Why Diagnosing Conflict Matters More Than Resolving It
From the Conflict Management collection
The average employee spends 2.8 hours a week dealing with conflict, according to a 2008 CPP study of 5,000 full-time employees across nine countries. In the United States alone, that amounts to roughly $359 billion in paid hours a year. Most organizations treat this as a cost to be reduced — a problem of difficult people and poor communication skills. The research on team performance suggests a different reading: conflict is an information system, and the question is whether an organization can extract the signal before the relationship damage drowns it out.
The first step is classification. Karen Jehn's research in the 1990s separated three kinds of conflict: task conflict (disagreement about the content of the work), relationship conflict (interpersonal friction — annoyance, tension, animosity), and process conflict (disputes over logistics, delegation and procedure). Corinne Bendersky and Nicholas Hays added a fourth in 2012: status conflict, contests over relative standing in the group — whose expertise counts, who gets credit, who speaks last. The same argument in the same meeting can be any of the four, and each behaves differently.
The popular version of this research holds that task conflict is healthy and relationship conflict is harmful. The evidence is more conditional. De Dreu and Weingart's 2003 meta-analysis found both types associated with lower team performance. De Wit, Greer and Jehn's 2012 meta-analysis — 116 studies, 8,880 groups — refined the finding: relationship and process conflict are reliably damaging, while task conflict ranges from harmful to helpful depending on whether it stays separated from relationship conflict. In most teams, it does not stay separated. Simons and Peterson (2000) showed that trust governs the conversion: in low-trust teams, criticism of an idea is read as criticism of its owner, and task conflict becomes relationship conflict through misattribution.
The implication is precise. The skill is not having more conflict or less conflict. It is having the right kind and keeping the kinds apart — which requires naming the type before choosing a response.
A second diagnostic layer concerns the source. Much workplace conflict is structural rather than personal. Lawrence and Lorsch demonstrated in 1967 that departments facing different environments develop different goals, time horizons and working styles. The resulting cross-functional friction is a normal by-product of specialization — it would recur with different people in the same seats. Steven Kerr's 1975 paper documented how reward systems set colleagues against each other. Rogers and Blenko (HBR, 2006) traced recurring disputes to ambiguous decision rights: two people who each reasonably believe the decision is theirs.
Checking the structure before assessing the personalities is the diagnostic move most managers skip. If the conflict comes from the incentive design, a conversation about interpersonal styles will be irrelevant. If it comes from ambiguous decision rights, the resolution is to clarify ownership, and the personalities are downstream. If it originates in status contests — which rarely announce themselves, appearing instead as disagreements about the work — resolving the stated issue leaves the underlying dynamic untouched.
Once the type and source are identified, the response can be chosen rather than defaulted into. Thomas and Kilmann mapped five conflict responses in 1974 on two axes — assertiveness and cooperativeness: competing, accommodating, avoiding, compromising, collaborating. Each response is correct in specific conditions. Competing fits when speed and stakes are high and one party has better information. Accommodating fits when the issue matters more to the other side. Avoiding is legitimate when the issue is trivial, tempers need time, or the cost of engaging exceeds the benefit. The skill is matching the response to the situation. The common pattern is relying on one or two responses by habit, regardless of what the situation calls for.
Ury, Brett and Goldberg (1988) added another frame. Studying wildcat strikes in coal mines, they distinguished three routes to resolution: reconciling underlying interests, determining who is right under some standard, or determining who is more powerful. Interest-based resolution generally costs less and lasts longer, but rights and power serve as the necessary backstop when interests cannot be reconciled or one party is acting in bad faith.
The individual's repertoire matters, but it operates within a system. Pondy (1967) described conflict as an episode with stages — latent, perceived, felt, manifest, aftermath — and the aftermath stage determines what comes next. Every conflict leaves residue: either a stronger basis for cooperation or the latent conditions for the next episode. Jehn and Mannix (2001) found that high-performing groups showed consistently low relationship conflict and a pattern of task conflict that was managed without migrating into personal friction.
At the organizational level, some teams and companies build disagreement into how they decide. Eisenhardt, Kahwajy and Bourgeois (HBR, 1997) studied twelve top management teams and found six tactics shared by those that argued hard and stayed cohesive — including working with more data, generating multiple alternatives, and maintaining common goals. Each tactic works by keeping the argument about the issue rather than letting it migrate to the people. Pixar's Braintrust, Amazon's disagree-and-commit protocol, and Bridgewater's idea meritocracy each institutionalize disagreement, and each carries costs that the organizations' own accounts tend to understate.
The diagnostic habit — classifying the type, checking the source, choosing the response — catches errors that are otherwise invisible. A manager who treats every conflict as a communication problem will solve some and entrench others. A manager who checks whether the conflict is task or relationship, structural or personal, and whether competing or collaborating or avoiding fits the situation, will waste less time on conversations that were never going to help and spend more time on the structural changes that prevent recurrence.
The same diagnostic extends beyond individual episodes. Organizations that build disagreement into how they decide — through data-rich debate, authentic dissent mechanisms, or disagree-and-commit protocols — get better decisions, but each design carries costs that must be weighed honestly.
Conflict is an information system. A disagreement signals that two people hold different data, different goals, or different claims on the same resource — and the organization benefits from learning which. The leverage lies in the conversion: turning a conflict about people back into a conflict about the work, and turning a conflict about the work into a decision.