Beyond the quick fix: How workforce data can drive deeper organizational problem-solving

Using data to help identify systemic issues in an organization’s workforce, workplace, and work

Whether by applying analytics, machine learning, or human judgment, sense-making is what helps organizations convert data into insights, decisions, and actions that have the power to make improvements, from innovation to agility to worker performance and well-being. But without the right context, even simple measurements can undermine efforts to turn data into value.

Below, we explore several examples of how organizations might be limiting their analysis to the surface level and how deeper analysis can reveal systemic issues and lead to opportunities for transformation.

In the workforce

• What the data shows: A talent shortage
• What the data might be revealing: An opportunity to adjust hiring systems to reveal hidden opportunities

According to a 2023 survey conducted by ManpowerGroup, 77% of employers globally are experiencing difficulty in filling roles—a 17-year high.⁵ But organizations may have less of a talent shortage than they think. If data is highlighting a talent shortage in an organization, consider whether the underlying problem might be that hiring systems and processes are ignoring the hidden and often overlooked talent at their fingertips. Automated hiring systems and AI screening tools, for example, may be screening out qualified talent, creating “hidden workers”—individuals who may have been overlooked due to experience or resume gaps, for example.⁶ Accessing diverse and hidden talent may require adjusting recruiting and retention policies to focus on candidates’ adjacent skills, cultural fit, and team fit instead of only the education and experience listed on resumes.

When one telecommunications company, for example, needed to hire employees with machine learning skills, it didn’t search for candidates who held degrees or had experience in machine learning or AI; instead, it analyzed the profiles of thousands of workers who identified themselves as machine learning experts to interpret the aggregation of skills, experience, and pathways relating to these workers’ machine learning skill development. The company then created algorithms to search for and hire based on those new metrics—increasing the talent pool by at least three times what the company had estimated. After hiring workers with adjacent skills, the company quickly built on that foundation to train the hired workers with the specific machine learning skills they needed.⁷

Another issue that may be manifesting as talent shortage is lack of worker mobility within the organization. Data on transferable or adjacent skills, interests, and activities of workers can be tapped to match employees with new opportunities, projects, learning, or roles. This can also help employees understand which skills could make them more marketable and employable as the organization evolves. The same telecommunications company that adjusted its recruiting practices also recognized the importance of expanding its internal talent pool and developed technology to enable its workers to assess their fit for different roles by comparing their skills profile against a list of skills required to qualify for the roles.⁸

• What the data shows: Diversity, equity, and inclusion (DEI) efforts aren’t moving the needle.
• What the data might be revealing: Representation metrics aren’t addressing the “equity” and “inclusion” parts of the equation.

Organizations are under pressure from investors, customers, and employees alike to make progress on DEI initiatives. According to a 2023 study from Harvard Business Review, while 97% of human resource leaders believe that their organization has made changes that improved DEI, just 37% of employees strongly agree.⁹ If data shows that organizational DEI efforts aren’t impactful, consider whether the underlying problem might be a need for more focus on equity and inclusion. Representation metrics can be easy to obtain, but they may not address the issues that underpin DEI efforts. New advances in technology, however, may be able to provide better data to help measure the impact of the underlying issues.

For example, implicit biases across the workforce may be seeping into communications and could negatively impact inclusion metrics. Organizations that have voluntary data sharing agreements or employee opt-in measures in place may be able to identify the biases, gaps in inclusion efforts, and toxic pockets of conversation that undermine broader DEI and belonging efforts through text analysis of work, communications, and performance feedback platforms. Some AI tools can also provide personalized feedback on factors such as tone and language choices in emails to help limit biases or microaggressions.

Data that shows a lack of progress in DEI efforts may also be revealing an issue related to people’s interactions with leaders, mentors, and resource or affinity groups. Organizational network analysis (ONA)—a way to measure and graph connections and patterns of collaboration between people within and across organizations—performed on this data could be helpful in identifying inequities. It could also reveal the degree of belonging workers feel.

