1 of 11

Quality Assurance process

Technical Reporting information session #5

28 October 2025

​

Mariagiulia Mariani, QA Lead – PPU

​

1

2 of 11

Objective and Content

Provide an overview of the 2025 Quality Assurance process, timeline and share key insights from previous learning sessions

* 2025 Quality Assurance process & timeline

* Key insights and takeaways from previous learning sessions

* QA approaches for pooled results vs. w3/bilat results

* Leveraging AI to strengthen the QA process

* Brief presentation of the QA module (refresher)

2

3 of 11

What is QA in CGIAR

 

Aims to

Improve data quality, consistency, and reliability

Inform learning and continuous improvement

Strengthen CGIAR accountability and transparency

QA includes

Checking reported data against agreed criteria

Supporting data quality improvement (pre-QA)

Streamlining QA processes to ensure value > cost

QAed data are used for

Featured in the Results Dashboard

Integrated into all 2025 Technical Reporting products (SP/A annual technical reports, Portfolio Narrative)

Included in the CGIAR Annual Report

CGIAR’s Quality Assurance Process: A snapshot of what it is and what is does

3

4 of 11

2025 Quality Assurance process: what is QAed�

All pooled results submitted via the PRMS Reporting Tool by CGIAR Science Programs/Accelerators are QAed.

​

Result Categories:

  • Outputs: Knowledge products, Innovation development, Capacity sharing, Other
      • All knowledge products (KPs) that are neither PR papers nor MELIA studies undergo QA directly by CGSpace curators.
  • Outcomes: Policy changes, Innovation use (including IPSR pathways), Other

​

Process:

  • 1 Round: Low-priority fields → screened against criteria; final responsibility with reporting teams
  • 2 Rounds: High-priority fields → lead assessors and third-party review ensures accuracy
    • Not QAed: Data points not verifiable in a cost-efficient way (e.g. collaborators)

​

�The third-party broker mediates when needed

High-priority fields​ undergoing the 2nd Round

Result level​

Result type​

Evidence ​

KP type (if MELIA)  ​

Innovation Readiness Level​

Policy Change Stage ​

Core and Complementary Innovation Use Levels (Evidence-Based)​

The full list of data fields undergoing QA will be shared soon with the 2025 QA Assessors’ Guidance, with no significant changes compared to 2024.

4

5 of 11

QA approaches for pooled results vs. w3/bilat results

Pooled Results:

    • Subject to centralized CGIAR-wide QA ensuring credibility, consistency, and adherence to system-wide standards.
    • Multi-round, structured review with dual independent assessors, dialogue, and third-party arbitration for priority data fields.
    • Fully integrated in PRMS with AI-assisted validation supporting efficient data quality control and transparent reporting in official dashboards and reports.

W3 / Bilateral Results:

    • Included in 2025 technical reporting as part of a unified CGIAR portfolio but handled differently due to funding and reporting requirements diversity.
    • QA primarily conducted at Center level, focusing on minimum data standards (see Annex 3 in the TRA) and evidence to support the claim.
    • P/A review results for ToC alignment, geographic focus, and duplication avoidance before final submission.
    • More detailed assessments will be added from 2026 onwards.

Transparency in Reporting:

  • Combined presentation of pooled and W3/bilateral results in reports with distinct visual identities and legends clarifying QA scope differences.
  • Transition year with plans to progressively harmonize QA across all funding streams.

5

6 of 11

2025 Quality Assurance timeline and process�

 

Anonymous assessors

Agree/disagree and make adjustments

Two assessments are made (only one comment per field)

SP/As

Lead assessors

Final decision is taken on unresolved disagreements

Highlight:

 - unresolved disagreement 

 - unimplemented changes

3rd party 

PPU/PCU

ROUND 1: for all fields

ROUND 2: for priority/core fields

Implementation of changes in the PRMS

Highlights: 

  • One main batch is planned, with 8 working days window for submitters to review and address comments.
  • Extended submission for Knowledge Products – hard reporting deadline is February 19.

​

Data submission

Programs/Accelerators

Data submission on January 30

Feb 2 – Feb 9

Feb 10 - 20

Feb 20-22 

Feb 23-25

From Feb 26

Extended submission of KP in PRMS by Feb 20

6

7 of 11

Key insights and takeaways from previous learning sessions

1. Focus on Outcome-Level Data�QA efforts will concentrate on outcome-level results to strengthen strategic alignment and ensure data quality.

2. Quality of Evidence (especially for Outcomes)�Submitters must provide clear, credible, and well-documented evidence for each outcome.�For lengthy documents, indicate exact sections or page numbers where relevant information can be found.

3. Relevance of Reporting�Results should be clearly linked to the ToC and planned indicators.�Avoid overreporting and unnecessary detail, which reduce strategic value and complicate QA.

4. Timely and Responsive Engagement�Submitters are encouraged to address QA comments promptly.�The 2024 cycle showed fewer pending comments, improving overall efficiency.�(Note: responding to accepted comments is optional.)

5. Improved Transparency�Enhancing transparency on how data are used and QA is conducted builds trust.�For instance, QA certificates for innovations with high readiness levels reinforce the credibility of the process.

​

​

7

8 of 11

Leveraging AI to strengthen the QA process�

    • Launched the first system-wide pilot of AI-assisted QA tools for predicting Impact Area scores and Innovation Readiness Levels.
    • AI efficiently supported assessors scoring by processing of extensive evidence.
    • Human oversight ensured balanced assessments, mitigating AI limitations in handling complex information.

2024 AI testing achievements

    • Scale up AI use with refined models for Impact Areas and Readiness Levels using improved prompt engineering. The goal is to reach a stage where AI independently handles these data points, and QA assessor intervene only in cases of discrepancies between the submitter and the AI-generated result.
    • Expand AI-assisted QA into new domains, including Innovation Use and Impact Area subcategories.
    • Test integration of AI upstream (in PRMS) to improve data quality through enhanced inputs and pre-validation (e.g. for titles).

2025 AI development/use plans

8

9 of 11

Brief presentation of the QA module (refresher)�

Would additional technical training in January be useful?

How to access the QA Module in the PRMS Reporting Tool

​

9

10 of 11

10

CGIAR QA PLATFORM  Programs/Accelerators

USER MANUAL

  1. HOME

1. Side menu

This is the main menu in the QA Platform. It allows Initiatives to navigate between the different indicators and the home page.

​

2. Indicator dashboard

This provides graphical representations of the QA process, organized by indicator.

​

3. New comments report

This new functionality allows users to download the full list of comments for each P/A, with an option to filter by indicator category if desired.

​

3

2

1

10

11 of 11

Additional Resources

Info sessions

  • Live Virtual Tour of the QA Platform to be scheduled in January if requested
  • Support sessions during the QA process

​

Contact

  • Support with technical issues in the QA platform: PRMSTechSupport@cgiar.org
  • For general enquiries on the QA process: performanceandresults@cgiar.org 

​

Support Materials

​

Please see P&R for further updates included webinar recording, PPT slides, Guidance documents and FAQs.

​

11