Few decisions a project manager makes are as quietly consequential as the decision about how work will be distributed among the people available to perform it. Long before a schedule is published or a budget is approved, someone must decide whether the six tasks sitting in a given phase of the work breakdown structure will be handed to six workers operating in parallel or to one group of six workers moving through the tasks in sequence. The decision looks administrative. It is not. Klastorin and Mitchell (2021) observed that in earlier chapters the term resource is used generically to describe workers, organizations, or materials, but that when the discussion turns to project teams the term must be narrowed to individuals, workers, or full-time equivalents, because the composition and effectiveness of a project team can be an integral part of a project’s ultimate success or failure (p. 440). That narrowing matters. Once resources are understood as people rather than as interchangeable units of capacity, the arithmetic of allocation acquires a behavioral dimension that pure scheduling models do not capture.

This paper addresses a single question with several interlocking parts. When a project manager assigns work, how should that manager decide whether tasks are best performed by a group of workers or by individual workers acting independently? What is the benefit of each arrangement, and what is the trade-off that arises from choosing one over the other? The analysis proceeds in four parts. The first part develops the decision framework itself, working through the relationship between work content, resource allocation, and elapsed duration, and arriving at the concept of breakeven team efficiency as the quantitative pivot on which the decision turns. The second part sets out the benefits that accrue specifically to individual assignment. The third part sets out the benefits that accrue specifically to group assignment. The fourth part examines the trade-offs, which are more numerous and less symmetrical than the simple framing of the question would suggest, and which extend well beyond schedule arithmetic into cost, efficiency, motivation, and risk.

The approach taken here is both descriptive and prescriptive. It is descriptive in that it reports what the project management literature, and the empirical research that literature draws upon, actually says about how team and individual assignment affect project outcomes. It is prescriptive in that it attempts to convert those findings into guidance a working project manager could apply. That dual purpose imposes a discipline. Descriptive accuracy requires acknowledging that the evidence is mixed, that different studies point in different directions, and that the models used to formalize the decision rest on assumptions rarely satisfied in practice. Prescriptive usefulness requires committing to a position anyway, because a project manager who waits for unambiguous evidence before staffing a task will never staff anything. The reconciliation offered here is that the models are valuable less for the numbers they produce than for the structure of reasoning they impose.

The starting point for the decision is structural rather than preferential. Whether concurrency is even available as an option depends on the precedence network, not on the manager’s inclination. Klastorin and Mitchell (2021) framed the choice as one between completing tasks concurrently by assigning team members individually to tasks and completing tasks sequentially with team members assigned to tasks in groups (p. 456). Concurrency presupposes that at least two tasks are eligible to be worked at the same time, which is a property of the network’s topology. A strictly serial chain of tasks offers no opportunity for parallel individual work regardless of how many workers are available, because the second task cannot begin until the first is complete. In such a network the only meaningful question is how many workers to place on each task in turn. Conversely, a network with wide parallel branches presents the manager with a genuine choice, and it is in those networks that the analysis developed below has purchase.

The second precondition is the relationship between work content and duration. Klastorin and Mitchell (2021) invoked the expression t = w/R, in which task duration equals work content divided by allocated resources, as the basic mechanism by which additional workers compress elapsed time (p. 457). The expression carries an important implication that is easy to overlook. Work content is a property of the task; it does not change when the manager changes the staffing. Duration is a property of the assignment; it changes inversely with the resources applied. A task requiring eighty hours of work content requires eighty hours of work content whether it is performed by one person over two weeks or by two people over one week. What the manager is purchasing when adding a second worker is not less work but less elapsed time, and the price of that compression is the subject of most of what follows.

The textbook develops the choice through a deliberately small example that repays close attention. Klastorin and Mitchell (2021) posited a project of four tasks, in which Task A must precede Task C and Task B must precede Task D, worked by two cross-trained resources named Barb and Bob who can each perform any task with equal effectiveness (p. 456). Task durations for a single worker are discrete random variables with expected values of five, eight, ten, and seven weeks for tasks A, B, C, and D respectively, giving a total expected work content of thirty weeks for a single resource (pp. 456–457). The smallness of the example is a feature. With two workers and four tasks, the configurations can be enumerated exhaustively, which allows the trade-off to be observed directly rather than inferred from a simulation whose logic would be harder to inspect.

