Category: Ideas & Perspectives
Abstract / Executive Summary
The growing availability of data has created opportunities for better decision-making, more efficient public services and improved environmental management. However, the volume of data collected by an organisation does not necessarily translate into better outcomes. Data that are inaccurate, incomplete, inconsistent, outdated or poorly documented can undermine analysis, distort reporting and weaken institutional trust. This article examines why data quality is often more consequential than data quantity, drawing on established research in information systems, the FAIR Data Principles and practical considerations in digital transformation.
The analysis distinguishes data volume from data quality and examines the characteristics that make information useful for a specific decision. It considers how poor-quality data affect analytics, artificial intelligence, public administration, environmental monitoring and institutional accountability. It also addresses an important qualification: large datasets can improve analysis when they provide representative coverage, capture rare events or enable meaningful comparisons. Quantity and quality are therefore complementary, but additional volume cannot reliably compensate for systematic errors or poorly defined measurements.
The article connects these principles to Eywa Systems’ documented work in national databases, geospatial platforms, environmental information systems and field data collection. Its company profile describes the use of structured databases, survey instruments, field data-quality controls and reporting dashboards, illustrating how quality considerations can be incorporated into system design and implementation. These examples demonstrate relevant practices rather than independently proving their effectiveness.
The central conclusion is that organisations should prioritise fit-for-purpose data, establish clear quality standards, validate information throughout its lifecycle and make data governance part of operational practice. The objective is not to collect less data by default, but to ensure that the data collected can support trustworthy, transparent and defensible decisions.
Keywords: Data quality; data governance; data management; decision intelligence; digital transformation; data integrity; environmental monitoring; information systems.
Main article
1. Introduction
Organisations increasingly rely on data to allocate resources, evaluate performance, manage services and understand complex systems. Governments use administrative records to plan public services; environmental agencies combine field observations and geospatial information to monitor ecosystems; and businesses use operational data to identify risks and improve processes.
The ability to collect information has expanded considerably, but collecting more data does not automatically improve the decisions made from it. A database may contain millions of records while still producing misleading results if its entries are inaccurate, its categories are inconsistent or important groups are missing.
This distinction matters because data are rarely valuable in isolation. Their value depends on whether they represent the phenomenon being examined, whether their meaning is understood and whether they are suitable for the decision at hand. Wang and Strong (1996) established an influential framework showing that data quality extends beyond accuracy to include contextual appropriateness, representation and accessibility.
The central question is therefore not simply how much information an organisation possesses, but whether that information can be trusted and used appropriately.
This article examines the relationship between data quality and quantity, the consequences of poor-quality information, and the practical measures organisations can adopt to improve data reliability. It also considers applications in public-sector digital systems and environmental management, including capabilities and documented assignments described by Eywa Systems.
2. Conceptual Background: What Makes Data High Quality?
Data quality refers to the degree to which data meet the requirements of their intended users and purposes. There is no single measure that adequately captures quality in every situation. A dataset suitable for exploratory research may be insufficient for regulatory reporting, financial reconciliation or a government decision affecting public resources.
2.1 Accuracy is necessary, but not sufficient
Accuracy concerns whether recorded values correctly represent the relevant facts or observations. An incorrect date, misclassified business or erroneous emissions figure can undermine subsequent analysis.
However, a dataset can contain individually accurate values and still be unsuitable for its intended purpose. For example, a survey may correctly record the responses of everyone interviewed but exclude remote communities that are central to the research question. The recorded answers may be accurate, while the resulting conclusions are not representative of the wider population.
Wang and Strong (1996) distinguish intrinsic quality from contextual, representational and accessibility dimensions. Their research supports a broader understanding: useful data must be credible, relevant to the task, clearly represented and accessible to those who need them.
2.2 The principal dimensions of data quality
The following dimensions provide a practical foundation for assessing whether a dataset can support reliable decisions.
| Dimension | Meaning | Example of a quality problem |
| Accuracy | Values correctly represent the relevant facts or observations. | An incorrect payment amount is recorded. |
| Completeness | Required information is present. | Survey records omit essential location fields. |
| Consistency | Values follow compatible definitions and rules across records or systems. | Two departments use different definitions of an active business. |
| Validity | Values conform to specified formats, ranges and business rules. | A date field accepts an impossible calendar date. |
| Timeliness | Information is sufficiently current for its intended use. | A report relies on outdated project status records. |
| Uniqueness | Records represent distinct entities appropriately, without unintended duplication. | The same organisation appears twice in a national registry. |
| Relevance | Information is appropriate to the question being addressed. | A dashboard contains extensive activity data but excludes outcome indicators. |
| Traceability | The origin, transformations and definitions of data can be understood. | A reported figure cannot be linked to its source records or calculation method. |
These dimensions are related but not interchangeable. Validation rules can detect an invalid date format without establishing whether the date itself is true. A complete dataset can contain duplicate records, while a timely dataset can remain systematically biased. Quality assurance must therefore examine several dimensions together rather than rely on a single score.
