If You Can't Measure Productivity in a Large Organization at the Individual or Department Level, What Do You Do?

It is nearly impossible to granularly know which employees and departments in any large organization really contribute to firm outcomes.

Since we cannot really measure such contributions, brute force layoffs might be the only way to find out how efficient any large organization really can be, as hard as that is on workforces.

The “attribution problem” in large organizations will be familiar to anybody who has had to assess rewards based on actual firm results when no single department actually can claim credit for most financial outcomes.

In large firms, value is rarely created by a single employee or department working in isolation. Product launches, involve research and development,, engineering, marketing, logistics, sales, legal compliance, and customer support. 

When the launch succeeds, how much credit belongs to each? When it fails, who bears responsibility?

The assessment problems (rewards, for example) are significant:

• Incentive misalignment. If individual contributions cannot be cleanly attributed, performance-based pay systems break down as heroes and free riders alike are in the same category

• Resource misallocation. Budget committees and executives tend to reward departments that can demonstrate value legibly (sales, manufacturing) while others are hard to evaluate (compliance, infrastructure, knowledge management)

• Strategic distortion. Firms may over-invest in activities whose productivity is visible and under-invest in activities whose productivity only shows up indirectly or with long lags. (research, culture-building, and organizational learning).

Small firms can rely on direct observation and social norms to attribute effort. But large entities have interdependencies that increase nonlinearly while political dynamics increasingly shape who gets credit. 

The result is that large firms have measurement costs that scale faster than the firm itself.

Study / Work	Authors	Year	Core Argument	Key Finding on Attribution
Production, Information Costs, and Economic Organization	Alchian & Demsetz	1972	The firm exists to monitor team production where individual contributions are inseparable	Joint production makes metering individual inputs prohibitively costly; hierarchy is an imperfect but necessary solution
Moral Hazard in Teams	Holmström	1982	Optimal contracts under joint output are necessarily inefficient	No budget-balanced contract can achieve first-best effort when output is jointly produced; someone must absorb the residual
Pay and Organizational Performance	Milgrom & Roberts	1988	Influence activities consume resources when attribution is ambiguous	Workers invest in lobbying for credit; firms respond with rigid rules that reduce gaming but also reduce flexibility
The Balanced Scorecard	Kaplan & Norton	1992	Firms rely too heavily on financial metrics that obscure true value drivers	Single-metric systems systematically misattribute productivity; multidimensional scorecards partially address but don't solve the problem
Knowledge-Creating Company	Nonaka & Takeuchi	1995	Knowledge is produced collectively and tacitly across departments	Tacit knowledge creation is inherently non-attributable; firms that try to credit individuals destroy the collaborative conditions needed for it
Measuring and Managing Performance in Organizations	Austin	1996	Measurement systems alter behavior in dysfunctional ways	The more precisely you measure a subset of outputs, the more workers optimize for measured outcomes at the expense of unmeasured ones (a formalization of Goodhart's Law)
Why Firms Differ	Nelson	1991	Organizational routines are the unit of productive capability, not individuals	Productivity resides in collective routines that cannot be decomposed into individual contributions without destroying the routine itself
Incentives vs. Control in Organizations	Prendergast	1999	A comprehensive survey of incentive design under imperfect attribution	Subjective performance evaluation introduces favoritism; objective metrics introduce gaming; no system cleanly resolves joint production
The Knowing-Doing Gap	Pfeffer & Sutton	2000	Firms systematically misidentify what drives performance	Attribution errors lead firms to copy visible practices (structure, perks) while ignoring invisible ones (culture, trust) that actually drive productivity
Social Capital and Value Creation	Tsai & Ghoshal	1998	Inter-unit resource exchange creates value that no unit can claim individually	Network ties between departments generate productivity gains that appear in no unit's performance metrics
Team Incentives and Worker Heterogeneity	Hamilton, Nickerson & Owan	2003	Empirical study of a garment manufacturer's shift to team production	High-ability workers joined teams despite lower individual pay, suggesting team attribution problems are partially self-selected but still persist
Paying for Performance: A History of Compensation	Milkovich & Wigdor	1991	Review of compensation system effectiveness across firm sizes	Attribution ambiguity grows sharply with firm size; large firms revert to tenure-based pay as a second-best solution
Multitasking and Soft Information	Baker	1992	When agents perform multiple tasks, incentivizing one distorts effort on others	Firms with complex interdependent workflows face a fundamental tradeoff: reward measurable tasks and lose unmeasured effort
Corporate Budgeting is Broken	Jensen	2001	Budget processes in large firms produce systematic gaming	Attribution uncertainty incentivizes sandbagging and internal negotiation over actual productivity improvement
Hidden in Plain Sight: Productivity and the Org Chart	Weil	2014	Fissured workplace structures (outsourcing, franchising) obscure which entity drives productivity	As production is fragmented across legal entities, attribution becomes structurally impossible even in principle

The point is that in large organizations,  it will be difficult to impossible to know with certainty how much individuals and whole departments actually contribute to firm outcomes, with or without AI. 

Few of us are perhaps happy that significant layoffs are happening, or will happen, in many large firms in many industries. On the other hand, none of us can actually “prove” that all those workers, in all those areas, are actually contributing to outcomes we can measure. 

This might be one of those instances where all we can do is act, see what happens, and then adjust if necessary.

If we assume that, for most firms, higher costs are not a good thing, then lower costs, with same or higher output, are preferable, generally speaking.

And if we cannot measure the impact of contributions, then the equivalent of pruning a plant might not be irresponsible.