Consider first the team configuration. If Barb and Bob work every task together, they must process the tasks sequentially, because two people cannot be in two places at once. Klastorin and Mitchell (2021) assumed initially that a two-person team completes any task in half the time a single worker would require, which yields an expected project makespan of one half of thirty, or fifteen weeks (p. 457). The precedence network has effectively been dissolved. When every task is worked by the whole team in turn, the constraints that A precede C and B precede D impose no binding restriction beyond requiring a feasible ordering, and the makespan collapses to the simple sum of the halved durations. There is no critical path in any interesting sense, because every task lies on the only path there is.

The individual configuration behaves quite differently. Assigning Bob to tasks A and C and Barb to tasks B and D allows the two chains to advance simultaneously, and the project makespan becomes the length of the longer of the two chains rather than the sum of all four durations. Because the individual task durations are random variables, the makespan is itself a random variable. Klastorin and Mitchell (2021) enumerated all twenty-four possible outcomes, formed by three possible durations for Task A multiplied by two each for B, C, and D, and computed an expected makespan of 16.433 weeks with realizations ranging from as few as twelve weeks to as many as nineteen (pp. 457–458). Concurrent individual work is therefore roughly ten percent slower in expectation than teamed sequential work in this example, an increase of about 1.433 weeks (p. 457).

That result is counterintuitive enough to warrant an explanation, and the explanation is the crux of the decision. The team configuration wins in this example only because of an assumption smuggled in at the outset. Klastorin and Mitchell (2021) were explicit that in assuming Barb and Bob complete a task in half the time a single worker needs, they had implicitly assumed no loss of efficiency due to communication and coordination when the two work together (p. 457). Two people working as a team are assumed to deliver exactly two people’s worth of output. That assumption is generous, and in most real settings it is false. The moment the team delivers less than the arithmetic sum of its members’ individual capacities, the advantage begins to erode, and at some level of degradation it disappears entirely.

The point at which it disappears can be calculated, and the calculation is the single most useful analytical product of this section. Klastorin and Mitchell (2021) denoted team efficiency as Eff and solved the relation thirty multiplied by Eff equals 16.43, yielding a breakeven efficiency of 0.548 (p. 457). The interpretation is direct. When Barb and Bob working together can complete a task in no more than 54.8 percent of the time an individual would require, they should be assigned as a team; if they require more than 54.8 percent of the individual’s time, the manager should assign them to work individually (p. 457). The threshold is not fifty percent, which is what naive reasoning would suggest, because concurrent individual work carries its own penalty in the form of path variability. The team is permitted a measure of coordination loss and still wins.

Breakeven team efficiency deserves to be treated as the organizing concept of the assignment decision rather than as a numerical curiosity attached to one example. Klastorin and Mitchell (2021) noted in the chapter summary that understanding how the decision to assign individuals or teams affects project makespan, and learning to calculate breakeven team efficiency, provides a means of determining whether individual or team assignments are most appropriate (p. 474). The number 0.548 is specific to the network and duration distributions of that example and should not be carried into other projects. What generalizes is the procedure. Compute the expected makespan under concurrent individual assignment, divide it by the total expected work content, and the quotient is the efficiency a team must achieve to be the better choice. A manager who cannot estimate the team’s likely efficiency at least approximately has no basis for the decision at all.

Before any of this arithmetic applies, however, the workers must be substitutable. The example assumes Barb and Bob are cross-trained so that either can perform any task with equal effectiveness (Klastorin & Mitchell, 2021, p. 456). Where that condition fails, the assignment problem becomes a skills problem first and an efficiency problem second. Wang et al. (2022) examined precisely this dependency in their treatment of the multi-skilled resource-constrained project scheduling problem, observing that multi-skilled resources increase scheduling flexibility and expand the alternatives available for project scheduling while simultaneously making the underlying problem harder, because the manager must now decide not only which resource works when but which skill that resource exercises. Their guidance to practitioners is unambiguous on the point that a project team should be established with as many multi-skilled resources as the organization can supply, precisely because flexibility of assignment is what makes the later choice between individual and team configurations meaningful.

