Center for the Study of Systemic Reform
   in Milwaukee Public Schools






Jeffrey Choppin
University of Wisconsin-Madison
Wisconsin Center for Education Research


Paper presented at the American Educational Research Association
Annual Conference in New Orleans, Louisiana
April 2, 2002


This paper reports results from a study supported by a grant from the Joyce Foundation and the Wisconsin Center for Education Research. Any opinions, findings, and conclusions are those of the author and do not necessarily reflect those of the supporting agencies.

            This paper describes in detail the results of work in six Milwaukee Public Schools (MPS) on using data to inform decision-making. The work was the result of collaboration between a team from the Wisconsin Center of Education Research (WCER), headed by Norman Webb, and teams from each of the six schools (henceforth labeled QSP teams). The case studies represent a synthesis of data collected over the course of the two-year project.

Brief Review of the Literature on Using Data in Schools 

The Quality School Portfolio project was conceived to help schools adopt a school-based version of Deming's Total Quality Management (TQM) for use in school improvement planning and decision-making (Deming, 1986, cited in Schenkat (1993) ).  Schmoker, borrowing from Deming's work (Schmoker, 1996; Schmoker & Wilson, 1995) , argues that short-term, measurable results are the key to long-term improvement.  Schools, Schmoker argues, would improve if they focused on short-term, local data to guide the achievement of long-term goals. Elmore and Rothman (cited in Massell (2001) ) concur: "The theory of action of the basic standards-based reform model suggests that, armed with data on how students perform against standards, schools will make the instructional changes needed to improve performance."  Schmoker suggests a process of collaboration and brainstorming, combined with data collection and analysis based on student assessment, was responsible for numerous examples of schools increasing test scores, dramatically so in a number of his examples (Schmoker & Wilson, 1995) .  

            This vision of school improvement is not without its critics. Kohn (1993) describes a number of problems and inconsistencies experience by schools adopting a TQM approach.  One of the goals of TQM, according to Kohn, is that the improvement process be internally driven.  This characteristic conflicts with the reliance on standardized test scores as a measure of student achievement on the part of many school-based TQM proponents: most standardized tests are externally developed without regard to a particular school’s improvement process.   Kohn further argues that applying workplace terminology to schools is inappropriate because it places undue emphasis on how well students learn particular content rather than on what content students are learning. This can lead to a focus on rote skill development at the expense of broader learning goals. 

            Nevertheless, the national standards and accountability movements have pressured schools into using data in their school improvement plans (Massell, 2001) .  Massell describes efforts in eight states to use data to inform decision-making and notes the increased emphasis on the use of data at the state, local, and school levels over the last twenty years.  This, Massell argues, is the result of the standards and accountability movement, which has put pressure on schools to increase their test scores.  The states’ role has gone beyond emphasizing the use of data in school improvement planning to providing professional development to school personnel. Massell reports that these efforts have resulted in increased demand for data.  In some cases, local districts or schools have supplemented state efforts in providing assessment data. 

Fullan (Fullan, 1999; Fullan & Hargreaves, 1998) argues that school improvement models need to be coherent.  Fullan and Hargreaves (1998) suggest that schools need to develop assessment literacy, defined as: (1) teacher capacity to analyze student data and make sense of it; (2) develop school improvement plans based on data; and, (3) enter the debate about the uses and misuses of achievement data. Some local efforts to develop this literacy are discussed below. 

Herman and Gribbons (2001) and Nichols and Singer (2000) describe specific local efforts to get schools to use data. Nichols and Singer describe how the use of “data mentors,” plus other measures, have helped schools to interpret their data and increase their test scores. Herman and Gribbons describe their efforts in two school groups to use school data profiles; the profiles, which were portrayed in graphic form, provided “information-at-a-glance.”  Herman and Gribbons describe how one school used data to investigate some of the school’s inequities.  In another school, efforts to promote data use were met with resistance.  In these cases, the district kept longitudinal data and the analysis was not complex. The authors cite a need for data beyond standardized test data to understand and monitor students’ progress.  The non-performance indicators, such as attendance and dropout data, also proved problematic and Herman and Gribbons suggest improvements in the way they are calculated to be more useful.  For example, the district calculated attendance rates by looking at the average number of absences, while schools would be more interested in data on the percentage of days a student attends school. 

