Organizational Design, Job Evaluation and Position Classification: How Structure Decisions Shape Coordination, and Why Job Evaluation Is Not Fully Neutral

Organizational design is the choice of how to group jobs, departments and decision rights so that work actually gets coordinated rather than merely divided. Job evaluation is a separate, later step: a systematic way of comparing the relative worth of jobs to each other, independent of who currently holds a job or what that person happens to be paid today. The two are taught together in this lesson because the first decision determines the shape of the second. A structure decides which jobs report to which other jobs, how many layers sit between the front line and the top, and how centralized decision-making is; job evaluation then has to compare jobs within whatever shape that structure produced. Get the structure wrong and job evaluation inherits the confusion, comparing jobs whose boundaries were drawn arbitrarily in the first place.

One boundary needs to be stated plainly before either topic starts. Job evaluation, as this lesson teaches it, produces a ranking or a score that says job A is worth more, less or the same as job B. It does not, by itself, say what job A should actually be paid. Turning a job evaluation result into an actual pay grade and a salary range is Module 11's subject, specifically Lesson 52. Treating the two as the same step is a common practical error, and this lesson avoids it deliberately: everything below stops at the comparison, not the price tag.

A functional structure groups people by the kind of specialist work they do: engineering together, finance together, marketing together, human resources together. Its advantage is depth. Specialists sitting among other specialists develop their craft faster, share techniques more easily, and benefit from economies of scale that come from having enough volume in one function to justify specialized equipment, tools or expertise that a smaller, scattered group could not support. A functional structure also tends to be relatively easy to supervise, since each manager oversees people doing recognizably similar work, which simplifies both performance judgment and the span of control question addressed later in this lesson.

The cost shows up whenever work has to cross functional lines, which in most organizations is most of the time that actually matters commercially. Launching a new product or closing a deal that needs legal, finance and delivery moving together requires coordination across departments with no single boss below the top of the organization. Each function optimizes for its own metrics, consistent with how it is organized and measured, and the resulting friction is not a sign of people behaving badly; it is the structure doing exactly what a functional grouping is built to do, concentrate expertise rather than customer or product outcomes. Tom Burns and G. M. Stalker's 1961 study, published as The Management of Innovation, gave this a name: a highly formalized structure of this kind, which they called mechanistic, performs well in a stable environment and comparatively poorly in a turbulent one demanding frequent, fast cross-functional adjustment. A functional structure is not a mistake; it is a bet that the environment will stay stable enough for its coordination weaknesses to matter less than its specialization strengths.

A divisional structure groups work by product line, market segment or geography instead of by specialism, with each division typically running its own functions, its own engineering, its own marketing, its own finance, as a largely self-contained unit. The business historian Alfred Chandler studied this shift directly in his 1962 book Strategy and Structure: Chapters in the History of the American Industrial Enterprise, examining how DuPont, General Motors, Standard Oil of New Jersey and Sears independently arrived at a decentralized, divisionalized structure. His central argument, since shortened into the phrase structure follows strategy, was that a centralized, functionally organized company could manage one core business reasonably well, but once a company diversified into several genuinely different product lines or markets, a single centralized structure could not process the resulting decision load fast enough, and each company restructured into divisions to push decision-making closer to where the relevant market information actually was.

That is a historical explanation, built from four specific companies in a specific industrial era, and it should not be read as a universal law that divisionalizing is always the right answer once a company grows. A single-product company gains comparatively little from dividing into product-based units; it mainly loses the economies of scale a unified function was providing, while duplicating finance, HR and marketing capability across divisions that do not actually need to be separated. Divisionalization solves a specific problem, the information and coordination load created by genuine diversification, and introduces a specific cost: duplicated overhead and a risk that divisions optimize for their own results even when the company would be better served by sharing resources across division lines. Whether that trade works in a given company's favor depends on how different its businesses actually are from each other, not on size by itself.

A matrix structure overlays a second dimension, usually product, project or geography, on top of the functional one, so that an employee reports to both a functional manager and a product or project manager at the same time. The design intention is to get the specialization benefits of a functional structure and the coordination benefits of a divisional one simultaneously, without having to pick between them. Stanley Davis and Paul Lawrence's 1977 book, Matrix, is the detailed account of what that intention costs in practice. They documented a specific set of recurring failure patterns, which they called matrix pathologies, including what they termed decision strangulation, where every decision of any size gets escalated for joint sign-off because neither manager has unilateral authority to resolve it, and a tendency toward excessive overhead, as the organization adds coordinating roles and meetings specifically to manage the ambiguity the dual-reporting design itself created.

The mechanism behind these pathologies is not mysterious once stated directly. Dual authority is a structural feature, not a management failing that better training could remove, and it produces a genuine, recurring question: whose priority wins when the functional manager and the product manager disagree about how an employee's time should be spent this week, with no lower authority positioned to resolve it quickly. Resolving that question, over and over, is real managerial time, and that time is the overhead cost Davis and Lawrence described, not an exaggeration of it. A matrix is, in Burns and Stalker's terms, a bet on an organic structure: better suited to a turbulent environment needing fast cross-functional integration, worse suited to a company that would be better served by the lower overhead of a cleaner functional or divisional design.

