Data Quality in M&E
Poor data quality is the most common reason M&E findings go unused. This guide covers what data quality means in practice, the five dimensions every practitioner should assess, and how to run a targeted data quality assessment that improves decisions without consuming your entire budget.
What is data quality in M&E?
Data quality in M&E means the data you collect is accurate enough, complete enough, and timely enough to support the decisions your program needs to make. It does not mean perfect data. It means data that is fit for its intended purpose.
This distinction matters because the traditional approach to data quality (exhaustive cleaning of every data point) consumes M&E resources that would have more impact elsewhere. Many programs spend weeks cleaning data that no one uses for any decision, while the indicators that actually drive resource allocation and program adjustments receive no special scrutiny at all.
What is a data quality assessment (DQA)?
A DQA is a systematic review of whether your data meets the quality standards required for its intended use. Most donors require at least one formal DQA per project cycle. USAID, FCDO, and the World Bank all have DQA requirements built into their monitoring frameworks, and many request annual or mid-term DQAs for major programs.
A DQA typically covers: tracing data from source documents to reported values, checking for completeness and consistency, validating indicator definitions against collection practice, and documenting findings. The output is a report that either confirms data integrity or identifies specific gaps to remediate before the next reporting cycle.
A common DQA failure pattern
The most common DQA failure is not inaccurate data; it is a DQA conducted too late, covering the wrong data, and producing a report that nobody acts on. A DQA done two weeks before a final evaluation, covering data already submitted to the donor, finds problems it cannot fix. A well-timed DQA at month 6 of a 24-month project finds the same problems while there is still time to correct the data collection process and retrain enumerators.
The Five Dimensions of Data Quality
These five dimensions are used by USAID, FCDO, and most major donors to assess whether M&E data is fit for reporting and decision-making.
Validity
The data measures what it is supposed to measure. A count of "households with improved water access" is only valid if the definition of "improved" is applied consistently by every enumerator, in every community, across the full data collection period.
Indicator definitions are ambiguous or interpreted differently across teams.
Reliability
The data collection method produces consistent results when applied under the same conditions. If two enumerators ask the same household the same question and get different answers, reliability is compromised, usually by unclear methodology, poor training, or biased question design.
Enumerator-to-enumerator variation in results for the same population.
Timeliness
Data is available when decisions need to be made. A quarterly data collection cycle that produces results two months after the reporting deadline has nearly zero decision value, even if the data is accurate. Timeliness is specifically about the relationship between data availability and decision timing.
Program decisions are being made on outdated data or without any data.
Precision
Data is specific enough to be useful. An indicator measured to the wrong level of disaggregation, or reported as a range so wide it cannot distinguish between success and failure, lacks precision. Precision failures often trace back to indicator definitions that were not specific enough at the design stage.
Results reported as broad ranges or without disaggregation that decisions require.
Integrity
Data has not been manipulated, falsified, or inadvertently corrupted between collection and reporting. Integrity failures include enumerator fraud, transcription errors during manual data entry, and formula errors in analysis spreadsheets. It also covers the chain of custody: who had access to data and when.
Unexplained gaps in datasets, implausibly round numbers, or results that cannot be traced back to source data.
Running a Targeted DQA
A 4-step process focused on the data that matters most. Not every indicator needs the same level of scrutiny; this process starts with prioritization.
Identify which data matters most
Not all data needs the same level of scrutiny. Start by listing the decisions your program needs to make this reporting cycle. Which indicators directly feed those decisions? Those are your tier-1 data: they get rigorous checks. Everything else gets spot-checks.
A single tier-1 indicator that directly informs a scale-up or pivot decision is worth more scrutiny than ten background indicators that end up in an appendix.
Check completeness and consistency
Pull the dataset for each tier-1 indicator. Check: are all expected records present? Do values fall within plausible ranges? Do totals reconcile across collection points? Completeness and consistency failures are the most common and usually the easiest to fix if caught early.
Sort by enumerator or data collector ID before checking consistency. Systematic errors almost always cluster around one person or one collection point.
Verify against source documents
For a sample of records (10-20% for critical data), trace the value back to its source document: the field form, the intake register, the photograph. If the number cannot be traced to a source, it cannot be defended to a donor or auditor.
Do spot-checks during data collection, not only at reporting time. Catching a systematic error at week 2 of a 12-week collection is much cheaper than at week 12.
Document findings and remediation
Record what you found, what you changed, and what you left unchanged and why. A DQA that produces no documentation did not happen, as far as your donor is concerned. Findings should feed directly into the next data collection cycle, not sit in a report nobody reads.
Keep a running data quality log, not a one-off report. Log issues as they are found so patterns accumulate over time.
Free DQA Scorecard
We are building a free interactive DQA Scorecard that walks you through a structured assessment of each indicator across all five quality dimensions, and produces a prioritized action plan. Built for field teams, no spreadsheet setup required.
Indicator Quality Starts at Design
Most data quality problems are set in motion before data collection begins. Indicators that are poorly defined produce data that cannot be cleaned into usability. The problem is upstream.
Specific definition
Every indicator needs a single, unambiguous definition that any enumerator can apply without judgment calls.
Clear data source
Where the data comes from is specified, not "program records" but "monthly intake register at facility X."
Disaggregation plan
Outcome indicators should specify at minimum one disaggregation dimension (sex, age, location) documented in the indicator plan before data collection starts.
Realistic frequency
Collection frequency matches the rate of change in the indicator. Measuring an outcome monthly when change occurs over years wastes resources and produces noise.
Browse 2,900+ donor-aligned indicators with definitions, data sources, and disaggregation guidance already built in.
Browse the indicator libraryUsing AI for data quality work
AI tools can accelerate two of the most time-consuming DQA tasks: cleaning messy datasets and detecting anomalies that signal collection problems. The AI for M&E guide collection includes dedicated guides on cleaning messy data and spotting anomalies in monitoring datasets.
Free DQA Tools
Practical templates for running your own data quality assessment.
Data Quality Assessment Checklist
A structured checklist for assessing each indicator across the five quality dimensions. Includes documentation fields and an action plan template.
More M&E methodology guides
Practical, plain-language guides for every phase of the M&E cycle.
Browse all guides