The issue on non-performance indicators is particularly relevant to the case studies discussed below.  The school system in which the schools are located tested only 4th, 8th, and 10th grades on an annual basis.  This limited assessment data pushed the schools to consider other types of indicators to measure improvement. Such data use contrasts with the literature discussed above, which focused primarily on test scores as forms of measurement. 

The literature shows an increased use of data in many local districts and schools.  The anecdotal evidence provides a number of examples of small-scale successes in solving particular school problems and in raising test scores.  This study contributes to this small but growing body of literature on efforts by school staffs to use data to make decisions. 

Major Findings 

            This section will report findings from the six MPS case study schools in terms of changes in data practices as a result of the QSP project. The Phase 1 schools, Forrester Middle School and Garden Heights Middle School[1], began their involvement in the QSP project in January, 2000. The four Phase 2 schools, two middle and two high schools began their involvement in August, 2000. We found clear distinctions in the level of success in using data for decision-making between the Phase 1 schools and the Phase 2 schools. The discussion below will highlight those differences and describe the most prominent trends.  

            The greatest differences that distinguished the efforts of the Phase 1 schools were: 1)  the creation of reports at regular intervals on a variety of indicators; and, 2) the creation of in-house databases to capture data normally sent to the district, or to generate data at a finer grain-size than required by the district. The indicators in the reports, mainly discipline-referral information, were aggregated at the school team level (called Families at Garden Heights and Teams at Forrester). The data necessary to measure the indicators were created by school-run databases; these databases required dedicated data entry.  

Garden Heights and Forrester were able to incorporate data as a regular part of their decision making process. Garden Heights’s efforts were tightly focused on the principal’s attempts to inform the school teams about their rates of discipline referrals and resource usage. This was designed in part to put some pressure on the teams to be more proactive and take more responsibility for student behavior. The principal also used this data to justify a hiring decision. The six-week reports created by the Forrester principal also centered on discipline issues, although there was a greater effort to include academic data and to link attendance, discipline, and classroom-level academic data to district-level assessment scores. On an annual basis, the Forrester QSP team used demographic and assessment data from the incoming sixth graders. The Forrester team’s efforts to use data appeared ready at the end of out two-year study to expand to additional  staff members and to affect a greater number of instructional decisions. Forrester’s principal also used the results of his data analysis to inform hiring decisions. Both principals felt that the use of data would produce greater accountability among their staffs.   

Only Forrester personnel, however, appeared ready to alter their current decision making processes as a result of the QSP project. The principal made attempts to expand the number of staff involved in using data; this included training additional staff on the QSP software and on data analysis. He also held a data workshop for about twenty teachers that coincided with the technical assistance visit from the WCER research staff.  Garden Heights’ data use was expanded, but the school’s decision making process did not appear to be greatly affected.  In a June, 2001, focus group, several staff members reported that the principal still controlled most of the data and the way it was used.  They expressed the opinion that the same group of people was still responsible for making decisions in the same ways as before.  

The four Phase 2 schools were unable to establish any regular use of data to inform their decision making during the course of the project. However, these teams continued to express an interest in using data, but needed greater access to reliable data and increased technical capacities. An additional consideration for the Phase 2 group was the inclusion of two high schools. The data management system for the two high schools proved more problematic than the system at the middle school level. These high schools reported access and accuracy problems. Furthermore, the two high schools were structured by departments according to academic discipline, while the four middle schools had school teams that were multidisciplinary.  

