Fraud. AI-generated responses. Professional survey respondents. In recent years, these have become some of the most widely discussed topics within the research industry. As digital methodologies have evolved, so too have the challenges associated with maintaining confidence in the data we collect.
Michael Deeming, Data Solutions Manager
While our approach to managing data quality in social research must continue to evolve, confidence in research still depends on the same foundations: clear objectives, robust methodology, appropriate quality assurance, professional judgement and ultimately, impact.
But focusing solely on these threats misses the bigger picture. At M·E·L Research, we believe data quality is built throughout the research process, not simply checked at the end. The challenge is not simply to identify poor-quality data, but to do so without excluding the genuine experiences and legitimate differences that research is intended to capture.
Why data quality matters
For research to have impact, it must be built on trust. Trust that the evidence accurately reflects the people, experiences and behaviours it is intended to represent.
Whether research is used to shape public policy, improve services or understand customer experiences, confidence in the findings ultimately depends on the quality of the underlying data. High-quality data should accurately and reliably reflect the population the research is intended to represent.
However, researchers must now also contend with increasingly sophisticated fraud, survey fatigue, disengaged respondents and AI-generated content.
Addressing these challenges requires a broad approach to data quality that considers every stage of a study, from defining the research objectives through to analysing and reporting the findings.
The changing data quality landscape
Digital data collection has transformed the way research is conducted. It has improved the speed and efficiency of fieldwork and provided access to larger and more diverse audiences. However, it has also introduced new risks that require different approaches to maintaining quality.
These challenges are often most visible in online research, where the scale and accessibility of data collection can increase the risk of fraudulent, duplicate or low-quality responses. However, the principles of data quality apply equally to telephone, face-to-face and mixed-mode studies.
Respondent fraud has become more sophisticated. Bots, duplicate accounts and professional survey respondents can be increasingly difficult to identify.
At the same time, advances in artificial intelligence have created opportunities for both researchers and those attempting to circumvent quality controls. AI can help researchers identify unusual response patterns and suspicious answers more efficiently, but it can also be used to produce plausible open-ended responses that are harder to detect through traditional checks.
Importantly, not every data quality issue is the result of deliberate fraud. Survey fatigue, declining engagement and increasing competition for people’s attention can all affect the quality of responses.
A participant who rushes through a survey, answers inconsistently or submits a generic AI-generated response may not be intentionally trying to deceive the researcher. Nevertheless, the resulting data may still be unsuitable for analysis.
These developments have made maintaining data quality more complex. Researchers must continually adapt their approaches, not only to identify unreliable responses, but also to ensure that genuine experiences and legitimate differences remain represented in the evidence. Excluding valid responses can be just as damaging as retaining poor-quality data, reducing confidence that the findings accurately reflect the population the research is intended to represent and ultimately limiting the impact of the research.
Building quality by design
At M·E·L Research, building data quality into every stage of a project means making informed decisions from the outset, rather than relying on quality checks once fieldwork is complete. Every stage of the research process contributes to the integrity of the final dataset, from defining clear research objectives and selecting the right methodology through to recruitment, questionnaire design, fieldwork, analysis and reporting. Each decision influences the quality of the evidence organisations ultimately rely upon.
The research lifecycle below illustrates how we embed quality throughout a study. While the specific activities vary between projects, the principle remains the same: confidence in the findings is built through a series of informed decisions made throughout the research process, not a single quality check at the end.
Building quality into the research process begins with clearly defined objectives and selecting an appropriate methodology. Researchers must understand who the research needs to represent, how participants will be reached and what evidence is required to answer the research questions.
At M·E·L Research, this increasingly means using blended methodologies where appropriate to ensure that people who may be digitally excluded are not unintentionally excluded from the research itself. In our work with the Bar Standards Board, we combined online and offline approaches to reduce the risk of digital exclusion and ensure the research reflected a broader range of experiences.
Careful sampling and recruitment help ensure that the right people take part. Clear screening criteria reduce the risk of unsuitable participants entering a study, while proportionate verification measures provide additional assurance where the risk of fraud is higher.