In one example of how ONA can be used to map the impact of these connections, a large service-based organization wanted to better understand whether it was achieving true gender diversity within its teams. On the surface, gender diversity appeared balanced, with approximately 44% of the teams made up of women. But how were the teams functioning? Using ONA, the organization was able to map its networks by gender, revealing that while some gender clustering was present, the center of the network was solidly gender-mixed, confirming that diverse perspectives were represented at the core of the organization. This is the kind of deep analysis organizations can use to extend beyond traditional representation metrics and gain a more nuanced understanding of diversity.¹⁰

In the workplace

• What the data shows: Workers aren’t on their computers as often as leadership thinks they should be to be productive.
• What the data might be revealing: Outdated productivity metrics are an inaccurate proxy for human performance.

Thanks in large part to the COVID-19 pandemic, remote and hybrid work have transitioned from an occasional option for a handful of individual employees to the new normal across many industries. But a growing sense of “productivity paranoia”—a term coined by Microsoft to describe the fact that only 12% of leaders are confident that employees working remotely are being productive¹¹—is casting a shadow on workplace data. If data is bringing employee productivity into question, one problem may be that organizations aren’t tracking the right productivity metrics.

With advances in technology enabling more knowledge work, traditional productivity metrics may not account for the increase in the “invisible” work many workers are engaging in as organizations shift to more open-ended work models and structure roles and responsibilities around problems to solve rather than a set of repeatable tasks. In a Deloitte study, a majority of human resources leaders (79%) said that worker roles are evolving into something broader and more integrated, often embracing adjacent job functions, and a majority of workers agreed: Seventy-one percent surveyed said they are already performing work outside of their stated scope of job responsibility.¹² (Read more on productivity metrics in the Deloitte Insights article, Outcomes over outputs.)

Among in-office workers, outdated organizational productivity metrics may be limiting leaders’ perspectives on what really matters in boosting human performance in the physical workplace. Data on how individual workers move, work, and interact throughout the day, for example, can greatly improve understanding of how to optimize workplace design to help improve human performance. A major energy company recently used workplace badge data to analyze where and how different groups were interacting while planning an office relocation. It found that as cross-functional teams became more dispersed, they had fewer informal interactions and instead relied too heavily on occasional formal meetings. The company used this finding to plan the location of team members during relocation to create more informal connection opportunities, improving workflow efficiency by 5.3%.¹³

• What the data shows: High turnover and burnout rates
• What the data might be revealing: Lagging/reactive workforce risk indicators aren’t surfacing problems early enough to mitigate them.

With employee burnout on the rise globally—42% report that they feel burned out at work, the highest rate since 2021¹⁴—organizations are right to be concerned about indicators of workforce risk. Employees experiencing burnout are 3.4 times as likely to look for a new job in the coming year and twice as likely to feel disconnected from their teammates, leaders, and the company’s values as those who aren’t.¹⁵

When data seems to be identifying increasing rates of turnover, burnout, compliance challenges, reputational damage, and DEI issues, that may be a sign that the indicators the organization is using to track workforce risk are reactive rather than proactive: They highlight a problem retroactively at the organizational level rather than helping surface it earlier at the worker level where it can be addressed before it escalates.

One way leaders can use work and workforce data to help stay ahead of workforce risk issues is by analyzing organizational culture for indicators that a problem may be brewing. Natural language processing, text analysis, computational linguistics, and audio analysis of business meetings and communications can help identify and quantify potential declines in sentiment across the workforce. The data can be assembled in a dashboard and used to help inform organizational strategies.

One global technology organization, for example, had long relied on the intuition of its managers and information from its human resources (HR) department to assess turnover intentions among employees. But the company’s chief HR officer realized that this information could be better collected and assessed algorithmically. Under her guidance, the organization created a machine learning algorithm that analyzes multiple variables and millions of data points to identify employees at risk of leaving the organization. This algorithm then provides managers with recommendations for conversations that may help retain these at-risk employees.¹⁶

In the work

• What the data shows: Organizational productivity is lagging.
• What the data might be revealing: The interactions and processes between functional domains need attention.

Many organizations are structured around capabilities, with different functional areas operating in the best interest of their own priorities. This often leads to organizational information silos or competition for resources. For example, production and maintenance departments may find themselves at odds over when machines should be taken offline for routine maintenance. Or marketing and media departments may be collecting the same information using different software programs, creating unnecessarily repetitive work.