The obstacles faced by all six teams included a dearth of data, issues of technical capacity, and lack of personnel resources. The lack of data available to schools was affected by both the district’s data management policies, as well as its assessment practices. The district is currently developing a plan to increase testing so that most grades will be assessed every year; this will increase the quantity of assessment data the schools will receive. MPS policies regarding data flow back to schools were limited as a result of the current transition the district information system infrastructure is undergoing.  The transition of the system has created barriers to data collection and retrieval.  The completion of the district data warehouse and query tools should alleviate many of the barriers to data access experienced by the MPS schools in the 2000-2001 school year. 

The lack of school-level technical capacity remains an enduring problem. It became clear to us that these capacity problems crossed technical, analytical, and organizational areas in each of the six schools.  Despite the WCER research staff’s efforts, school staff members still displayed a lack of confidence and ability to create functional databases and to analyze the data they were able to access.  The capacity issue is aggravated by the need to master and integrate several software packages and data systems.  Furthermore, the ability of the school teams to generate appropriate questions and draw proper conclusions will require more training. Each school also faced problems of data management.  There was only so much time and few available personnel to dedicate to the task of data entry and analysis.  This was why the principal at Garden Heights insisted that this process would only work if there were a seamless way to incorporate the data entry into the appropriate software packages.  

In summary, the two schools (the two Phase 1schools) that successfully incorporated data had: 1) strong leadership from the principal; 2) invested considerable time and effort; and, (3) had two years of involvement in the project. The most intriguing of the four Phase 2 schools, Norris, did not appear to be following the pattern of strong principal leadership, but relied more heavily on technology personnel to provide the impetus for using data. 


            In January, 2000, QSP research staff interviewed members of the two Phase 1 school teams. A total of nine team members were interviewed regarding:  the nature of the school staff’s decision-making processes, the staff’s use of data; criteria for school success; and the staff’s most pressing concerns. In addition, staff members at both schools were given a survey that inquired about the same topics. Fifty-three surveys were returned. The results of these interviews and surveys were compiled and analyzed to determine baseline data on the two schools.   

            Technical assistance and training were provided for the Spring, 2000, semester at the two schools. A total of four training days were allotted for each school; the training focused on decision-making processes, software operation, and data analysis. Technical assistance was provided by telephone and by visits from the Milwaukee liaison. In addition to school-level assistance, WCER staff members were also involved at the district level, partly in an effort to gain access to data for the schools. 

            In August, 2000, WCER research staff interviewed members of the four Phase 2 schools. The protocol used during the first round of interviews was used for these interviews. In addition, the four school staffs were given the same survey that was administered to the original two staffs. A total of 33 interviews were conducted and 44 surveys were completed.  These interviews and surveys comprised the baseline data for the four new schools.   

            The Phase 2 schools were allotted three days of training, which focused on decision-making processes, QSP software operation, and data analysis. Technical assistance was provided throughout the school year by telephone and e-mail with WCER research staff and by visits by one of the WCER research staff members. In addition, three of the schools had day-long technical assistance visits in May and June to further facilitate the teams’ efforts. In May and June, 2001, the WCER research staff conducted focus groups at each of the six school sites. These focus groups were queried regarding their efforts to use data and to assess the impact of the QSP project.   

Research Questions 

            The case study data were analyzed to obtain answers to the project’s four research questions:  

1.      What are the data needs of schools?

2.      How can the quality and flow of data to schools be improved?

3.      What level of data analysis is useful to schools?

4.      How can schools use data effectively to meet their needs? 

What are the data needs of schools? 

            All six of the school QSP teams expressed a desire for rapid access to a wide variety of data in convenient formats. In general, they wanted data that would allow them to track their students’ academic and behavioral progress on a regular basis. The majority of the QSP teams wanted both historical data on the students, as well as data generated throughout the school year, such as current attendance, discipline, and academic data.  Current academic data would include grades, aggregated at both the student and teacher levels, and school-administered assessments. Several teams explicitly commented on the need to integrate data collection and analysis into regular school routines. All of the teams commented on their lack of access to this wide variety of data.  

