The Gallup 'best friend at work' finding

The "best friend at work" question emerged from Gallup's multi-year effort to distill the conditions of great workplaces into a small number of survey items. Starting in the late 1980s, Gallup researchers — notably the consulting psychologist Donald Clifton, who later became Chairman of Gallup — conducted tens of thousands of interviews with employees across organizations to identify the experiential conditions that distinguished highly productive business units from average ones. The research method was empirical rather than theory-driven: Gallup asked workers to describe the best work environments they had experienced and analyzed the themes.

Friendship emerged repeatedly as a theme without being directly asked about. Workers describing excellent work environments consistently mentioned specific colleagues — people they trusted deeply, communicated honestly with, and genuinely enjoyed being around. The research team noticed that these descriptions sounded more like friendship than like professional collaboration. The Q12 instrument, finalized through meta-analysis in the late 1990s, included the best-friend item as the most direct operationalization of the relational condition workers were describing. Its inclusion in a business engagement instrument was deliberate provocation: an assertion that something so apparently personal belonged in the same analysis as customer satisfaction and profitability.

The Q12 is a twelve-item survey instrument validated through Gallup's meta-analysis of data from 2.7 million workers across 27 countries and 100,000 business units. Each item was selected based on its ability to predict business outcomes — engagement, performance, retention, customer satisfaction, safety — at the business-unit level. The items range from immediate management quality ("My supervisor, or someone at work, seems to care about me as a person") to growth and recognition to the best-friend item.

The meta-analysis supporting Q12 — published formally in Harter, Schmidt, and Hayes (2002) in the Journal of Applied Psychology — found that business units in the top quartile of Q12 engagement outperformed bottom-quartile units by 22 percent in profitability, 21 percent in productivity, 10 percentage points in customer satisfaction, and 37 percent in absenteeism. The best-friend item contributes to the composite Q12 score; Gallup has not published item-level effect sizes for proprietary reasons, but subsequent academic replication has confirmed that friendship-related items predict engagement over and above the composite of other items.

The most serious methodological critique of the best-friend finding concerns causal direction. The cross-sectional survey design of most Q12 applications cannot distinguish between two plausible accounts: (1) having a best friend at work produces engagement and performance, or (2) high-performing, well-managed business units create conditions in which friendship can form, and the best-friend item is capturing organizational quality rather than friendship per se.

Gallup's longitudinal data — accumulated over decades of repeat measurement in stable organizations — provides some traction on this question. Business units that increase their best-friend scores over time show subsequent improvements in performance outcomes at rates greater than units that improve other Q12 items. This temporal ordering is consistent with a causal contribution of friendship to performance, though it does not eliminate the possibility of confounded improvement (management quality improving both the friendship conditions and performance simultaneously).

The academic literature provides complementary evidence. Experimental and quasi-experimental studies of team formation, onboarding, and social network development consistently find that relationship formation precedes and predicts performance improvement rather than the reverse. The contact hypothesis literature, from Allport through contemporary field experiments, shows that structured positive contact produces genuine relationship formation, which then predicts cooperation and collaboration outcomes. These findings are consistent with the causal interpretation of the best-friend finding.

The original Gallup datasets showed consistent gender variation: the best-friend item was a stronger predictor of engagement for women than for men. This finding has proven robust across Gallup's subsequent surveys and has been replicated in independent organizational studies. The interpretation is disputed, with several competing accounts.

One account emphasizes socialization: women have historically been socialized to invest more heavily in dyadic close friendships, to derive more identity and support from them, and to be more explicit about their relational needs at work. On this account, women's engagement is more friendship-contingent because friendship is more central to their self-concept and social functioning. Another account points to organizational power dynamics: in organizations where women are systematically excluded from informal male networks of sponsorship and information, a close friendship with a colleague who provides access, advocacy, and information is worth more structurally — hence a larger engagement effect. A third account notes that women tend to be more underrepresented in leadership roles, making peer relationships a more important source of support, validation, and career resources.

These accounts are not mutually exclusive and each has supporting evidence. The gender variation in the best-friend effect is substantively important because it suggests that organizational investments in supporting friendship formation will have heterogeneous returns — and that failure to support cross-demographic friendship formation will tend to benefit majority-group employees disproportionately.

The best-friend effect also varies by role type and industry. Gallup's data consistently shows stronger effects in customer-facing roles — retail, healthcare, hospitality, call centers — than in technical or autonomous individual-contributor roles. This is theoretically consistent: friendship predicts performance through the mechanisms of trust, psychological safety, and discretionary communication, which matter more in roles where coordination and customer interaction are central to performance.