There is a further reason the individual configuration underperforms in the four-task example, and it is a reason with wide application. Parallel paths merge, and a merge point inherits the worst of its inputs. Because the project makespan under individual assignment is determined by the longer of the two chains, an unlucky realization on either chain delays the whole project, while a lucky realization on one chain buys nothing unless the other chain is also lucky. Klastorin and Mitchell (2021) drew the general lesson that a project manager must consider the effects of parallel paths and statistical variation and understand how they affect project performance (p. 458). This asymmetry is why concurrency is less valuable than it appears when durations are uncertain. Sequential teamed work has a single path and therefore aggregates variability additively rather than through a maximum, which under many distributions yields a lower expected completion time.

Turning to the affirmative case for individual assignment, the first and most obvious benefit is breadth of progress. Klastorin and Mitchell (2021) characterized the trade-off as one between completing multiple tasks concurrently and thereby making progress on multiple fronts, and completing individual tasks more quickly while progressing on fewer dimensions at any given moment (p. 456). Breadth has value that expected makespan does not capture. A project advancing on four fronts simultaneously surfaces problems in all four areas early, while a project advancing on one front at a time discovers the problems embedded in the fourth task only after the first three are complete. For projects in which the principal risk is discovering late that a requirement was misunderstood or a technical approach is unworkable, the informational value of parallel progress may outweigh a modest penalty in expected duration.

The second benefit of individual assignment is the complete absence of coordination overhead. A single worker on a single task has no one to synchronize with, no status meetings to attend regarding that task, and no need to reconcile a partial result with anyone else’s partial result. Klastorin and Mitchell (2021) defined project efficiency as coding duration divided by makespan, and team efficiency as one minus the ratio of coordination time to time spent directly delivering work content, and their tabulation shows both measures at one hundred percent for a team of one and declining monotonically thereafter (p. 464). A team of one is, by construction, perfectly efficient. Every hour it spends is an hour applied to the work. This establishes that any coordination cost, however small, is one individual assignment does not incur.

Empirical work outside the project management literature supports the productivity claim in per-capita terms. Mao et al. (2016) reported an online experiment in which forty-seven teams ranging in size from one to thirty-two participants worked on a realistic crisis mapping task, and found that average individual effort in the largest groups was roughly thirty percent lower than for workers operating independently. Their analysis also showed that per-capita productivity was highest for individuals and decreased with team size, a pattern consistent with the classical social loafing literature. The finding is relevant to project staffing because the compensation scheme tied each participant’s pay to their own monitored activity time as well as to team performance, which should have suppressed free riding. That effort still declined with group size suggests the effect is not simply a matter of poorly designed incentives.

Individual assignment also preserves clear accountability, and it accommodates a category of work that groups perform badly. Klastorin and Mitchell (2021) cited Cain’s observation that many creative people are most successful when working alone, and her reference to Steve Wozniak’s advice to inventors and engineers to work alone, “not on a committee. Not on a team” (p. 443). The same discussion characterizes the modern organizational enthusiasm for group work as a new groupthink, and frames the existence of project teams as defining a trade-off between worker privacy and solitude on one side and collaboration and synergy on the other (p. 443). A project manager staffing a task that requires sustained individual concentration, original design work, or the kind of thinking that is disrupted rather than assisted by interruption should weight this consideration heavily.

There is also a resource-management argument for spreading workers across concurrent tasks. Klastorin and Mitchell (2021) described the resource leveling problem as the scheduling of noncritical activities between their early and late starting times so as to minimize peak resource requirements and smooth utilization across the project makespan (p. 404). Concentrating an entire team on one task at a time produces a spiky resource profile in which demand for a particular skill is intense during one interval and absent in the next. Distributing workers across parallel tasks flattens that profile. The textbook notes that there are both explicit and implicit costs to adding or removing workers, including learning and start-up costs that leave new workers less than fully productive for some period (p. 404), and that leveling can minimize hiring and firing costs along with communication and training costs (p. 430).