2.3 Data quantity and data quality are different properties
Data quantity describes the volume of information available, often measured by the number of records, observations, variables or files. Data quality concerns the reliability and suitability of that information.
A large dataset may provide broader coverage and support more detailed analysis. Yet its additional records can also reproduce the same errors, omissions or measurement biases at greater scale. Conversely, a smaller dataset with carefully defined variables, documented sources and appropriate coverage may be more useful for a specific operational decision.
The distinction is not an argument against collecting substantial datasets. It is an argument against treating volume as a substitute for quality.
3. Literature and Evidence Review
3.1 Quality must be evaluated in context
A major contribution of Wang and Strong (1996) was to examine data quality from the perspective of data consumers rather than restricting the concept to technical accuracy. Their framework recognises that users need information that is appropriate to their task, understandable and accessible.
This has a direct practical implication: quality standards should begin with the decisions that the data are expected to support. A national registry, for instance, may require unique identifiers and consistent administrative classifications. An environmental monitoring system may place greater emphasis on observation dates, geographic coordinates, measurement methods and the provenance of reported indicators.
The appropriate standard depends on the intended use. Nevertheless, contextual flexibility should not become an excuse for weak controls. Minimum requirements for accuracy, documentation, security and integrity remain necessary wherever the consequences of error justify them.
3.2 Reusability depends on more than collecting information
Wilkinson et al. (2016) introduced the FAIR Data Principles: data should be findable, accessible, interoperable and reusable. These principles were developed to improve the management and reuse of research data, with particular attention to making data usable by both people and machines.
FAIR is not a complete data-quality framework, and FAIR compliance does not guarantee that data are accurate or unbiased. Its contribution is different: even good data have limited value if researchers or systems cannot discover them, interpret their structure, access them appropriately or combine them reliably with other information.
This distinction is especially relevant to organisations operating multiple databases and reporting platforms. Consistent formats, clear metadata and documented interfaces can support integration, but they cannot correct an inaccurate underlying observation automatically.
Together, these foundational contributions suggest that data quality and data usability must be considered alongside the architecture through which information is stored, exchanged and interpreted.
3.3 What the evidence does not establish
The literature supports a multidimensional understanding of data quality, but it does not justify a universal numerical rule stating that quality always matters more than quantity. The relative importance of each depends on the analytical task.
For example, a small sample may fail to capture rare events or meaningful differences between population groups. In these circumstances, increasing the number of observations may improve the analysis, provided that the additional data are collected using suitable methods. A larger, well-designed dataset can be both more representative and more informative.
The defensible conclusion is narrower and more useful: data volume alone is an inadequate measure of decision readiness. Organisations need sufficient data and appropriate quality controls, with neither treated as a substitute for the other.
4. Analysis and Discussion: Why Data Quality Matters More Than Data Quantity
4.1 Poor-quality data can amplify errors in decision-making
Every analytical process depends on assumptions about the information it receives. If those assumptions are incorrect, sophisticated tools may produce precise-looking results that do not reflect reality.
Consider a hypothetical public-sector registry in which the same business is recorded under several different identifiers. A dashboard may overstate the number of registered businesses, while an eligibility assessment may fail to recognise that multiple records belong to the same organisation. Adding more records without resolving the duplication would not solve the problem.
The consequences extend beyond individual calculations. Decision-makers may allocate resources using distorted totals, report misleading performance figures or direct interventions towards the wrong groups. The underlying issue is not the absence of data, but the failure to establish whether the available information is fit for purpose.
More sophisticated analytics cannot, by themselves, establish that source data are correct. The solution requires appropriate definitions, validation, reconciliation and checks against authoritative evidence.
4.2 Data quality determines the usefulness of analytics and AI
Business intelligence and artificial intelligence depend on the data used to configure, train, test or operate their systems. Missing values, inconsistent labels and historical biases can affect the outputs of analytical models. However, the precise effect depends on the model, the nature of the defect and the task being performed; not every data error necessarily produces a material error in every result.
Quantity can help when additional observations improve coverage or capture patterns that would otherwise remain undetected. Yet more observations do not automatically remove systematic measurement errors or biased selection.
For organisations adopting AI, the practical priority should therefore be to understand data lineage, identify relevant quality risks and test whether the information represents the intended population and operational context. Data preparation and evaluation are essential parts of responsible deployment, not merely preliminary technical tasks.