            The school teams wanted access to academic data, such as grades and standardized test results, and data on attendance and discipline. The teams felt it was important to monitor non-academic indicators, in part because these data were readily available, but also because the teams felt that a causal link existed between these factors and learning.  

            The teams wanted access to historical and current data. Historical data refers to data from previous years and could include grades, standardized test scores, and Special Education data. Current data refers to data collected throughout the year, such as attendance, discipline and current school-year grades.  

            The teams expressed a need to receive data that were already in a format convenient for entry into a database or computer program (like QSP) for analysis. The district’s School Management System (SMS) has that potential, but at the time of our study it lacked a functioning export feature, and it lacks the range of reporting features of QSP.   

            The case studies provide examples of the ways individual school teams varied in their expression of data needs. Teams at the two Phase 1 schools were more sophisticated in the way they expressed their needs; they were interested both in establishing baseline data and in measuring indicators at regular intervals to monitor progress. The two Phase 2 high schools had great difficulty obtaining trustworthy data of any sort; this situation constrained the efforts of team members to get beyond data access issues to the utilization of data for decision-making. At the two Phase 2 middle schools, the staff members of one demonstrated increasingly sophisticated notions of how they might be able to collect and utilize data to monitor their students’ progress.   

How can the quality and flow of data to schools be improved? 

            All of the school teams experienced difficulty in gaining access to data maintained by the district. The data the schools did receive were not formatted conveniently for school use, were not electronically stored, were at an inappropriate level of aggregation, and were not linked to other essential data.  Some data, such as attendance records, were sent by the district to schools in large stacks of paper. Schools had to enter these data into their computers in order to do analyses. Data were either aggregated only at the school level or available only at the individual level. Compiling data on one student might require accessing several different database systems, and, often these data were of questionable accuracy. The database in use at the high school level, the School Management System (SMS), had features such as the ability to aggregate data above the student level or to report attendance accurately; these features often did not work properly or were not exportable to, or compatible with, other database systems in place at the school.  

            The accuracy of data was cited by all of the QSP teams as a problem. Attendance data from the SMS at the high school level were viewed as particularly problematic, but district data on Special Education, state assessments, and grades were also questioned by both the middle and high schools.

             Nevertheless, the school teams expressed a desire for greater access to the historical and operational data maintained by the district. A number of team members expressed frustration with the district’s reluctance to provide schools with data in an electronic format. In late Spring of 2001, the district began training school-level staff on Brio software, which promises to facilitate access to the district’s data warehouse.  

What level of data analysis is useful to schools? 

The level of analysis can be expressed along several dimensions: temporal, grain size, and statistical analysis. The temporal dimension refers to the time intervals used by schools in their regular collection and analysis of data; for example, the interval could be six weeks or it could be a year, depending on the availability and use of the data.  Watson, referring to the data rather than to its collection, calls this characteristic temporal resolution (Watson, 2001). For example, a school might collect and aggregate discipline referral data every week. This datum’s temporal resolution is weekly.  The grain size (Watson, 2001) refers to the size of the group a particular datum is describing, or to the number of items in an assessment that a particular score represents. For example, a school team might want to look at the school’s median test score on a mathematics assessment; or the members might decide to look at the scores of the 6th graders from a particular feeder school. Statistical analysis is represented by the type of charts and the techniques schools employ to analyze data. A simple descriptive analysis might include pie charts and histograms; a medium level of sophistication would include statistics that describe distributions, such as measures of central tendency and variance.  A more sophisticated analysis would include some inferential statistics. 

            School teams were interested in looking at data at regular intervals, although only the two Phase 1 schools were able to accomplish this. The Phase 2 schools expressed an interest in generating data that would regularly measure the progress of students along a variety of indicators, but did not have the capacity, or access to appropriate data, to accomplish this. The school teams did look at data generated on an annual basis, such as state or district standardized assessments. These data were frequently used in the school’s Educational Plans, the district-mandated school improvement plans. 