When productivity data is trending downward, one approach is to try to optimize the organization’s individual functions. But the data may be revealing a bigger problem to be addressed: how functional domains are working together. Focusing on individual efficiencies without a unified view of how the organization is functioning across domains can result in failure to solve underlying problems that may be at the core of lagging productivity.

Organizations can use network data to identify bottlenecks or spots where more interaction between functional areas can improve a process and to accordingly adjust their communication patterns. Communications or work application data can help improve work and collaboration patterns within or across functional groups, potentially reducing time to market.

Work and workforce data can also be used to inform workplace design that optimizes collaboration between various organizational functions or activities. For example, a tier 1 automotive supplier used AI-powered video analytics to provide detailed visibility into the activities that workers perform in the factory. The analytics revealed a slowdown in stations whose configurations inhibited workers. Consequently, the organization reconfigured these stations and improved line balancing, thereby reducing overall cycle time and boosting overall productivity.¹⁷

• What the data shows: Organizational processes are inefficient.
• What the data might be revealing: Organizational processes don’t reflect the way work really gets done.

An org chart or process map may tell the official story about how an organization collaborates, but it’s often not the full story. When the data seems to indicate that an organization’s processes or charts are lacking, there is likely more to the story about how work gets done—and who’s doing it—than the processes or structures mapped out seem to be revealing.

Rather than trying to work out organizational processes on paper, new sources of data can help leaders understand hidden work patterns and identify who’s working on what and how they’re doing it. ONA can highlight informal structures and roles to help inform a new organization design or help leaders spot collaborations that need nurturing. ONA can also identify hidden gems in the workforce—teams with strong connections or energizer employees who can infuse new ideas into the organization. Mining enterprise transaction systems, workflow data, or video analytics can help identify the root causes of issues, the tasks underlying each job, and the strongest opportunities for improvements.

When a computer and technology company, for example, wanted to better understand how its workers connect and communicate beyond its org charts, it leveraged aggregated work metadata—deidentified emails and communications data from collaborative tools, calls, and meetings—to chart how collaboration was actually happening. What they learned was that the strongest teams were the ones that formed and disbanded quickly around projects, regardless of location, title, or department. In one instance, communications data revealed an intense collaboration between several labs across different locations. Despite geographical differences, the distributed teams actually functioned as a single unit, allowing leaders to successfully fine-tune work strategies and processes.¹⁸

Understanding data in context

As organizations seek to move beyond quick fixes and use data to help solve deeper, more complex problems, how can they ensure that they’re getting the full picture of what the data is telling them? There are three actions leaders can consider taking to help ensure they are not missing important context in their data analysis.

• Bring data from different domains and sources together for analysis. To help avoid oversimplification of problems and ensure accurate cause-and-effect reasoning, consider analyzing data from across domains. Consider the example of monitoring truck driver fatigue again. While the data collected led to actionable outcomes, they weren’t necessarily the right outcomes. Focusing on the individual worker instead of analyzing data from across the ecosystem oversimplified the problem and overlooked the organizational and structural problems at the root.
• Make sure you’re measuring what you should—not just what you can. It can be easy to become more enticed by the data and numbers than the actual goal. Organizations should always ask themselves: just because it can be measured, does it really need to be—and if so, why? Part of responsible data collection is ensuring that the data collected reflects the metrics that are most important to an organization’s goals and objectives.
• Identify potential biases in data collection algorithms. Organizations should ensure that data collection and use are fair, equitable, and ethical. New advances in technology can help. Vendors now offer testing tools designed to take a continuous, automated approach to testing against biases. Others offer monitoring and governance solutions meant to monitor, measure, and improve machine learning to help ensure that models are delivering accurate, transparent, and fair results. Other advances, such as supplementing human data with synthesized data created by software, can help enable AI to fill the gaps of “edge cases” that haven’t happened in the real world and include more inclusive and less biased datasets.

Data may highlight the symptoms of an organizational issue, but treating these symptoms may only offer a temporary solution. Organizations should be willing to pursue a deeper analysis and allow data to surface organizational changes that may be more challenging or difficult to implement. In the end, going beyond the quick fix can help organizations achieve the kind of transformation that moves them closer to their goals and outcomes.