eduKateSG Learning Node Series · 0082

How Research–Practice Partnerships Work | Build a Two-Way Street Between Research and Educational Practice

Education has a familiar coordination problem. Researchers can produce careful studies that arrive after the decision window has closed. Schools and systems can make urgent decisions using local experience without enough comparative evidence. Both groups may be intelligent, serious and committed—and still fail to learn from one another.

A research–practice partnership changes the relationship. Researchers do not merely arrive, collect data and leave. Practitioners do not merely receive a finished report and decide whether to use it. The two sides build a longer-lived collaboration around problems that matter in practice, create a shared agenda, negotiate what counts as useful evidence, interpret findings together and preserve enough trust that inconvenient results can still travel across the boundary.

The partnership is not a delivery pipe from research to practice. It is a joint learning system in which the problem, evidence and decisions move in both directions.

The 50-Second Read

• A research–practice partnership, or RPP, is a long-term collaboration between researchers and practitioners organised around educational improvement.
• Strong RPPs begin with consequential problems of practice rather than with a researcher’s preselected study topic.
• The relationship is deliberately mutual: practitioners contribute contextual, operational and professional knowledge; researchers contribute methodological, theoretical and comparative knowledge.
• Joint agenda-setting matters because a rigorous answer to an irrelevant question has little practical value.
• Shared authority does not mean everyone performs the same role. It means important decisions about questions, interpretation and use are not owned by only one side.
• Data agreements, governance rules, meeting routines, timelines and boundary-spanning roles are infrastructure, not administrative decoration.
• Evidence use should happen throughout the partnership, not only after a final report.
• Long-term relationships can support cumulative learning across multiple studies instead of restarting trust and context with every project.
• The William T. Grant Foundation describes RPPs as long-term, mutually beneficial collaborations that promote the production and use of rigorous research around problems of practice.
• A major 2016 review by Coburn and Penuel characterised RPPs by long-term collaboration, problems of practice, mutualism, intentional routines and original data analysis, while emphasising that the evidence base was still developing.
• Current RPP frameworks increasingly stress equity and the deliberate redistribution of whose knowledge and priorities count.
• The relationship itself is a form of research infrastructure: when it is weak, even excellent evidence may not reach the decision that needs it.

Canonical Owner Boundary

This page owns the research–practice partnership mechanism: a sustained cross-boundary relationship in which research organisations and education practitioners jointly frame problems, produce or interpret evidence and use that evidence for improvement. How Education Works | Educational Research remains the umbrella owner for the broader research enterprise. How Teacher Inquiry Works owns practitioner-led inquiry close to the classroom. How Professional Learning Communities Work owns recurring educator communities inside professional practice. The next Learning Node on networked improvement communities owns multi-site improvement networks. An RPP can contain all of these forms, but its distinctive mechanism is the durable bridge between research and practice organisations.

1. The Research–Practice Gap Is Often a Relationship Gap

It is tempting to describe educational research as a knowledge-transfer problem: researchers know something, practitioners do not, so the solution is better dissemination. Sometimes that is true. Often the problem begins earlier.

The research question may not match the decision practitioners face. The measure may not represent the outcome they care about. The study may finish after budgets, staffing or curriculum choices have already been made. The report may use categories that do not map cleanly onto local systems. Practitioners may possess contextual knowledge that never entered the research design.

An RPP tries to move the connection upstream, before the question, data and interpretation are fixed.

2. Start With a Problem of Practice

A problem of practice is not simply a topic such as attendance, mathematics achievement or teacher retention. It is a problem that practitioners are responsible for acting on and that evidence could help them understand or improve.

“Attendance is falling” is a signal. “Which transition points, transport constraints, school responses and family circumstances explain the rise in repeated absence among a particular group, and which responses are feasible within the district’s current system?” is closer to an actionable problem of practice.

The problem gives the partnership a reason to exist beyond producing another publication.

3. Long-Term Relationships Change the Questions You Can Ask

A one-year study can answer some questions well. It struggles with others: implementation across leadership turnover, long-run teacher learning, changes in organisational routines, or whether an intervention keeps working after initial support fades.

Long-term partnerships accumulate context. Researchers learn how data are produced, which policy cycles matter, where school autonomy begins and ends, and why apparently simple operational changes are not simple. Practitioners learn what different designs can establish, what uncertainty remains and why some claims require slower evidence.

Trust compounds, but so does shared technical language.