            All four middle school teams were interested in looking at data on groups of students. The most common way the school teams wanted to group the data was according to internal school teams of teachers and students. For example, one principal wanted to compare rates of discipline referrals among the various teams (the school-level division of teachers and students)  at his school. Another example involved looking at 8th graders who needed to pass the MPS Middle School Proficiencies.  The school teams analyzed data aggregated at the school level or grade level as well. Examples include looking at the students’ performance on a standardized assessment, or at the distribution of mathematics scores of an incoming 6th grade cohort.  

            The school teams mostly relied on QSP to produce simple descriptive analyses, such as frequency tables, pie charts, and histograms. In many cases, they had never seen these types of visual representations of characteristics of their students used. The two Phase I teams occasionally utilized QSP to produce the distribution of means to compare cohorts of students, but otherwise the schools did not engage in analysis much beyond simple description.  

How can schools use data effectively to meet their needs? 

            The two Phase 1 schools were most effective in using data to inform their decision-making. Both of these schools prepared reports that were disseminated to their internal school teams at regular intervals. The reports differed, but they both contained data regarding the number of discipline referrals. The Garden Heights report did this in a more detailed way and also included data on resource usage of each internal team. The Forrester report contained information about grades and attendance, in addition to the discipline data. It is important to note that these schools focused on these data in part because they were the most accessible and because both schools felt that discipline was an important indicator to monitor. 

            The Forrester QSP team tended to use data to inform a broader set of decisions than did Garden Heights. Members used data as a basis for resource decisions, proportional distribution of students to school teams, school policy, and to attempt to align grading policies with achievement results.  This ambitious agenda reflected the principal’s strong long-term interest in data use. 

The teams at the four Phase 2 schools provided a number of examples of how they used data to inform their decision-making. These examples can be analyzed in terms of: 1) who was using the data; 2) what data were used; and 3) what kinds of decisions were made as a result of using data. The examples were wide-ranging on all three criteria. 

The three groups of people from the Phase 2 schools who reported using data were teachers, administrators, and support personnel (counselors, Special Education teachers, curriculum specialists). For example, an administrator at one of the high schools used attendance and assessment data to alter the 9th-grade schedule. A counselor at one of the middle schools used attendance and discipline referral data to select students into a program to help 8th graders pass the MPS Middle School Proficiencies. 

The data mentioned were reading scores from a variety of assessments, grades, pass/fail rates, Honors Level discipline data, attendance rates, MPS Proficiency scores, and enrollment rates.  For example, pass/fail rates at one of the high schools were used to abandon the 9th grade family team structure. In another school, low mathematics scores on a statewide assessment motivated the school to switch to block scheduling. 

Team members made decisions that influenced instruction, scheduling, course offerings, school structure, student privileges, team performance, resource allocation, and program enrollment.  For example, one school used a software program to track students’ discipline referrals; students with positive ratings were granted privileges for extracurricular activities.  At another school, low reading scores on an internally administered assessment were used by the librarian as a basis for purchasing more appropriate reading books. 

            The difference between the Phase 1 schools and the Phase 2 schools was that the examples of data use in the later schools seemed to be isolated and uncoordinated. The two Phase 1 schools had principals who were strong advocates of a data-driven approach and this was reflected in the more integrated manner in which they incorporated data into their decision making process. 

Focus Group Results 

Towards the end of the two years of the project, each QSP school team participated in a focus group. The purpose of these focus groups was to assess the results of the efforts of the school teams to use data to inform their decision-making. Some of the comments have been summarized above. Other comments respond in a tangential but relevant way to the research questions. Some comments are also included below. 

Necessary Conditions 

            The QSP school teams expressed a variety of conditions that would facilitate the use of data to inform decisions. These conditions reflected both the resource commitment necessary for data collection and analysis and the cultural change necessary to implement such a process. 