For sensitive topics, we use trauma-informed recruitment approaches and carefully designed pre-screening to protect participants while ensuring the research reaches those with relevant lived experience.
Where appropriate, we work with specialist clinical leads to design and oversee recruitment and screening, ensuring safeguarding and participant wellbeing are embedded throughout. We have also partnered with Enna Global, a leading neurodiversity and neuroinclusion consultancy, to strengthen the neuroinclusive quality of our research, communications and participant engagement, embedding accessibility throughout every stage of the research journey.
Questionnaire designis equally important. Clear wording, logical question flow and an appropriate survey length encourage meaningful participation while reducing respondent burden. At M·E·L Research, accessibility is considered from the outset, from using inclusive language to tailoring the appearance of surveys with features such as dyslexia-friendly fonts and layouts.
Where appropriate, we also give participants the option to have survey questions read aloud and to record open-ended responses instead of typing them, helping to remove barriers to participation. This approach informed our work with the British Dyslexia Association, enabling participants to engage with the research in ways that suited their individual needs. Reducing barriers to participation helped ensure the findings reflected a broader range of perspectives and strengthened confidence that the evidence accurately represented the people the research was intended to capture.
Quality should continue to be monitored while fieldwork is underway rather than waiting until data collection has finished. Real-time monitoring allows researchers to identify unusual response patterns, investigate anomalies and take corrective action before poor-quality data becomes embedded within the final dataset.
Technology plays an increasingly important role in supporting this process. Behavioural metrics, automated quality checks, machine learning and artificial intelligence can all help identify potential issues more efficiently. However, these tools are most effective when combined with robust methodology, professional judgement and human oversight.
Quality assurance continues after fieldwork. During analysis, researchers should validate the data, investigate anomalies and apply appropriate statistical techniques where required. Reporting should communicate findings accurately and transparently, providing appropriate context and acknowledging any limitations so that decision-makers can understand the strength of the evidence.
At M·E·L Research, we believe effective quality assurance should strengthen confidence in the evidence without creating unnecessary barriers for genuine participants. There is no single technology or quality check that can guarantee reliable data. Instead, confidence is built through multiple layers of proportionate quality assurance, tailored to the methodology, objectives and risks of each study.
Applying controls too rigidly can create a different kind of risk. Genuine participants may be excluded because their responses appear unusual, while accessibility needs, language differences or individual circumstances can produce patterns that automated systems interpret incorrectly. Ironically, these are often the perspectives that research most needs to understand. Effective quality assurance is therefore not about removing as many responses as possible, but about distinguishing unreliable data from genuine experiences to ensure the evidence remains both robust and representative.
Ultimately, data quality depends on combining robust methodology, appropriate technology and professional judgement. When quality is embedded throughout the research process and applied proportionately to the risks of each study, organisations can have confidence that the evidence they rely upon is robust, representative and fit for purpose.
Looking ahead
The challenges surrounding data quality will continue to evolve. Artificial intelligence will become more capable, patterns of respondent behaviour will continue to change, and new methods of data collection will introduce fresh opportunities as well as new risks.
Researchers will therefore need to keep adapting the tools and techniques used to protect the integrity of the evidence they collect.
However, the foundations of reliable research remain unchanged.
Technology can strengthen good research practice, but it cannot replace it.
For organisations commissioning research, this means asking not only what the findings show, but also how confidence in those findings has been established. Quality should be considered in the design of a study, monitored during its delivery and demonstrated transparently in its reporting.
As researchers, our responsibility extends beyond collecting information. We must ensure that the evidence we provide is robust, transparent and appropriate for the decisions it is intended to inform. And at M·E·L Research, that philosophy underpins the way we design, deliver and quality assure our studies.
In an increasingly complex research landscape, confidence will depend not only on how quickly data can be collected, but on how confidently researchers can stand behind it.