Individual assignment carries a schedule risk that deserves mention here rather than in the trade-offs section, because it is a risk specific to the arrangement. A worker alone on a task with a comfortable deadline may consume the deadline. Klastorin and Mitchell (2021) discussed Parkinson’s law, the hypothesis that work expands to fill the time available for its completion, and showed through a two-task example that imposing a deadline of twenty-four days on a project with an expected duration of twenty-four days raises the expected duration to twenty-five days when workers behave according to that law (pp. 428–429). A procrastinating worker who delays starting a successor task produces a similar though smaller effect, yielding 24.3 days in the same example (p. 429). Teamed work is not immune to these behaviors, but mutual visibility within a team provides at least some check on them that solitary work does not.

The affirmative case for group assignment begins with the fact the four-task example was constructed to demonstrate. A team finishes any single task faster than an individual does. Where a task sits on the critical path, that compression translates directly into a shorter project. Klastorin and Mitchell (2021) concluded from the example that given a set of tasks that could be performed simultaneously, and assuming a project team can complete a task in time proportionate to team size with little or no loss of efficiency, it appears prudent to assign those tasks to a team and perform them sequentially (p. 458). The conditional clause is doing considerable work in that sentence, but the underlying logic is sound. Concentrating capacity on the constraint is the standard move in any system with a binding bottleneck, and a critical task is a bottleneck by definition.

The second benefit is skill coverage, and it is often decisive before any efficiency calculation is reached. Klastorin and Mitchell (2021) presented a mobile application development project requiring six distinct competencies, namely wireless technology, application development, mobile operating systems, user interface design, machine learning, and database management, drawn from a pool of six candidate workers none of whom possesses all six (pp. 449–450). Formulated as a set covering problem, the smallest team satisfying every skill requirement contains three members (p. 450). No individual assignment is feasible for a task requiring skills that no single worker holds. In such cases the manager is not choosing between team and individual work at all; the task itself dictates a group, and the remaining decision concerns the group’s composition and size rather than its existence.

Groups also correct errors that individuals working separately cannot, and this is the finding from the experimental literature that most complicates the simple productivity comparison. Mao et al. (2016) compared their real teams against synthetic teams assembled by pooling the outputs of an equivalent number of independent workers, and found that aggregating independent work reliably improved recall while degrading precision, because combining many independently generated reports increases the chance that at least one is correct and simultaneously increases the chance that at least one is wrong. Real teams, able to see and reconcile one another’s work, achieved better precision without much sacrifice in recall, and their collaborative activity more than doubled for the largest teams relative to independent workers. Compared directly, the gains from coordination dominated the losses from reduced individual effort, so that the largest teams outperformed the equivalent pool of solo workers. Brooks’s law, which holds that adding people to a late project makes it later, would have predicted negative marginal returns beyond some size, and the data did not show this even though participants in larger teams described coordination as burdensome. Perceived overhead and actual overhead are not the same quantity, and for project work in which quality is measured by the absence of defects rather than by volume of output, the asymmetry between precision and recall is a substantial argument for teamed assignment.

Group work also improves planning quality, which matters because planning is itself a project task subject to the same assignment decision. Klastorin and Mitchell (2021) reported a study finding that the quality of project plans produced by groups was significantly higher than the average quality of plans produced by individual members, and that plans produced by interacting groups were consistently better than those produced through nominal group techniques designed to suppress premature commitment (p. 468). The same study could not support the stronger hypothesis that a planning team consistently outperforms the best individual member working alone; only among the poorest-performing groups did the best member outperform the group as a whole (p. 468). The practical reading is that group planning reliably beats the average individual and is a safe default when the manager cannot identify in advance who the best individual planner would be.

A related benefit concerns problem solving as distinct from decision making. Klastorin and Mitchell (2021) summarized research showing that homogeneous groups produce fewer innovative solutions and are much more likely to accept inferior solutions, while nonhomogeneous groups produce higher quality solutions on problems involving quality alone (p. 442). They further reported a study distinguishing informational diversity, social category diversity, and value diversity, and found that informational diversity correlates positively with conflict and also positively with group performance (p. 443). Tasks that require the team to create a solution rather than select among established alternatives therefore benefit disproportionately from group assignment, provided the group is not composed of people who think alike. The assignment decision should therefore be sensitive to task type, with creative and analytical tasks favoring groups and routine execution favoring individuals.