            The resource requirements included the need for personnel who have both time and technical expertise. The required expertise was defined as the ability to enter and analyze the data, as well as the ability to interpret the results. This included the abilities to determine what questions to ask and what variables to compare. Another resource requirement involved finding convenient processes and software for collecting and analyzing data. 

The necessary cultural changes entail the need to establish a focus for data collection and analysis. This meant that data collection and analysis need to follow an established plan that had been discussed by all of the involved stakeholders. The other cultural change needed involves the creation of what one team member called “non-defensive atmosphere,” which would promote the use of data for school-wide improvement and thus make school staff more comfortable in using data.  

Change/Impact of the Data Process as a Result of the Project 

            The two Phase 1 school teams reported a more substantial impact from the QSP project than did the four Phase 2 teams. Phase 1 teams reported an expanded role for data use in their decision-making processes and in a school-wide awareness of the impact of the use of data at both sites. 

Forrester’s team reported a stronger focus in its data collection; it also reported that the use of data had helped to reduce the emotional aspect of important decisions and had helped to expand the number of staff members involved in decision-making. The Garden Heights’s team also reported increased use of data to inform decisions, but the attempts were less focused and created greater levels of anxiety among the staff than those at Forrester.  

The four Phase 2 school teams reported an increased awareness of the difficulties of collecting and analyzing data to help inform decision making. This awareness included the need for future efforts to be more focused and to include multiple sources of data to analyze a problem. Beyond that, their efforts were still in the preliminary stages at the end of the academic year. The four schools reported obstacles in the gathering of data, as well as in developing the technical capacity to use data efficiently.  

Future Data Uses and Processes 

            The data uses that the QSP school teams would like to implement in the future fall into two categories: individual student profiles and school team-level profiles.  

            Four of the six teams would ultimately like to create extensive electronic student profiles.  This would allow teachers to easily retrieve information regarding a student’s past academic performance, Special Education history, and discipline referral information.  One principal also expressed an interest in having students monitor their progress via electronic profiles.  Ideally, this information would be updated at regular intervals to help teachers track their students’ academic progress. 

            The two Phase 1 already provide their school teams with regular reports containing about half a dozen measures (which differ at the two sites).  Both of these QSP school teams would like to see the number of measures expanded to help determine the effectiveness of various interventions and teaching approaches.  In addition, two of the other QSP school teams also expressed an interest in creating school team-level indicators.  These would help the school teams track the progress of their students on important indicators, both behavioral and academic. 


Fullan, M. (1999). Change forces: The sequel. Philadelphia: Falmer Press.

Fullan, M., & Hargreaves, A. (1998). What's worth fighting for out there? New York: Teachers College Press.

Herman, J., & Gribbons, B. (2001). Lessons learned in using data to support school inquiry and continuous                 improvement: Final report to the Stuart Foundation . Los Angeles, CA: Center for the Study of Evaluation.

Kohn, A. (1993). Turning learning into a business: Concerns about total quality. Educational Leadership, 51(1), 58-61.

Massell, D. (2001). The theory and practice of using data to build capacity: State and local strategies and their effects.                 In S. H. Fuhrman (Ed.), From the capitol to the classroom:  Standards-based reform in the states.  One                 hundredth yearbook of the National Society for the Study of Education, Part II. Chicago: University of                 Chicago Press.

Nichols, B. W., & Singer, K. P. (2000). Developing data mentors. Educational Leadership, 57(5), 34-37.

Schenkat, R. (1993). Deming's quality: Our last but best hope. Educational Leadership, 51(1), 64-65.

Schmoker, M. (1996). Results: The key to continuous school improvement. Alexandria, VA: Association for                 Supervision and Curriculum Development.

Schmoker, M., & Wilson, R. B. (1995). Results: The key to renewal. Educational Leadership, 52(7), 62-64.

[1] The names of all schools mentioned in this paper are fictitious.

Information Staff Reports Links