A larger training dataset may improve model performance under suitable conditions. It cannot be assumed to correct missing populations, inappropriate labels or unreliable ground-truth measurements.
4.3 Reliable data strengthen institutional accountability
Public institutions must often explain how decisions were made, demonstrate compliance with requirements and provide consistent reports to oversight bodies, funders or the public. These responsibilities require more than large databases.
Data quality supports accountability by making records consistent, calculations reproducible and decisions traceable to documented information. Audit trails, standardised definitions and documented transformations help reviewers understand how reported figures were produced.
This is particularly relevant to digital public services and national information systems, where several agencies or teams may contribute data to a shared platform. If each institution uses different definitions or reporting periods, the resulting information may be difficult to reconcile even when individual systems function correctly.
The design of the information system must therefore address the governance of data: who creates records, who validates them, which definitions apply, who can change them and how corrections are documented.
4.4 Environmental monitoring illustrates the stakes
Environmental and climate programmes often combine information from field observations, administrative records, satellite imagery, geographic information systems and monitoring indicators. These sources may differ in resolution, collection frequency, classification and measurement method.
Combining them without appropriate checks can create misleading comparisons. A geographic boundary may change, an observation may refer to a different reporting period, or two indicators may use incompatible definitions. A consolidated dashboard can make such information appear coherent without resolving the underlying differences.
Quality controls should therefore include documented indicator definitions, appropriate geographic and temporal references, source provenance and procedures for reviewing anomalous values. Where uncertainty remains, it should be communicated rather than hidden by a single summary figure.
Eywa Systems’ company profile describes assignments involving the National Forest Monitoring System and Cambodia Environmental Management Information System, including the integration of more than 40 indicators spanning land use, land-use change and forestry, carbon, forest, biodiversity and socioeconomic domains. The profile also describes national carbon and safeguards information systems in Papua New Guinea.
These assignments illustrate the practical relevance of managing structured environmental information across multiple domains. They do not, by themselves, establish a quantified improvement in data accuracy or environmental outcomes. Such a conclusion would require suitable baseline measurements, evaluation methods and independently verifiable results.
4.5 Quality begins at the point of collection
Correcting unreliable information after it has entered a database can be expensive and difficult. Errors may already have propagated into reports, decisions and other connected systems. Prevention therefore deserves attention alongside later-stage cleaning.
Useful measures include designing clear data-collection forms, defining mandatory fields where appropriate, applying validation rules, standardising classifications and training the people responsible for recording information. Automated checks should be combined with human review because a value can satisfy a technical rule and still be factually incorrect.
Eywa Systems’ company profile describes the Nationwide SME Database assignment for the Global Green Growth Institute in Papua New Guinea. The documented scope includes survey-instrument design, enumerator training, a field data-collection application, a structured database, reporting dashboards and a field data-quality regime. At the time described in the profile, four of six contract deliverables had been accepted and field data collection was continuing.
This example illustrates how data quality can be addressed across collection, storage and reporting rather than treated solely as a database-cleaning exercise. Because the assignment was ongoing, its documented scope should not be interpreted as proof of completed nationwide data-quality outcomes.
4.6 Interoperability makes consistent definitions essential
Data are often distributed across applications, departments and partner organisations. When these systems exchange information, differences in naming conventions, identifiers, units and classification rules can reduce its usefulness.
Interoperability—the ability of systems to exchange information and use it meaningfully—therefore depends partly on shared definitions and appropriate technical standards. A common data format is helpful, but semantic consistency matters as well: two systems may both contain a field labelled “project status” while using different meanings for the values recorded.
A practical response is to establish shared data dictionaries, unique identifiers where appropriate, documented interfaces and rules for reconciling conflicting records. The FAIR Principles further support discoverability, interoperability and reuse through clear metadata and accessible, well-described data structures (Wilkinson et al., 2016).
The goal is not simply to connect more systems. It is to ensure that information retains its meaning as it moves between them.
5. Challenges, Limitations, and Counterarguments
5.1 Larger datasets can be necessary
The strongest counterargument is that quantity can materially improve data quality in the broader sense of analytical usefulness. More observations may improve coverage, support subgroup analysis, increase the chance of observing rare events and reduce sampling uncertainty when the sampling design and assumptions are appropriate.
This matters in environmental monitoring, population studies and AI applications. A small dataset can be systematically incomplete because it covers only a limited area or period. In such cases, collecting additional observations is essential.
However, sample size alone does not guarantee representativeness. A very large dataset collected through a biased process may still exclude important populations or systematically mismeasure the phenomenon under study.