4. A Shared Agenda Is Not a Compromise List

Weak partnerships often build an agenda by letting each side add its favourite question. The result is a long list with little cumulative logic.

A stronger agenda identifies a small set of connected uncertainties around a real improvement problem. One project may map variation, another study mechanisms, a third test an implementation change and a fourth examine long-term outcomes. The sequence builds knowledge rather than merely filling a calendar.

5. Mutual Benefit Must Be Designed

Researchers may gain publications, data access, grants and theoretically important cases. Practitioners may gain analysis, evidence, tools, decision support and a clearer model of the system. Those benefits do not automatically align.

A partnership should make the exchange explicit. If one side bears the workload while the other captures most of the recognised value, the relationship will eventually become extractive.

6. Shared Authority Does Not Mean Identical Jobs

Researchers are not school principals. Principals are not statisticians. Community organisations are not merely recruitment channels. Students and families are not data points with opinions attached.

Shared authority means each form of expertise can shape consequential decisions: what problem deserves attention, how it is defined, which outcomes matter, what interpretations are plausible and how findings are used. Roles remain differentiated while voice becomes more balanced.

7. Power Must Be Made Visible

A partnership can call itself collaborative while one organisation controls the funding, data, publications, meeting agenda and final interpretation. Formal equality on a steering committee does not neutralise those asymmetries.

Modern RPP frameworks therefore pay increasing attention to power: who can say no, whose timetable dominates, whose knowledge is cited, who owns products, who decides when evidence is sufficient and who receives credit.

Equity is not achieved by adding more chairs to the room if the decision rules remain unchanged.

8. Governance Turns Good Intentions Into Repeatable Behaviour

Partnerships need routines for choosing projects, resolving disagreements, approving data use, reviewing findings, communicating uncertainty and deciding which products are public.

Without governance, every difficult moment becomes a fresh negotiation. With too much governance, the collaboration becomes a bureaucracy. The goal is enough structure that trust does not have to carry every decision by itself.

9. Data Agreements Are Trust Architecture

Educational data can be sensitive, messy and politically consequential. Before analysis begins, partners need clarity about access, security, linkage, retention, disclosure, publication, re-use and the rights of students, families and staff.

A data agreement is not only a legal shield. It expresses the partnership’s model of responsibility. It tells participants whether information can travel safely without turning local vulnerability into external risk.

10. Data Quality Is Often an Organisational Finding

Researchers may discover that attendance codes vary across schools, intervention participation is inconsistently recorded or a supposedly stable category changed definition midway through the period.

That is not merely a nuisance before the “real” study. It can reveal how the system itself works. If a district cannot reliably tell which students received an intervention, implementation monitoring may be part of the improvement problem.

11. The Partnership Must Respect Decision Time

A mathematically elegant analysis delivered three months after a budget decision may have little operational value. Conversely, a same-day dashboard may be too weak to support a high-stakes policy choice.

RPPs can design evidence products around decision windows: rapid descriptive analyses for immediate planning, deeper causal work for longer-horizon policy, implementation studies during rollout and synthesis when strategic direction is reconsidered.

Rigor and timeliness are not enemies, but they must be negotiated rather than assumed.

12. Rigor and Relevance Are Different Axes

A relevant question can be studied badly. A rigorous study can answer a question nobody needs. Strong partnerships refuse the false choice.

Methodological rigor concerns whether the evidence supports the claim. Practical relevance concerns whether the claim can improve understanding or decision-making in the system. Both can be designed for, and both can fail independently.

13. Local Context Is Not Noise to Be Removed

Research often seeks findings that travel. Practice happens in particular places. The partnership must learn which parts of context matter to the mechanism.

If a programme works only where teachers receive common planning time, that condition is not an embarrassing footnote. It may be part of the causal architecture. Understanding context helps evidence travel more responsibly.

14. Boundary Spanners Keep Two Worlds Connected

Some people become unusually valuable because they understand both research and practice. They can translate a policy problem into a researchable question, explain methodological constraints without condescension, notice when researchers misunderstand an operational rule and help practitioners see when a local pattern has broader theoretical significance.

These boundary spanners may be research directors, data leads, former educators in universities, practitioner-researchers or dedicated partnership staff. Their work is coordination-intensive and easy to underestimate because much of it appears as conversation rather than a finished product.

15. Co-Interpretation Is Where Context Meets Method

Suppose an analysis shows that one group of schools improved faster. Researchers can test statistical explanations. Practitioners may know that those schools changed timetables, staffing or assessment routines at the same time.

