Will your numbers survive a funder's check? A practical guide to data quality

5 min read

Data quality is the degree to which program data is fit for the decisions and reports it supports. Good data measures what it is meant to measure, is collected the same way every time, is complete and on time, and is protected from manipulation. A data quality check compares reported figures with source records and fixes the system that produced any errors.

A report is only as credible as the figures behind it. When a funder checks a reported number against attendance sheets or service registers and the two do not match, trust drops fast. Errors found after a report goes out can lead to queries, corrections, and delayed funding. Errors found before the report goes out are just a task list.

The six dimensions of data quality

Data quality has recognized dimensions, and they come from long practice in global health programs. Data Diamond’s six dimensions are adapted from the user manual for the Routine Data Quality Assessment (RDQA) tool. MEASURE Evaluation developed the tool with partners including the World Health Organization and the Global Fund. The manual’s list also includes confidentiality (MEASURE Evaluation).

At Edit Right Advisors, we check data with Data Diamond, the data quality model in our Learning Loop framework. It tests six dimensions, each answering one question:

Dimension	The question it answers
Validity	Does the data measure what it is meant to measure?
Reliability	Is the data collected the same way every time?
Precision	Is the data detailed enough for the decisions it supports?
Integrity	Is the data protected from deliberate bias or manipulation?
Completeness	Is all the expected data present?
Timeliness	Is the data available when it is needed?

The goal at the top of the diamond is simple: data that is fit for decision making. Quality is judged by the use the data must serve, not by an abstract standard.

What each dimension looks like in practice

The dimensions become clearer with a plain example. Imagine a program that reports how many people completed a training course each quarter. The example is illustrative.

• Validity: the indicator counts people who completed the course, but staff record everyone who attended the first session.
• Reliability: one site counts a participant who missed one session as complete, and another site does not.
• Precision: the funder asks for results by sex and age, but the register records only names.
• Integrity: a site manager adjusts the count to meet a target before it is reported.
• Completeness: two of eight sites did not submit registers this quarter.
• Timeliness: registers arrive six weeks after the quarter ends, after the report is due.

Each of these problems produces a wrong number. Each one also has a different fix, which is why a single “check the data” step is not enough.

The four-step data quality cycle

Data Diamond pairs the six dimensions with a four-step cycle that repeats each reporting period. Its logic follows the RDQA approach, which combines verifying the data, assessing the system that produces it, and developing action plans to improve both.

1. Assess

Select the indicators, sites, and period to check, and review the data system behind them: the tools, forms, staff roles, and data flow. Start with the indicators that matter most to the funder and to your own decisions.

2. Verify

Check reported figures against source records. Recount a sample of registers or forms and compare the totals with what was reported. Record each difference and where in the data flow it arose.

3. Correct

Agree on corrective actions with owners and dates. Fix the reported figure if it is wrong, and fix the cause as well: a form, a definition, a training gap, or a missing step in the data flow.

4. Monitor

Recheck before the next report, then begin the cycle again. Improvements that are not rechecked tend to fade.

Where data quality usually breaks

Many data errors come from the system, not from careless people. Look first at these points:

• Indicator definitions that staff interpret differently
• Paper forms copied by hand into spreadsheets
• Totals added up at several levels, with no check between them
• No backup when the one person who compiles data is away
• Reporting deadlines set before the data can realistically arrive

Fixing one of these often removes a whole class of errors at once. That is why the system review in step one matters as much as the recount in step two.

How data quality connects to your wider monitoring

Data quality is one part of a monitoring, evaluation, and learning (MEL) plan, not a separate project. In The Edit Right Architecture, Data Diamond carries out the data quality part of MEL Map, our MEL plan model, and supplies trusted data to Learning Ladder for decisions. Verified data also supports the “accurate” level in Proof Pyramid, the quality check we apply to funder reports.

Frequently asked questions

How often should we check data quality?

At least once in each reporting period, and before any major report or evaluation. The schedule should fit the program and its reporting calendar.

Do we need to check every indicator?

No. Start with the indicators that drive funding decisions and your own management decisions. Rotate the rest across reporting periods.

Is a data quality check the same as an audit?

No. A data quality check tests whether program data is fit for decisions and reporting. It is not a financial audit and does not certify financial reports.

Talk it through

Data Diamond is part of SS-03 Monitoring, Evaluation, and Learning (MEL), delivered through Learning Loop. If a report or evaluation is coming up and you are not fully confident in the numbers, Book a free call. We will talk through your data and where to check first. You can also reach us at hello@editright.io or +1 314 998 0221.

Sources

• MEASURE Evaluation, Routine Data Quality Assessment Tool: User Manual (2017). Accessed October 27, 2026.