Group assignment functions as risk mitigation in a sense that goes beyond schedule performance. Klastorin and Mitchell (2021) argued that project teams should always be viewed as part of the project’s risk management strategy, and that a well-designed project team is an effective risk mitigation device, giving the example of adding members with relevant market expertise when a competitor threatens to reach market first (p. 441). A task worked by a single individual has a single point of failure. If that person leaves, falls ill, or is reassigned, the task stops and whatever undocumented understanding they carried leaves with them. A task worked by two or three people degrades gracefully under the same shock. This consideration rarely appears in makespan calculations but is often decisive for tasks whose failure would be catastrophic rather than merely inconvenient.

The literature on scheduling under uncertainty reinforces this point from a different direction. Wang et al. (2022) modeled resource availability as a stochastic process and allowed multi-skilled resources to switch skills dynamically, so that when a resource with a needed skill becomes unavailable an idle resource holding the same skill can be redirected to cover the gap. Their numerical experiments found that dynamic scheduling performs better as resources hold more skills and as more resources are supplied, and worse as the probability of unavailability rises and as the number of distinct skills required increases. The mechanism that makes this work is redundancy of coverage, which team assignment supplies almost incidentally. A team of three cross-trained workers on a task is also an insurance policy against any one of them being absent.

Against these benefits stand the trade-offs, and the central one has already been stated in outline. Klastorin and Mitchell (2021) put it as a choice between simultaneous work and greater efficiency from teaming (p. 457). The manager who assigns individuals to tasks buys concurrency and pays for it in slower completion of each individual task and in exposure to the variability of parallel paths. The manager who assigns teams buys faster completion of each task and pays for it in serialization and in whatever efficiency the team loses to coordination. Neither purchase is obviously better, and the breakeven calculation exists precisely because the answer depends on parameters that vary from project to project. What follows examines the specific mechanisms through which teaming loses efficiency, since these determine whether the breakeven threshold is met.

The dominant mechanism is communication overhead, and its mathematics are unforgiving. Klastorin and Mitchell (2021) observed that interpersonal complexity, measured by the number of pairwise links among team members, increases as a function of team size and may raise the total hours required to complete a task or project (p. 459). The number of pairwise links on a team of N members equals N times N minus one, divided by two, which grows as a quadratic function of team size (p. 459). A two-person team has one link; a six-person team has fifteen; a twelve-person team has sixty-six. Because each link represents a channel that must be maintained, coordination effort grows far faster than the team’s productive capacity, which grows only linearly. The textbook notes this is especially pronounced in software projects, where code written by different people must be integrated into a single product (p. 459).

The textbook’s second worked example makes the consequences concrete and produces a result that should unsettle anyone who staffs by intuition. Klastorin and Mitchell (2021) modeled a project requiring fifty thousand lines of code, a programmer productivity of fifteen hundred lines per week, one hour per week of coordination per pairwise link, a programmer cost of $4,200 per week, and overhead of $3,000 per week (p. 460). Coding duration falls from 33.33 weeks with one programmer to 11.11 with three and 3.70 with nine, but coordination time rises throughout, and total cost traces a convex curve minimized at three programmers, at $186,333 (p. 461). A single programmer costs $240,000 and nine programmers cost $287,111 (p. 461). The cheapest arrangement is neither the smallest nor the fastest.

A structural finding embedded in that example deserves separate emphasis because it contradicts a common managerial assumption. Klastorin and Mitchell (2021) stated that when the costs of communication and coordination are correctly included in total project cost, direct resource costs are always an increasing function of team size, assuming all resources carry the same direct cost rate (p. 461). The chapter summary restates this as the proposition that direct project costs are always a nondecreasing function of team size, meaning a project’s direct labor costs can never be reduced by increasing team size (p. 475). Adding people can shorten a project. It cannot make the labor cheaper. Any managerial argument for a larger team must therefore rest on the value of time or on indirect cost recovery, never on direct labor savings, and a manager who claims otherwise has made an arithmetic error.

The most practically troublesome trade-off is that the team size minimizing cost and the team size minimizing duration are different numbers. Klastorin and Mitchell (2021) found that in the same coding example the makespan-minimizing team size is nine programmers while the cost-minimizing size is three (p. 461), and warned that team size affects makespan and cost differently when choosing the size of a team for a task (p. 464). There is no team size that is simultaneously optimal on both dimensions, which means the manager cannot avoid an explicit judgment about which objective governs. A project with severe schedule penalties and generous funding should staff toward nine; a project with a hard budget and flexible dates should staff toward three. What the manager must not do is choose a number without deciding which curve is being optimized.