In healthcare settings, the effect is particularly robust. Research on nurse team cohesion and physician collegial relationships consistently shows that close collegial bonds predict patient safety outcomes, including near-miss reporting, error detection, and communication quality during handoffs. These are high-stakes environments where coordination failures kill people, and where friendship — or its proximate kin, genuine collegial trust — is a meaningful predictor of the quality of communication that prevents those failures. The best-friend finding in healthcare is not about job satisfaction; it is about whether patients receive safe care.

The specific language of the Q12 item — "best friend" rather than "close colleague," "trusted coworker," or "good friend" — was deliberate and has been controversial. Gallup chose the strongest possible relational language because their qualitative interviews showed that the most productive work environments were characterized by relationships that workers themselves described using that language. Using weaker language might capture a broader range of relationships but would miss the intensity of the specific condition that predicted performance.

The controversy around the label reflects real cultural variation. In some organizational cultures, calling a colleague a "best friend" is entirely natural. In others — particularly in more formal or hierarchical professional cultures in Northern Europe and East Asia — the language is jarring in a professional context, which may cause respondents to answer the item based on cultural discomfort with the label rather than the underlying relational reality. Gallup has developed adapted language for international surveys; the finding holds across most cultural contexts studied, suggesting the underlying phenomenon is not culturally specific even if the label is.

The history of organizational resistance to the best-friend finding is itself instructive about corporate culture. When Gallup began including the item in engagement surveys in the late 1990s, multiple major corporate clients objected — arguing that encouraging friendship at work was unprofessional, that it created accountability problems, that it exposed the organization to discrimination risk if friendship networks formed along demographic lines, and that it was simply inappropriate for a business measurement instrument.

These objections were not entirely without basis. Close friendships can create accountability problems; friendship networks do tend to form along demographic lines. But the objections functioned mainly as rationalizations for the prior commitment to a view of the workplace as a purely transactional space where personal relationships were beside the point. The data did not support that view. What the data showed was that organizations that maintained the purely transactional fiction — pretending that their employees' human need for connection was irrelevant to organizational outcomes — were leaving significant performance on the table. The objections to the item were often most vigorous from the organizations whose engagement scores were lowest.

Gallup's engagement tracking between 2020 and 2023 documented a measurable decline in the best-friend item, particularly among younger workers and remote employees. The pattern was distinct from the general engagement trend: some engagement items held relatively stable during the remote-work transition while the best-friend item declined more sharply. This is consistent with the mechanism: the best-friend condition depends on the accumulated social history of shared experience, and remote work interrupted both the maintenance of existing friendship and the formation of new ones.

The decline has practical implications. If the best-friend condition predicts engagement, and engagement predicts performance, then the friendship erosion documented since 2020 represents a latent performance risk that does not necessarily show up immediately in output metrics — discretionary effort declines gradually, trust erodes slowly, the willingness to speak up about problems fades over time. Organizations that experienced the remote transition in 2020 with strong pre-existing social networks are now, five years on, managing with thinner networks than they had, and many have not taken deliberate steps to rebuild them.

The best-friend finding connects to a related literature on mentorship and sponsorship in organizations. Formal mentorship programs — designed to produce the career guidance and organizational knowledge transfer that informal mentorship provides through relationship — consistently underperform because they try to replicate the outputs of relationship without creating the relationship itself. The mentee and mentor meet on schedule, complete their assigned discussions, and often form no genuine connection. The programs produce credential and activity data; they rarely produce the sustained mutual investment that genuine mentorship involves.

The reason mentorship works when it works is the same reason friendship predicts engagement: it creates a relationship in which someone with more experience and organizational power has a genuine stake in the other person's development. That stake produces the advocacy, honest feedback, and informal information sharing that formal mentorship programs try to mandate. Organizations that understand the best-friend finding correctly should recognize that their formal mentorship programs are attempts to engineer a relational condition; the question is whether the engineering is producing the condition or merely its simulation.

The management literature on translating the best-friend finding into practice is less developed than the measurement literature, for understandable reasons: you cannot mandate friendship, and attempts to do so through forced team-building produce the opposite of the genuine connection the item captures. The practical implications are more structural and environmental.