5.2 Quality assurance has costs
Validation, documentation, staff training, data governance and periodic audits require time and resources. Excessive controls can delay collection or create unnecessary administrative burdens. Organisations must therefore prioritise quality dimensions according to the consequences of error and the decisions the data support.
Not every field needs the same level of scrutiny. A minor formatting inconsistency may have limited consequences in one context, while an incorrect financial value, geographic coordinate or eligibility indicator may have substantial implications in another.
5.3 Quality is not always directly observable
Organisations may find it difficult to determine whether every record is correct, especially when independent ground-truth information is unavailable. Completeness and format validity can often be measured automatically; factual accuracy may require field verification, external records or expert review.
A high quality score can consequently create false confidence if it measures only the dimensions that are easiest to quantify. Quality assessments should state what has been tested, which defects remain possible and how uncertainty affects interpretation.
5.4 Data quality does not guarantee good decisions
Even accurate, complete and timely information can be used badly. Decision-makers may misunderstand the evidence, ignore relevant context or prioritise objectives that are inconsistent with public needs. Institutional incentives, professional judgement and governance arrangements also influence outcomes.
Data quality is therefore a necessary foundation for many evidence-based decisions, not a sufficient guarantee of sound judgement. It must be combined with appropriate analysis, transparent reasoning and accountable decision-making.
6. Implications for Organisations and Digital System Design
Organisations can translate these principles into a practical approach to data management.
First, define the decision before defining the dataset. Identify what the information must support, which errors would matter and which quality dimensions are essential. This prevents data collection from expanding without a clear operational purpose.
Second, establish ownership and shared definitions. Assign responsibility for important datasets, document their meaning and specify who can create, validate, amend and approve records. Shared definitions reduce avoidable inconsistencies between teams and systems.
Third, build quality controls into collection and processing. Use suitable validation rules, duplicate detection, cross-field checks and review procedures. Automated controls should flag questionable records rather than silently convert uncertainty into apparently reliable information.
Fourth, monitor quality throughout the data lifecycle. Quality can deteriorate when information becomes outdated, systems change or new sources are integrated. Periodic checks should examine accuracy, completeness, consistency, timeliness and traceability, with findings assigned to responsible staff.
Fifth, make uncertainty visible. Reports should identify missing information, limitations in coverage and material differences in definitions. Decision-makers should be able to distinguish verified facts from estimates and provisional figures.
Finally, invest in usability and institutional capability. Reliable information must be understandable and accessible to authorised users. Training, documentation and clear reporting processes help ensure that quality controls continue to function after a system is deployed.
These priorities align with the delivery approach described in Eywa Systems’ company profile, which includes requirements validation, functional and user acceptance testing, defect tracking, documentation, training and institutional handover. The profile documents relevant practices and capabilities; it does not establish that every recommended control is implemented identically in every assignment.
The broader lesson is that data quality should be treated as an ongoing organisational responsibility rather than a one-off technical task. Its effectiveness depends on people, processes, technology and governance working together.
7. Conclusion
Data quantity can expand analytical possibilities, improve coverage and support the detection of patterns. Yet volume alone cannot establish whether information is accurate, representative, consistent or suitable for a particular decision. Poor-quality data can undermine dashboards, AI systems, public reporting and environmental monitoring, even when the underlying databases contain large numbers of records.
The research reviewed here supports a multidimensional approach to quality, encompassing accuracy, contextual relevance, representation, accessibility and reusability. Practical applications in public-sector and environmental information systems reinforce the need to consider data quality from initial collection through integration, analysis and reporting. The documented Eywa Systems assignments illustrate relevant implementation contexts, although they do not independently demonstrate quantified improvements in data quality.
The appropriate objective is therefore neither maximum data collection nor minimum data collection. It is the collection and responsible management of sufficient, fit-for-purpose information, supported by clear definitions, proportionate validation, transparent limitations and accountable ownership.
Organisations that invest in these foundations are better positioned to understand what their data can support—and to recognise when the evidence is not yet strong enough to justify a decision.
References
- Wang, R. Y., & Strong, D. M. (1996). Beyond accuracy: What data quality means to data consumers. Journal of Management Information Systems, 12(4), 5–33. https://doi.org/10.1080/07421222.1996.11518099
- Wilkinson, M. D., Dumontier, M., Aalbersberg, I. J., Appleton, G., Axton, M., Baak, A., Blomberg, N., Boiten, J.-W., da Silva Santos, L. B., Bourne, P. E., Bouwman, J., Brookes, A. J., Clark, T., Crosas, M., Dillo, I., Dumon, O., Edmunds, S., Evelo, C. T., Finkers, R., … Mons, B. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data, 3, Article 160018. https://doi.org/10.1038/sdata.2016.18