Whether you’re a public body, charity or organisation looking to understand your audience, evaluate impact, or inform future strategies, we’re here to help.
Get in touch with the M·E·L Research team today via our ‘Get in Touch’ form below to see how our expert researchers can support your goals. Alternatively, you can email the team at info@melresearch.co.uk.
About the author
Michael joined M·E·L Research in January 2022 and has over 20 years’ experience across research and data services. Throughout his career, he has developed a broad understanding of the research lifecycle, from research design and data collection through to data processing, analysis, reporting and visualisation.
As Data Solutions Manager, Michael leads the Data Solutions team, overseeing the delivery of survey scripting, data processing, reporting, visualisation and technical solutions that support every stage of the research lifecycle. He works closely with colleagues and clients to develop efficient, innovative solutions and automation that enhance data quality, improve processes and maximise the impact of research.
We are delighted to be working with Theos, the religion and society think tank, on an important new research project examining how people in Britain understand and navigate today’s emerging moral issues. Madeleine Pennington, Head of Research at Theos, commented: “We are pleased to be partnering with M·E·L Research on this timely and ambitious study. We are excited to generate new evidence about how people across Britain think about…
At M·E·L Research, we believe the strongest neuroinclusive social research is accessible by design. As part of this commitment, every member of our team recently completed specialist neurodiversity training with Enna Global, a leading neurodiversity and neuroinclusion consultancy.
The training marked the beginning of a wider collaboration. We’re delighted to be partnering with Enna Global to further…
We are pleased to announce the appointment of Senior Evaluation Lead Liz Murphy, alongside two new Insights Executives, Jamie Bains and Olivia Smith.
These appointments reflect our continued growth and investment in exceptional talent across our research delivery team. Together, Liz, Jamie and Olivia bring expertise spanning public policy, evaluation, social research, consumer insight, and culture and heritage, further strengthening our ability…
As artificial intelligence continues to transform industries across the world, it is easy to assume that technology will eventually replace the human role in research.
There is no question that AI is changing the way research is conducted and will continue to as technology develops. From automating repetitive tasks to analysing large datasets at speed, it is already improving efficiency across the market and social research industry. However, when the focus shifts from processing information to understanding people, human expertise remains indispensable.
Social research is fundamentally… social. It is about understanding humans’ motivations, fears…
I should probably start with a caveat. This is not a serious constitutional proposal. Or at least, it wasn’t when I started writing it. It began as a slightly tongue-in-cheek response to the latest instalment in what now feels like Britain’s longest-running political drama: “Who is in Number 10 this week?”
That might sound like an odd starting point for a piece on a research agency website. But many of our clients at M·E·L Research operate in spaces that are directly shaped by political decision-making: health, education, local government, regulation, justice, transport, skills, the environment and public services more broadly. For them, who makes decisions matters. So does how long those decisions…
New research conducted by M·E·L Research for the Molly Rose Foundation suggests that many young people continue to encounter harmful suicide, self-harm, depression and eating disorder content online, despite protections introduced under the Online Safety Act.
The research suggests that, despite significant regulatory changes, many children continue to encounter harmful content as part of their everyday online lives. A key question in the ongoing debate is the extent to which recommender feeds drive children’s exposure to harmful content.
The findings arrive at a time of intense debate about how best to protect children online…
We are pleased to see the publication of a new research report for the Legal Services Board (LSB) titled ‘AI in Legal Services’ exploring the opportunities and risks associated with AI in legal services. As artificial intelligence becomes increasingly embedded in everyday life, more consumers are turning to AI-powered tools for help with legal issues such as housing disputes, employment rights, debt and family matters. For many people who cannot afford traditional legal advice, these technologies offer a potentially transformative route to understanding their rights and accessing support.
We’re delighted to be working with Breast Cancer Now on a new policy research project exploring the future implementation of risk-stratified breast screening in the NHS. Breast Cancer Now is the UK’s leading breast cancer charity, combining world-class research, campaigning and support services to improve outcomes for people affected by breast cancer. Alongside funding…