Two further variables cut across all three structures and shape how any of them actually behaves in practice. Span of control is the number of people reporting directly to one manager. Centralization is the degree to which decision rights sit at the top of the organization rather than being pushed down to the people closer to the work. Neither variable is fixed by the structure type; a functional, divisional or matrix organization can each be built with a wide or narrow span, and with decisions held centrally or pushed outward.

The classical management view, represented by Lyndall Urwick's 1956 Harvard Business Review article, "The Manager's Span of Control," argued for a narrow span, Urwick suggested numbers in the range of five or six direct reports at senior levels, grounded in V. A. Graicunas's earlier mathematical argument that the number of possible relationships a manager has to track grows far faster than the number of direct reports itself. That argument is a logical and normative one, an argument about what a manager ought to be able to handle, rather than a measured study of how spans actually function inside real organizations. William Ouchi and John Dowling's 1974 study in Administrative Science Quarterly, "Defining the Span of Control," tested the classical assumption more directly and found something that complicates it: the formal number of people reporting to a manager was a poor proxy for how much actual direct supervision was happening, and the managers in their study spent, on average, only around half of their working time on direct supervisory activity at all, with the remainder going to other work entirely. That finding does not prove Urwick's prescription wrong in every case, but it does mean that a narrow span, by itself, does not guarantee close supervision, and a wide span does not automatically mean a manager is overstretched; what the manager actually does with the time matters as much as the headcount reporting to them.

Centralization interacts with span of control rather than operating independently of it. A wide span paired with high centralization is a workable combination at the operating-core level of a standardized business: a factory supervisor overseeing twenty machine operators whose jobs are already tightly specified needs relatively little discretionary decision-making in the moment. The same combination at a senior level, where each direct report is making genuinely different judgment calls, is far less workable, because the manager cannot meaningfully review that many distinct, non-routine decisions without becoming the bottleneck. The practical design question is therefore not "what is the correct span of control" in the abstract, since Ouchi and Dowling's findings suggest no single number answers that honestly, but what level of decision discretion the roles actually carry, and whether centralization matches that level or fights against it.

Once a structure exists, specific jobs exist inside it, described by the job descriptions and job specifications that Lesson 32 covered. Job evaluation takes that set of descriptions and asks a narrower, more disciplined question than it might first appear to ask: not "is this a good job" or "is this person doing well," but "how does the content of this job compare to the content of other jobs in the organization." The comparison is about the job as defined on paper, the duties, required skill, level of responsibility and working conditions, independent of the specific person filling it and independent of market salary data, which belongs to a separate compensation-survey step. Confusing job evaluation with performance appraisal, or with checking what a competitor pays for a similar title, is a common practical error, and it produces results that satisfy neither purpose well.

Four methods dominate how organizations have historically done this comparison, and they differ mainly in how much they break a job apart before comparing it. Ranking compares whole jobs against each other directly, typically by having a small panel order job descriptions from most to least demanding without decomposing them into separate dimensions first. It is fast and simple to administer in a small organization with a manageable number of distinct roles, and its main weakness is exactly that simplicity: a whole-job judgment offers no documented reasoning for why one job outranked another, which makes it hard to defend consistently as the organization grows or as a ranking gets challenged.

Classification, sometimes called the grading method, sorts jobs into a predetermined set of grades or classes, each defined by a written class specification describing the kind and level of work that belongs in it, and then slots each job description into whichever class specification it most closely matches. This is the lineage behind formal position classification systems; the United States' Classification Act of 1923 established one of the earliest large-scale versions of this approach in the federal civil service, grouping positions into occupational series and grades based on the duties and responsibilities of the position itself rather than the qualifications of whoever happened to hold it. Classification works well when an organization already has a stable, well-understood set of grade levels to sort into; it works less well when a genuinely new kind of job does not fit comfortably into any existing class and forces a judgment call about which grade's definition it resembles most, which reintroduces some of the same whole-job subjectivity ranking has.

Point-factor evaluation and factor comparison both break a job into separate compensable factors before comparing it to other jobs, which is the more analytical half of the four methods and the subject of the next section in more detail.

Merrill Lott is generally credited with the first systematic version of the point method, described in his 1926 book Wage Scales and Job Evaluation, which set out to assign separately weighted point values to distinct job characteristics instead of comparing whole jobs intuitively, an approach his own work framed as a more scientific basis for setting wage rates than the ad hoc methods common at the time. Eugene Benge, Samuel Burk and Edward Hay carried the idea further in their 1941 Manual of Job Evaluation, codifying a standard set of compensable factors, commonly grouped as skill, effort, responsibility and working conditions, that became a template other point-factor systems built on. That same grouping, skill, effort, responsibility and working conditions, later became the explicit statutory language the United States used in the Equal Pay Act of 1963 to define what counts as equal work for equal-pay purposes, which is one indication of how directly job evaluation's technical vocabulary fed into employment law rather than staying confined to compensation departments. Edward Hay went on, in 1951, to develop the Guide Chart-Profile Method through what became the Hay Group, scoring jobs on know-how, problem-solving and accountability; it remains one of the most widely used point-factor systems globally today.