Joint interpretation does not permit anecdote to overrule evidence. It helps identify plausible mechanisms, confounders and follow-up analyses that neither side would see alone.

16. Research Use Happens Before the Final Report

Traditional dissemination imagines a completed study moving into practice. RPPs can create use much earlier. Framing the question may reveal unclear goals. Building the dataset may improve operational definitions. Reviewing preliminary patterns may change implementation. Developing measures may force the system to state what success actually means.

The research process itself can become organisational learning.

17. Multiple Methods Belong in the Same Partnership

Different questions need different evidence. Administrative data can reveal patterns at scale. Interviews can explain how participants experience a policy. Observation can show enacted practice. Experiments or quasi-experiments can strengthen causal claims under appropriate conditions. Design studies can help build and refine new tools.

An RPP should not be loyal to one method. It should be loyal to the quality of inference needed for the decision.

18. Causal Questions Need Special Discipline

Schools change many things at once. Students select into programmes. Leaders target support at those already struggling. Context shifts. Simple before-and-after comparisons can therefore mislead.

When the partnership asks whether an intervention caused an outcome, design matters. Randomisation may sometimes be feasible; natural experiments, discontinuities, matching or other quasi-experimental designs may sometimes help. At other times, the honest answer is that the available evidence is descriptive.

A trusted partnership should make uncertainty easier to say, not harder.

19. Qualitative Evidence Can Explain Why the Number Moved

A district may observe improved mathematics scores after a curriculum change. That result matters. It does not tell leaders whether teachers used the intended routines, whether some components were ignored, how students experienced the tasks or which local adaptations helped.

Mixed methods let the partnership connect outcome, mechanism and implementation rather than treating one number as the whole story.

20. Community Knowledge Should Enter Before Findings Are Final

Families and students often experience system effects invisible to central offices. A transport policy that appears neutral in administrative data may create different burdens by neighbourhood. A digital learning requirement may assume device access, quiet space or language support that is unevenly distributed.

Community participation is strongest when it shapes the problem and interpretation, not when people are invited at the end to react to a finished slide deck.

21. Credit and Authorship Reveal the Partnership’s Values

Who is named on reports? Who presents findings? Whose intellectual contribution is recognised? If practitioners generate the problem, help interpret evidence and redesign the system but appear only in acknowledgements, the formal record may misrepresent where the knowledge came from.

Credit practices should reflect contribution rather than institutional prestige.

22. Publication Incentives Can Pull Away From Practice

Academic careers reward novelty, peer-reviewed publication and disciplinary contribution. Practice organisations reward timely decisions, reliable operations and measurable improvement. Those incentive systems overlap imperfectly.

A mature partnership designs a portfolio: some work becomes scholarly publication, some becomes technical guidance, some produces internal decision tools, and some improves data or routines without ever becoming a journal article.

23. Research Products Need Different Speeds and Formats

• A two-page decision memo may serve a policy meeting.
• An interactive data tool may support school teams.
• A technical appendix may preserve methodological transparency.
• A professional-learning resource may translate a finding into classroom use.
• A peer-reviewed paper may test and expose the work to scholarly criticism.

The evidence is not diluted by translation if claims remain accurate. Different users need different interfaces to the same underlying reasoning.

24. Negative Findings Test the Relationship

A programme championed by senior leaders may show little benefit. A highly regarded school may have weaker implementation than expected. A researcher’s preferred theory may not fit local evidence.

The moment inconvenient findings appear is when the partnership discovers whether trust is real. Can results be checked rigorously without being suppressed? Can uncertainty be communicated without humiliation? Can leaders change course without treating evidence as disloyalty?

25. Turnover Is a Scientific Problem Too

Partnerships often depend on a few people who know the history, data quirks and unwritten agreements. When they leave, years of relationship capital can disappear.

Institutionalise memory: document decisions, maintain shared definitions, onboard new participants, preserve project histories and build more than one relationship across organisational levels.

26. Cross-Domain Comparison: Translational Medicine

Health research has long confronted the distance between laboratory discovery, clinical evidence and routine care. Translational systems create interfaces among scientists, clinicians, patients, regulators and delivery organisations so that questions and evidence can move across stages.

Education is not medicine, but the coordination lesson transfers. Evidence becomes useful only when the people who produce knowledge and the people who operate the system can repeatedly align questions, conditions and implementation.