Physical space design that creates informal gathering opportunities — shared kitchens, comfortable common areas, proximity between people with interdependent work — supports the repeated incidental contact that friendship formation requires. Onboarding practices that create multiple informal interactions between new employees and teams in the first 90 days — not structured networking, but unstructured shared meals, walks, and casual conversations — accelerate friendship formation among new employees who are at greatest risk of low engagement. Managerial practices that treat social time as legitimate work time rather than distraction from work change the cultural accounting of friendship. None of these interventions create friendships; they create conditions under which friendships can form.

A recurring critique of the best-friend finding from outside organizational psychology is that it instrumentalizes friendship — treating what should be a freely chosen, intrinsically valuable human relationship as a productivity mechanism. On this view, asking "do you have a best friend at work?" as a means of predicting profitability is a kind of category error: it misunderstands what friendship is and colonizes an intimate domain with commercial purpose.

The critique has philosophical substance. Friendship is not, in its primary sense, a productivity tool. The Aristotelian account of virtue friendship — the highest form, based on genuine care for the other's good — is explicitly opposed to the utility-based friendship that the organizational account risks reducing all workplace connection to. But the critique does not fully engage with the finding's actual practical implications. The finding does not recommend that workers choose friends based on productivity returns; it recommends that organizations stop treating human connection as irrelevant to organizational outcomes and start creating the conditions under which genuine connection can form. The difference between creating conditions for genuine friendship and engineering a friendship substitute for productivity purposes matters, and the best organizational applications of the finding honor that difference.

The best-friend finding's long-term significance lies less in the specific item and more in what it established: that the quality of human relationships at work is a legitimate variable in organizational performance analysis, not a soft confound to be controlled away. Before Gallup, the dominant measurement paradigm in organizational psychology treated relationships as background conditions — things that could facilitate or impede performance but were not themselves performance inputs worth measuring directly. After Gallup, the evidence base for relational variables in organizational analysis has expanded substantially, with psychological safety, team cohesion, trust, and connection all established as organizational performance predictors in their own right.

The long arc of this research program is a quiet refutation of the Taylorist premise that the worker is an individual input to be optimized in isolation. The worker is a relational creature who performs better, stays longer, cares more, and contributes more freely when embedded in genuine connection with the people around them. This was probably obvious to anyone who had ever worked closely with people they genuinely liked. The contribution of Gallup's finding was to make it measurable, replicable, and businesscase-legible — transforming an obvious human truth into a defensible organizational claim.

---

1. Gallup. State of the Global Workplace: 2023 Report. Washington, DC: Gallup Press, 2023.

2. Harter, James K., Frank L. Schmidt, and Theodore L. Hayes. "Business-Unit-Level Relationship Between Employee Satisfaction, Employee Engagement, and Business Outcomes: A Meta-Analysis." Journal of Applied Psychology 87, no. 2 (2002): 268–279.

3. Rath, Tom, and James Harter. Wellbeing: The Five Essential Elements. New York: Gallup Press, 2010.

4. Rath, Tom. Vital Friends: The People You Can't Afford to Live Without. New York: Gallup Press, 2006.

5. Edmondson, Amy C. "Psychological Safety and Learning Behavior in Work Teams." Administrative Science Quarterly 44, no. 2 (1999): 350–383.

6. Hewlett, Sylvia Ann, Melinda Marshall, and Laura Sherbin. "The Sponsor Effect: Breaking Through the Last Glass Ceiling." Harvard Business Review Research Report, December 2010.

7. Gittell, Jody Hoffer. High Performance Healthcare: Using the Power of Relationships to Achieve Quality, Efficiency and Resilience. New York: McGraw-Hill, 2009.

8. Gallup. "What Is Employee Engagement and How Do You Improve It?" Gallup Workplace, 2023. https://www.gallup.com/workplace/285674/improve-employee-engagement-workplace.aspx.

9. Levin, Daniel Z., and Rob Cross. "The Strength of Weak Ties You Can Trust: The Mediating Role of Trust in Effective Knowledge Transfer." Management Science 50, no. 11 (2004): 1477–1490.

10. McPherson, Miller, Lynn Smith-Lovin, and James M. Cook. "Birds of a Feather: Homophily in Social Networks." Annual Review of Sociology 27 (2001): 415–444.

11. Microsoft. "Hybrid Work Is Just Work. Are We Doing It Wrong?" Microsoft Work Trend Index: Annual Report, September 2022.

12. Mayo, Elton. The Human Problems of an Industrial Civilization. New York: Macmillan, 1933.