Skip to content

Data Wise Improvement Process

Design Process · All design processes
Evidence · 3 claims (1 for, 1 mixed, 1 against) · 6 studies (2 quant-synthesis, 2 theoretical, 1 review, 1 qualitative), q1–q3 · 2 of 6 report an effect size · 1 claim rests on one study

Description

The Data Wise Improvement Process, developed at the Harvard Graduate School of Education, is a structured cycle through which school teams systematically examine student learning data, connect findings to instructional practice, and act on what they learn. It organizes collaborative data work into three phases — Prepare, Inquire, and Act — comprising eight steps, and is deliberately scaffolded by norms and protocols so that conversations about data stay focused on evidence rather than blame. The problem it solves is common in schools: assessment data are collected extensively but rarely translated into changed instruction, because teams lack a shared process for interpreting evidence and agreeing on next steps.

Implications

The process rests on the premise that data use is fundamentally a learning problem for adults, not a technical one: teachers need structured opportunities to build shared interpretations of evidence before they will change practice. Its protocols externalize and discipline group reasoning — much as think-aloud makes expert reasoning visible to novices — reducing the tendency for discussion to be dominated by status or anecdote. Research on data-informed decision making finds that structured, collaborative data use is associated with instructional change, but that effects depend heavily on leadership support, time, and data literacy Data-based decision-making improves instruction when teams have time, data literacy, and leadership support. [~M]

Context

Requirements

  • A school- or program-level team with protected, regular meeting time for data work
  • Valid, disaggregated assessment data available at a grain size that connects to instruction (item-level, strand-level, not just aggregate scores)
  • Assessment literacy: team members must understand what the measures do and do not show
  • A facilitator skilled in running protocols and keeping discussion evidence-based
  • Leadership that models inquiry and protects teams from blame-oriented data use

Constraints

  • Time-intensive; the full cycle typically takes a semester to a year, and under-resourced schools struggle to sustain it
  • Degrades into compliance-driven "data theater" when imposed top-down without teacher ownership [~M]
  • Only as good as the underlying assessments — teams can inquire deeply into weak or misaligned measures and draw misleading conclusions
  • Focused on instructional improvement at team/school level; not designed for individual student intervention planning or program evaluation
  • Requires psychological safety; in low-trust cultures, data discussions revert to defensiveness or avoidance

Grain Size

Program or school — the cycle operates at the level of instructional teams, departments, or whole faculties over months, not individual lessons.

Target Goals

  • Improved instructional quality through evidence-based identification of student learning problems
  • Formative use of assessment — shifting data use from accountability reporting to instructional decision making (Assessment for Learning) Assessment for learning improves achievement. [+S]
  • Collective teacher efficacy and shared instructional ownership
  • A durable culture of collaborative inquiry rather than one-off data reviews

Target Learners

  • Practicing teachers, instructional coaches, and school leaders engaged in professional learning
  • Teams in K–12 schools (the original context), adapted in higher education for program assessment and in teacher preparation as a case-based learning experience
  • Adults, not students — the "learners" in this pattern are educators studying their own practice

Theory

Supporting

  • Situated Learning (Lave & Wenger) — teachers learn data-informed practice by participating in an authentic professional community of inquiry, not by being trained in the abstract
  • Constructivism — the Inquire phase requires teams to build their own interpretations of evidence rather than receive conclusions from administrators
  • Social Learning Theory (Bandura) — norms, modeling, and collective efficacy shape whether teams sustain the work

Contradicting / Qualifying

  • Behaviorism — accountability regimes that treat data as incentive signals (rewards/sanctions) undermine the inquiry stance Data Wise requires; the same data can support either approach, but not both simultaneously

Claims

Supporting

  • Assessment for learning improves achievement. [+S] — the process operationalizes formative assessment at the institutional level
  • Structured collaborative data use is associated with instructional improvement when time, data literacy, and leadership conditions are met [~M] — see Schildkamp (2019) meta-review

Contradicting

  • None identified — though evidence for data-use interventions generally is more consistent on teacher learning than on measured student outcomes [~W]

Design

Sequence

The eight steps, grouped in three phases:

Prepare 1. Organize for collaborative work — form teams, establish norms, build trust (check-in routines and agreed discussion protocols) 2. Build assessment literacy — develop shared understanding of what each measure reveals and what it cannot 3. Create a data overview — the team compiles and disaggregates assessment results into a shared picture of student learning

Inquire 4. Dig into student data — use protocols to examine learner work and item-level results, identifying specific learning problems (class discussion structured by evidence rules) 5. Examine instruction — look for instructional practices that could explain the observed learning problems, without jumping to solutions 6. Develop an action plan — agree on concrete instructional changes, supports, and owners (action planning)

Act 7. Plan to assess progress — decide in advance what evidence will indicate whether the action plan is working (assessment) 8. Act and assess — implement the plan, then return to data to evaluate impact and cycle back

Affordances

  • Assessment for Learning — the entire cycle reorients assessment from summative judgment to evidence for improving instruction, with the "plan to assess" step building measurement of impact into every action plan
  • Collaborative Learning — the process is designed for teams, not individuals; protocols distribute airtime and force shared sense-making rather than individual interpretation
  • Purposeful Reflection — the Inquire phase institutionalizes structured reflection on the relationship between teaching and learning before action is taken

Personalization

Teams new to data work: Begin with a small, well-understood dataset (one assessment, one class period) and a tightly scripted protocol; build assessment literacy before attempting a full data overview.

Experienced teams: Compress Prepare steps and spend more time in the Inquire phase examining student work rather than test scores; rotate facilitation to build internal capacity.

Low-trust or high-accountability contexts: Leaders should model vulnerability by examining their own practice first, and explicitly separate Data Wise conversations from evaluation conversations; without this separation, participation becomes performative.

Higher education adaptation: Use course-level learning analytics and program assessment data in place of K–12 test data; the "examine instruction" step requires curricular rather than pedagogical framing.

  • Collaborative Inquiry — Data Wise is a highly specified instance of collaborative inquiry; the broader pattern applies when teams lack the full Data Wise infrastructure
  • Constructive Alignment — the "examine instruction" and "plan to assess" steps push teams toward aligning assessments, instruction, and intended learning outcomes
  • Authentic Assessment — data overviews are only as trustworthy as the assessments feeding them; authentic tasks often yield the richest evidence for the Inquire phase

Examples

K–12 school improvement — Boston Public Schools: Data Wise originated in work with Boston schools, where instructional teams used the eight-step cycle to move from aggregate test-score reviews to item-level analysis and shared instructional action plans (Boudett, City, & Murnane, 2013).

School networks — Data Wise Project, Harvard GSE: The project publishes protocols, rubrics, and case studies used by districts and networks internationally; schools typically run the full cycle annually, with team-level mini-cycles each term.

Teacher education: Preparation programs use Data Wise as a case-based learning experience, having candidates analyze a school's data overview and draft an action plan before entering the classroom — see Case-Based Learning.

Higher education program assessment: Departments have adapted the cycle to review capstone assessment results, examine curriculum, and plan revisions, replacing compliance-oriented accreditation reporting with inquiry.

Key Sources

  • Boudett, K. P., City, E. A., & Murnane, R. J. (Eds.). (2013). Data wise: A step-by-step guide to using assessment results to improve teaching and learning (Rev. ed.). Harvard Education Press.
  • Boudett, K. P., & City, E. A. (Eds.). (2014). Meeting wise: Making the most of collaborative time for educators. Harvard Education Press.
  • Datnow, A., & Park, V. (2014). Data-driven leadership. Jossey-Bass.
  • Schildkamp, K. (2019). Data-based decision-making for school improvement: Research insights and gaps. Studies in Educational Evaluation, 60, 76–90. doi:10.1080/00131881.2019.1625716
Page metadata
Field Value
type process
id data-wise-improvement-process
title Data Wise Improvement Process
description The Data Wise Improvement Process is an eight-step, team-based protocol for using student assessment data to drive instructional improvement in schools.
status review
grain_size program, school
author Boudett, City, & Murnane (Harvard Graduate School of Education)
generated by claude/unspecified on 2026-08-30
trust unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.