27. Cross-Domain Comparison: Flight-Test Engineering

An aircraft designer cannot learn everything from a model. Test pilots and engineers create a feedback system in which theoretical expectations meet operating reality. The pilot does not replace aerodynamics; the engineer does not replace the pilot’s direct experience of the machine.

RPPs work similarly when practitioner experience becomes structured evidence and research helps distinguish local impression from more general mechanism.

28. Cross-Domain Comparison: Intelligence Fusion

Complex intelligence problems require different sources with different strengths: signals, imagery, human reports, historical patterns and local expertise. No single stream tells the whole story. Integration requires provenance, confidence judgments and disciplined disagreement.

An education partnership also combines heterogeneous evidence. Administrative data may show where a problem concentrates; teachers explain instructional conditions; families reveal access barriers; researchers test alternative explanations. The gain comes from structured fusion, not from averaging every opinion into consensus.

29. Example: Chronic Absence

A district sees rising chronic absence and initially plans a generic attendance campaign. An RPP begins by disaggregating timing, school, age, transport, health, family circumstances and prior attendance patterns. Interviews show that some absences cluster after repeated minor disciplinary removals; in other areas, transport unreliability is more important.

The partnership therefore avoids treating absence as one mechanism. It helps the district test different responses by subgroup and location, while tracking whether interventions reduce absence rather than merely improve documentation.

30. Example: A New Mathematics Curriculum

A school system adopts a curriculum because prior trials were promising. The RPP does not simply rerun the efficacy question. It asks how implementation varies, which supports teachers use, where pacing breaks, how student reasoning changes and which adaptations preserve the instructional mechanism.

Quantitative outcomes, classroom observations, teacher interviews and implementation records are interpreted together. The result can improve local delivery while also producing knowledge about the conditions under which the curriculum travels.

31. Example: Artificial Intelligence in Student Writing

A system wants guidance on generative AI but the technology and student behaviour are changing faster than a conventional multi-year study cycle. The partnership builds a staged agenda: descriptive work on current use, rapid studies of specific feedback workflows, qualitative analysis of teacher decision-making, and longer-term examination of writing independence and assessment integrity.

The partnership does not ask one timeless question—“Does AI work?” It builds an updating evidence system around changing tools, practices and risks.

32. What the Major RPP Frameworks Emphasise

The William T. Grant Foundation describes research–practice partnerships as long-term, mutually beneficial collaborations that promote the production and use of rigorous research around problems of practice. Its later field work emphasises five broad principles: building trusting relationships, conducting rigorous research that informs action, supporting partners’ ability to use research, producing knowledge that can inform broader improvement and intentionally addressing power and equity.

Coburn and Penuel’s influential 2016 review highlighted long-term collaboration, focus on problems of practice, mutualism, intentional strategies for partnership and original analysis. Importantly, the authors also argued that RPPs themselves needed stronger study. A promising organisational form should not be exempt from evidence about whether and how it works.

Current Spencer Foundation programmes likewise treat RPPs broadly, including partnerships that use participatory approaches, design research and networked improvement. The common thread is not one research method. It is collaborative knowledge-building organised around improvement.

33. Failure Mode: The Partnership Is Really a Consultancy

A school commissions an analysis, researchers deliver recommendations and the relationship ends. That may be useful consultancy. It is not the same as a long-term RPP with shared agenda-setting and cumulative learning.

34. Failure Mode: Academic Capture

The partnership slowly reorganises around publishable questions. Practitioner meetings become recruitment opportunities. Timelines stretch to match academic incentives. Local priorities survive only when they fit a scholarly paper.

The remedy is governance that keeps the problem of practice visible and protects multiple forms of useful output.

35. Failure Mode: Practice Capture

The opposite failure also occurs. Leaders want research mainly to legitimise a decision already made. Methods that could reveal an inconvenient answer are discouraged. Findings are filtered until they become advocacy.

A partnership that cannot produce unwelcome evidence has lost the research half of its identity.

36. Failure Mode: Extractive Data Relationships

Researchers receive years of school data, produce papers and return little usable knowledge. Practitioners become data suppliers rather than partners.

Reciprocity should be concrete: useful analyses, transparent methods, capacity-building, shared interpretation, usable tools and recognition of practitioner intellectual contribution.

37. Failure Mode: Stakeholder Theatre

Students, families or teachers are invited to one workshop after the research agenda has already been decided. Their presence appears in photographs and reports but does not change a question, measure or interpretation.

