Report Finds Data Quality Issues Affect R&D Decision-Making at 63% of Organizations

Despite widespread confidence in the quality of their research data, many corporate R&D organizations continue to make important decisions using incomplete, outdated or inaccurate information, according to a new global study from Elsevier. The company’s Corporate R&D of the Future report, based on a survey of more than 1,000 R&D leaders, managers and researchers at research-driven organizations, highlights a disconnect between confidence in available evidence and the reliability of the information used to guide development decisions.

While 76% of respondents expressed confidence in their organization’s data quality, 63% acknowledged that their company had made an important R&D decision using information subsequently found to be inaccurate, incomplete or outdated. Additionally, 51% said their organization had faced situations in which decisions had to be made before all relevant evidence was available.

The findings also point to growing expectations for artificial intelligence (AI) in research and development. Although 81% of respondents believe AI will fundamentally change R&D decision-making over the next five years, fewer than half (47%) report that the technology is currently used regularly within their organization’s R&D operations. Another 40% said their organization is piloting AI applications.

Stuart Whayman, president of Corporate Markets at Elsevier, says the findings underscore the distinction between confidence in information and its actual accuracy. He noted that pressure to accelerate decisions can leave research teams relying on evidence that has not been fully validated or is no longer current. As AI adoption expands, he cautioned, the technology could magnify existing data weaknesses unless organizations improve how information is collected, connected and verified.

The survey found that 78% of respondents have access to the external evidence their organizations need. However, more than half of those with access (56%) still encounter obstacles when trying to use that information effectively. The most frequently reported challenges were fragmented information sources, cited by 42% of respondents, difficulty finding and synthesizing information (32%) and a lack of awareness about what evidence is available (30%).

Access also varied by geography. Respondents in Asia-Pacific (APAC) reported the highest availability of third-party evidence at 90%, compared with 77% in Europe and 73% in North America.

Although organizational adoption is still developing, individual researchers are already applying AI across a range of R&D activities. Data analysis was the most frequently cited application, at 71%, followed by literature searches and competitive intelligence, both at 70%.

Confidence in AI’s potential is closely tied to the quality of the information underpinning it. A total of 86% of respondents believe organizations combining AI with reliable, high-quality evidence will have an advantage over those that do not.

The report also identified several factors that could encourage broader AI adoption. Regulatory compliance was cited by 55% of respondents, while ease of use and integration with existing workflows and demonstrated factual accuracy and safety each received 54%. Regulatory considerations were particularly prominent among European respondents, 64% of whom said compliance with applicable laws and regulations would make them more willing to use AI.

Despite expectations for AI-driven improvements, implementing the technology across R&D operations remains a concern. Nearly half of respondents (48%) anticipate significant challenges in scaling AI within their organizations. The leading priorities identified for the next five years include developing AI capabilities across the workforce (65%), translating data and evidence into actionable insights (62%) and adapting AI applications to organizational knowledge and existing workflows (59%).

Concerns about implementation were most pronounced in North America, where 56% of respondents anticipated significant challenges in scaling AI. The corresponding figures were 50% in Europe and 34% in APAC.

The survey also examined the types of AI tools researchers use. For literature searches, general-purpose platforms such as ChatGPT and Copilot were more widely used than specialized scientific AI applications, with 48% reporting use of general-purpose tools compared with 23% using specialist solutions. For data analysis and modeling, specialized AI tools were the more common choice.

Overall, the findings suggest that the next phase of AI adoption in corporate R&D will depend on more than access to new technology. Organizations will also need to address fragmented information, improve data quality and develop the skills and infrastructure required to turn reliable evidence into informed research and development decisions.

Learn more at www.elsevier.com.