How to Conduct a Data Quality Assessment

Who This Page Is For

You manage or oversee monitoring data. You know the numbers going into your reports, but you are not confident those numbers are accurate. Maybe a field team is aggregating data manually. Maybe two sites report the same indicator differently. Maybe you just inherited a project and have no idea whether to trust what is in the database. This page walks you through conducting a Data Quality Assessment from start to finish.

A DQA is not a data cleaning exercise. It is a systematic check of whether your data quality assurance systems actually work. You are auditing the process, not just the numbers.

The Five Data Quality Dimensions

Every DQA scores data against the same five dimensions. Learn these before you start. They are the backbone of the entire assessment.

Step-by-Step: Conducting the DQA

Step 1: Select Your Indicators

You cannot assess every indicator in your logframe. That would take weeks and produce a report nobody reads. Pick 5-8 indicators using these criteria:

• Decision weight. Prioritize indicators tied to major reporting or funding decisions.
• Suspicion. Include any indicator where you have seen inconsistencies or received complaints about data quality.
• Complexity. Include indicators with multi-step data flows (collected at site level, aggregated at district, reported at national). More steps mean more opportunities for error.
• Mix of types. Include at least one output indicator and one outcome indicator. Include both quantitative and qualitative data sources if your program uses both.

See indicator selection for broader guidance on choosing the right indicators.

Step 2: Assemble the DQA Team

A DQA should not be conducted by the same people who collect and report the data. That defeats the purpose. Your team needs:

• A lead assessor who understands M&E systems and can interview staff about data processes. This is often a senior M&E officer from headquarters or a regional office.
• A data analyst who can pull records, trace numbers, and run concordance checks.
• An external member if possible. Even one person from outside the project (a partner organization, a technical advisor) adds credibility.

Two to three people is the right size. One person cannot conduct site visits and run verification checks simultaneously. More than four creates coordination overhead that slows the process.

Step 3: Prepare Your Verification Tools

Before visiting any site, prepare:

• Indicator reference sheets for each selected indicator: definition, data source, collection method, reporting frequency, responsible person.
• A data trail map for each indicator: where data originates, every handoff point, where it ends up in reports. Draw this as a simple flowchart.
• Verification worksheets with columns for: reported value, recounted value, difference, explanation.
• Interview guides for staff at each level of the data flow (field team, data entry clerk, site manager, M&E officer). Ask: What do you collect? How? Using what tool? How often? Who do you send it to? What training did you receive?
• A sample of records to recount. Pull 10-20% of records for your selected indicators from the most recent reporting period.

Step 4: Conduct the Site Visit

Visit at least two data collection or aggregation points per indicator. During the visit:

Observe the system. Watch how data is collected, entered, and stored. Are paper forms locked in a cabinet? Is the electronic database password-protected? Are forms completed in pen, or are pencil entries being erased and changed?

Interview staff. Ask open-ended questions. "Walk me through what happens when you finish a home visit" reveals more than "Do you fill in the form correctly?" Listen for gaps between what the protocol says and what staff actually do.

Recount the data. Take the source documents (registers, forms, tally sheets) and independently count the totals for your selected indicators. Compare your recount to the reported figures. Record every discrepancy.

Check the trail. Follow one indicator from the original source document through every aggregation step to the final reported number. Flag every point where a number changes without a clear, documented reason.

Step 5: Score Each Dimension

For each indicator, score all five dimensions. Use a three-point or five-point scale. A three-point scale is simpler and reduces false precision:

Score each dimension separately. An indicator can be strong on timeliness but weak on integrity. The per-dimension breakdown is what makes the DQA actionable.

Step 6: Calculate Concordance

Concordance is the quantitative core of the DQA. It compares your recounted values to the reported values.

Formula: Concordance = (Recounted Value / Reported Value) x 100

Calculate concordance for each indicator at each site. If you visited three sites and two show 98% concordance while one shows 74%, the problem is site-specific, not systemic. That distinction matters for your corrective actions.

Step 7: Write the DQA Report

Your report should be short and action-oriented. Structure it as follows:

1. Executive summary (half a page). Overall findings, number of indicators assessed, number of sites visited, headline concordance results.
2. Methodology (one page). Which indicators, which sites, who conducted the assessment, what period the data covers.
3. Findings by indicator (one page per indicator). Dimension scores, concordance results, key observations from site visits. Include a table for each indicator.
4. Cross-cutting findings (one page). Patterns that affect multiple indicators: training gaps, system weaknesses, process bottlenecks.
5. Action plan (one page). Specific corrective actions, who is responsible, deadline for completion.

Do not write a 40-page DQA report. Nobody will read it. Ten to fifteen pages including tables is sufficient for most programs.

Step 8: Follow Up

A DQA without follow-up is a waste of time. Within two weeks of completing the assessment:

• Share findings with every team that provided data. Do this in person or by video, not just by emailing the report.
• Assign corrective actions with names and deadlines. "Improve data quality" is not an action. "Retrain field staff on Indicator 3 definition by May 15" is.
• Schedule a follow-up check 60-90 days after the DQA to verify corrective actions were implemented.
• Update your data management protocols to address systemic issues. If three sites had the same problem, the protocol needs to change, not just the sites.

Use the DQA Scorecard to track dimension scores over time and see whether your corrective actions are working.

Common Mistakes

Mistake 1: Assessing every indicator. A DQA that covers 25 indicators takes so long that findings are outdated before the report is written. Prioritize 5-8 indicators. Do a thorough job on fewer indicators rather than a superficial pass on all of them.

Mistake 2: Letting the data team assess their own data. Self-assessment produces predictable results: everything looks fine. Bring in someone from outside the immediate data chain. Independence is what makes a DQA credible.

Mistake 3: Skipping the site visit. A desk-based DQA catches arithmetic errors but misses process failures. You cannot assess reliability, integrity, or timeliness without talking to the people who collect the data and observing how they work.

Mistake 4: Writing findings without an action plan. "Concordance for Indicator 4 was 78%" is a finding. Without a corrective action attached to it, it is useless information. Every finding below acceptable thresholds needs a named action, a responsible person, and a deadline.

Mistake 5: Treating the DQA as a one-time event. A single DQA tells you the state of data quality at one moment. Systems degrade. Staff turn over. New tools get introduced. Build DQA into your annual M&E calendar and repeat it.

DQA Checklist

Use this before, during, and after the assessment.

Preparation:

• Selected 5-8 indicators based on decision weight, suspicion, and complexity
• Assembled a DQA team with at least one member outside the data chain
• Prepared indicator reference sheets for each selected indicator
• Mapped the data trail for each indicator (source to report)
• Created verification worksheets with space for recounted values
• Pulled a 10-20% sample of records for recounting
• Drafted interview guides for staff at each data flow level

During the Site Visit:

• Visited at least two data collection or aggregation points per indicator
• Observed data collection, entry, and storage processes firsthand
• Interviewed staff using open-ended questions about their actual procedures
• Independently recounted data from source documents
• Traced at least one indicator through every aggregation step

After the Assessment:

• Scored all five dimensions for each indicator
• Calculated concordance for each indicator at each site
• Written a DQA report of 10-15 pages with findings and action plan
• Shared findings with all data-providing teams within two weeks
• Assigned corrective actions with names and deadlines
• Scheduled a 60-90 day follow-up check