Efficiency measures decline steadily as teams grow, and the decline is measurable. Klastorin and Mitchell (2021) tabulated project efficiency and team efficiency across team sizes and found both falling monotonically, with a fifteen-person team registering project efficiency of 83.6 percent and team efficiency of 80.3 percent against one hundred percent for a single worker (p. 464). Roughly one hour in six is being consumed by coordination rather than delivery at that size. The authors observe that this type of analysis provides a useful context for evaluating how well a project organization is using its resources to deliver project content (p. 464). Tracking these ratios across a portfolio gives a manager an empirical basis for estimating the team efficiency parameter that the breakeven calculation requires, converting an assumption into a measurement.

Behavioral trade-offs compound the mechanical ones. The reduced effort documented by Mao et al. (2016) was not confined to working less; participants in larger teams reallocated their effort toward less cognitively demanding subtasks, with time devoted to classifying events falling by roughly half while time spent on simple filtering rose by a comparable proportion. This is a subtler failure than idleness and considerably harder for a project manager to detect, because the team appears busy and its activity logs show work being done. The implication for assignment is that teamed work requires monitoring not merely of output volume but of output composition. A team that is completing the easy portions of a task at an impressive rate while the difficult portions accumulate is failing in a way that a conventional status report will not reveal.

Group harmony introduces a trade-off that runs opposite to intuition. Klastorin and Mitchell (2021) recounted a study by Brown and colleagues involving fourteen teams of MBA students completing a fifty-two task simulation of the preoperational testing phase of a nuclear power plant, in which teams were sorted into high and low performers by total project cost (p. 444). The teams that ultimately performed better were initially lower on every measure of group harmony and cohesiveness, and by project end no significant difference in harmony remained between the groups (p. 444). Low performers, meanwhile, maintained high cohesiveness throughout despite informal signals that their performance was below average (p. 444). The textbook’s conclusion is that some balance between cohesiveness and contentiousness is desirable, since teams with long tenure and high cohesiveness tend to avoid asking each other the difficult questions (p. 444).

Uncertain worker availability imposes a trade-off that individual assignment handles poorly and team assignment handles at a price. Klastorin and Mitchell (2021) noted that workers are seldom available one hundred percent of the time, whether because of illness, competing project demands, or nonproject operational obligations (p. 465). Their example concerns a one-day stress test requiring a minimum of seven call center technicians, each independently available with probability 0.90, which yields an expected availability of only 6.3 technicians if exactly seven are assigned (p. 465). Assigning eight raises the probability of having at least seven present to 0.813, and reaching ninety-nine percent confidence requires eleven, a resource safety stock of four (p. 466). The textbook draws the analogy to safety stock in supply chains, where inventory held in excess of expected demand buffers against demand uncertainty (p. 465).

That buffering has a cost, and the cost can be optimized rather than guessed. Klastorin and Mitchell (2021) extended the example by assigning a technician cost of $800 per day and a delay penalty of $10,000, and showed that the expected total cost is minimized at nine technicians rather than the eleven required for ninety-nine percent confidence (p. 467). The same answer follows from the newsvendor model, where the critical fractile of $10,000 divided by the sum of $10,000 and $800 equals approximately 0.926, and the smallest tabulated probability at least that large is 0.947, corresponding to nine technicians (p. 468). The lesson generalizes beyond the example. Confidence levels chosen by intuition, such as ninety-five or ninety-nine percent, are arbitrary, and determining team size to satisfy an arbitrary confidence level may increase project costs unnecessarily (p. 475).

