of data types. Many kinds of quantitative data from financial sources, operations research, and surveys are appropriate for statistical analysis, modeling, and simulations. Many of the social, political, and intrinsic value returns can be expressed in normalized scales, or may best be revealed in qualitative terms or in simple dimensions that are not suitable for much quantitative manipulation. To help guide the assessment, given the very large number of possible public return data types and variables, the framework provides two kinds of resources. The first is a general scheme suggesting variable types and sources for different value types (see Table 4). The second consists of summaries of the approaches and variables available in a range of existing ROI methods that can be of value in completing an assessment (see Table 5).

A wide range of possible public value data can be identified by the methods presented here. The framework approach is based on the assumption that virtually any kind of data can be useful in describing public value creation, from the most precise quantitative figures available from financial or physical measurements, to material as diverse as the content of blogs or observation of user or customer behavior. A conclusion about public value creation requires an inference, since value does not stalk about wearing a label. Valid inferences about value can be formed from qualitative as well as quantitative data, content analysis as well as statistics. Taking these four principles into account, choices made about how a specific analysis proceeds should be based on three criteria: 1) What constitutes the best data? 2) What kind of analysis is appropriate to that type? 3) Who will be the audience for the conclusions reached? The best kind of data available will be specific to the operational and stakeholder context. The kinds of analysis appropriate to various data types are shown in summary form in Table 4.(13)

Beyond these general considerations, the choice of variables and analyses for an assessment can draw on a volume of existing work on ROI methods for guidance. These methods, summarized in Table 5, vary widely in the number and type of variables used, the scope of public value considered, and the level of analytical detail and technique included. Some are intended for use prospectively, in planning for and justifying an investment. Others are aimed primarily at showing impacts of investments after the fact. They also vary in terms of the degree to which they deal with both internal and public value results of the investment, and whether they are designed specifically for IT or government investments generally. The summary of these methods or models in Table 5 can help in the selection of variables and analysis to fit the IT project.

The SROI (Social Return on Investment) model, the only private sector oriented one in the summary, is included for its special features. Its private orientation refers not to the commercial sector but to a private philanthropic orientation. The method was developed by a San Francisco foundation to assess its social and economic development programs. It illustrates not so much how to assess an IT investment but rather how to deal with personal and community impacts in a systematic and comprehensive way.