Participation becomes real when it can alter the work.

38. Failure Mode: Endless Relationship, No Decision

Partnerships can become so careful about consensus that research never reaches action. Trust-building is essential, but it should eventually make decisions easier, not permanently postpone them.

Every project should identify a decision horizon: what might change if the evidence points one way rather than another?

39. Failure Mode: The Final Report Arrives After the World Changed

Staffing changed. The policy was replaced. A new technology altered practice. The original question is still answered rigorously, but the answer has lost its operating context.

Long-term partnerships should revisit assumptions and use staged findings when the environment is moving.

40. A Practical RPP Operating Cycle

1. Identify a consequential problem of practice.
2. Map the organisations and communities that possess relevant knowledge or authority.
3. Agree what mutual benefit means for each partner.
4. Make power, decision rights and constraints explicit.
5. Build governance for agenda-setting, data use, publication and conflict.
6. Translate the broad problem into a sequence of researchable uncertainties.
7. Match each question to the method capable of supporting the needed claim.
8. Define decision windows before analysis begins.
9. Create data agreements and shared definitions.
10. Build in practitioner and community interpretation.
11. Share preliminary evidence with appropriate caution.
12. Track how findings enter decisions and implementation.
13. Record disagreements and boundary conditions.
14. Produce outputs for different users without changing the strength of the claim.
15. Evaluate the partnership as well as the intervention.
16. Preserve history so turnover does not erase knowledge.
17. Update the shared agenda from what has been learned.

41. A Partnership Charter Worth Writing Down

• Problem: What shared improvement problem justifies the partnership?
• Participants: Whose expertise and lived experience must be represented?
• Authority: Who decides questions, methods, release and use?
• Data: Who owns, accesses, secures and may reuse information?
• Timelines: Which decisions require evidence by which dates?
• Methods: How will rigor be protected when timelines are tight?
• Products: Which outputs are needed for practice, policy, community and scholarship?
• Credit: How will intellectual and operational contribution be recognised?
• Disagreement: How will conflicting interpretations be handled?
• Negative findings: What protection exists for inconvenient evidence?
• Succession: How will new members inherit partnership memory?

42. Missing-Node Scan: Questions a Mature RPP Should Ask

• Can every partner state the shared problem in compatible language?
• Would practitioners still choose the current research agenda if no publication were possible?
• Would researchers still trust the design if leaders preferred a different result?
• Do students, families or communities influence questions where their experience is central?
• Are data definitions stable enough for the claims being made?
• Can the partnership produce useful evidence before major decision windows close?
• Are preliminary findings labelled with appropriate uncertainty?
• Does practitioner knowledge change interpretation rather than merely decorate it?
• Can external evidence challenge a strong local belief?
• Are negative findings preserved and discussed?
• Does the partnership know how evidence is actually used after delivery?
• Are scholarly and operational incentives both visible?
• Can new personnel understand why earlier decisions were made?
• Is power concentrated in funding, data ownership or publication control?
• Does the relationship create capability that will outlast a single grant?

43. Authoritative Research Starting Points

Useful starting points include the William T. Grant Foundation’s Research-Practice Partnerships resource and its 2021 field review, Research-Practice Partnerships in Education: The State of the Field; Cynthia Coburn and William Penuel’s 2016 Research–Practice Partnerships in Education: Outcomes, Dynamics, and Open Questions; and the Spencer Foundation’s current Research-Practice Partnership Grants framework, which recognises multiple collaborative forms.

Together these sources describe an approach that is more demanding than dissemination. It asks institutions to build durable relationships, connect diverse expertise, produce rigorous and usable evidence, attend to power and learn across multiple cycles of work.

44. The Return Path

At the beginning, the district asks a university team to explain why a programme is underperforming.

Two years later, the relationship looks different. The partners have learned that programme participation was recorded inconsistently, that schools implemented different components, that teacher planning time predicts whether the intended routine appears, and that one subgroup faces an access barrier invisible in the original dataset. The district changes implementation support. Researchers refine their model. Community partners help redefine the outcome. A follow-up study tests whether the change travels across schools.

The value is not one answer moving from the university to the school system. It is an evidence relationship capable of producing better questions and better decisions repeatedly.

Research–practice partnerships work when evidence stops being something one side produces for the other and becomes something different forms of expertise build, test, interpret and use together.