Two further trade-offs arise from sustaining team assignments over time. Klastorin and Mitchell (2021) described a study by Katz of fifty research and development projects which found that project teams became increasingly isolated both inside and outside their organizations as group membership stabilized, that the resulting reductions in communication harmed technical performance, and that communication within the team also diminished with team longevity (p. 472). Highest performance occurred where average member tenure was between two and four years, with lower performance on either side (p. 472). A parallel decay affects the cross-training on which flexible assignment depends. Wang et al. (2022) observed that because multi-skilled resources are typically produced through cross-training, a worker who does not regularly practice a given skill tends to forget it, so that over a long enough horizon a multi-skilled workforce reverts to a single-skilled one. Assignment patterns that repeatedly place the same person on the same category of task therefore erode the very substitutability that makes the individual-versus-team choice available. Separately, performance incentives tied to project outcomes can trade one class of risk for another; the textbook recounts a case in which a team incentivized on capital cost delivered a plant under budget that was subsequently shut down, written off, and dismantled after severe corrosion appeared within three months of operation (p. 470).

Finally, the assignment decision itself generates conflict that the manager must be prepared to absorb. Klastorin and Mitchell (2021) reported an early classification of project team conflict into seven fundamental areas, among them project priorities, resources, cost estimates, scheduling and sequencing of work, and personality conflict (p. 470). Scheduling and sequencing appeared as a fundamental cause in three of the four project phases observed (p. 471). Since the choice between individual and team assignment is precisely a choice about resources and sequencing, it sits on top of two of the most reliable sources of intrateam friction. The textbook notes that experienced project managers overwhelmingly prefer collaboration as a resolution approach, followed by compromise, with avoidance least used (p. 471), which suggests that assignment decisions are better made with the team than announced to it.

Three propositions summarize the argument. The first is that the choice between individual and group assignment is a genuine trade-off with no dominant answer, in which concurrency and breadth of progress are exchanged for speed on individual tasks and depth of coverage, and the exchange rate is set by team efficiency. The second is that this exchange rate can be computed rather than guessed, through the breakeven efficiency calculation that divides expected concurrent makespan by total work content, and that a manager who has never performed this calculation for a representative project is staffing by intuition alone. The third is that team size interacts with cost and schedule through different functions, so that no single team size optimizes both, and that direct labor costs never fall as team size rises regardless of what schedule compression is achieved.

For practitioners, several concrete actions follow. Begin by establishing whether concurrency is structurally available, since a serial network makes the question moot. Where it is available, estimate team efficiency from historical data on comparable tasks rather than assuming proportional scaling, and compare that estimate against the computed breakeven threshold. Favor individual assignment for tasks requiring sustained concentration, original creative work, or clear personal accountability, and for portfolios where parallel progress surfaces risk early. Favor team assignment for critical path tasks, for tasks whose skill requirements exceed any single worker, for work whose quality depends on error detection, and for tasks where a single point of failure would be intolerable. Decide explicitly whether cost or duration governs before selecting a team size, cross-train deliberately so that assignment flexibility exists when it is needed, and size buffers against absence using cost ratios rather than arbitrary confidence targets.

These conclusions should be held with appropriate humility. The models examined here rest on assumptions that projects routinely violate, including independent task durations, homogeneous worker productivity, coordination cost proportional to pairwise links, and a stable relationship between resources applied and duration achieved. The empirical evidence is genuinely mixed; the same experimental study that documents substantial social loafing also finds that collaboration gains outweigh it, and the same textbook that endorses teaming for parallel tasks records that highly cohesive teams underperform contentious ones. Klastorin and Mitchell (2021) themselves conceded that the models are useful chiefly for illustrating the nature of the trade-offs, though they may become indispensable when the question is whether to assign forty people to crash a project rather than the planned twenty (p. 475). That concession is the right note on which to end. The value of the framework is not that it produces a defensible number but that it forces the manager to name the trade-off being made, and a manager who can explain why additional resources may make things worse is better equipped than one who merely suspects it.

References

Klastorin, T., & Mitchell, G. (2021). Project management: A risk-management approach (1st ed.). SAGE Publications.

Mao, A., Mason, W., Suri, S., & Watts, D. J. (2016). An experimental study of team size and performance on a complex task. PLOS ONE, 11(4), Article e0153048. https://doi.org/10.1371/journal.pone.0153048

Wang, M., Liu, G., & Lin, X. (2022). Dynamic optimization of the multi-skilled resource-constrained project scheduling problem with uncertainty in resource availability. Mathematics, 10(17), Article 3070. https://doi.org/10.3390/math10173070