The mechanics of the point method are straightforward to describe. A set of compensable factors is chosen, each factor is divided into several degrees of increasing demand, each degree is assigned a point value, and a job is scored by identifying which degree of each factor it requires and summing the resulting points into a single total. The total score, not a job title or a manager's impression, is what gets compared across jobs. This produces something ranking and classification do not: a documented, factor-by-factor rationale for why one job scored higher than another, which is considerably easier to defend when an employee or a regulator asks why two jobs are valued differently.

Position classification, as described above under the classification method, deserves a second look as a system rather than a single technique, because many large organizations, particularly in the public sector, still run on it directly rather than on a point-factor score. A classification system defines a fixed ladder of grades in advance, writes a class specification for each grade describing the kind and level of work it covers, and then places every position into the grade whose specification it matches most closely. The appeal is administrative consistency: once the grade ladder exists, assigning a new position to a grade is a comparison against a known standard rather than a fresh evaluation exercise each time. The limitation is the one already noted for the classification method generally: a genuinely novel role, one that does not resemble any existing class specification closely, forces a subjective judgment about which grade it belongs in, and that judgment is exactly where disputes about fairness tend to surface.

Whichever of the four methods an organization uses, the output is the same kind of thing: a ranking, a score or a grade that states the relative worth of jobs to each other. None of the four methods, on its own, produces a salary figure. Converting that relative ordering into actual pay grades and salary ranges, deciding how many grades to have, how wide each grade's salary band should be, and how progression within a grade should work, is the specific work of Module 11, Lesson 52 in particular. This lesson's job evaluation output is an input to that later design step, not a substitute for it.

Job evaluation is usually presented, including in the sections above, as a technical procedure: define factors, assign points, sum the score. That framing is useful for learning the mechanics, and it is also incomplete in a way worth stating directly rather than leaving implicit. The National Research Council's Committee on Occupational Classification and Analysis, publishing its findings as Women, Work, and Wages: Equal Pay for Jobs of Equal Value in 1981 under the editorship of Donald Treiman and Heidi Hartmann, reviewed a substantial body of job evaluation practice in the United States and found a specific, documented pattern: jobs held predominantly by women were, on average, scored and paid lower than jobs held predominantly by men even where the combined factor content of the two jobs, as measured by the very point-factor systems meant to be neutral, was comparable. The committee's explanation was not that evaluators were deliberately rigging scores. It was that the choice and weighting of compensable factors is itself a policy decision, and those decisions had historically been calibrated against a labor market already segregated by sex, so skills more associated with physically demanding or technically coded work tended to be weighted more heavily than skills more associated with caregiving, interpersonal coordination or sustained fine-motor dexterity, categories of work that happened to be concentrated in female-dominated occupations.

Richard Scholl and Elizabeth Cooper's 1991 article in Public Personnel Management, "The Use of Job Evaluation to Eliminate Gender Based Pay Differentials," examined the same mechanism from the other direction: under what conditions job evaluation can actually help correct a gender-based pay gap rather than reproduce it. Their analysis pointed to the same lever the committee identified, factor selection and weighting, and to a related practical issue: organizations running separate, incomparable evaluation systems for different job families, rather than one shared factor set applied consistently across them, made it structurally difficult to even detect whether a gap existed, let alone correct it. A numeric score can look objective by virtue of being a number, while still carrying forward whichever assumptions were built into the factors generating it. This lesson states that plainly: the methods above genuinely make job comparisons more documented and consistent than an unexplained whole-job impression, and they are also policy instruments whose factor choices deserve the same scrutiny as any other decision capable of embedding bias.

Two limitations deserve to be stated rather than smoothed over. First, the functional, divisional and matrix typology, and Mintzberg's five configurations alongside it, are useful categories for thinking about structure, not a menu with one correct item on it. Most organizations that have grown past a certain size are hybrids blending elements of more than one configuration, often deliberately, as the Provati case above illustrates; treating any one of the three structures as the universally correct choice for a growing company would overstate what Chandler's, Burns and Stalker's, or Davis and Lawrence's research actually supports, each of which describes conditions under which a structure tends to work well, not a structure that works well regardless of conditions.

Second, the comparable-worth-era research behind the gender-bias finding, Treiman and Hartmann's 1981 study in particular, was conducted on United States data within a legal and labor-market context that included active litigation and policy debate over comparable worth during the 1970s and 1980s. Whether the same mechanism operates with the same magnitude in other labor markets, including Bangladesh, where the legal and bargaining history around comparable worth differs substantially, has not been established with the same depth of evidence reviewed here. The structural risk described, that factor selection and weighting choices can embed existing occupational segregation into a seemingly neutral score, has no obvious reason to be confined to one country's labor market, but whether it operates identically elsewhere, and at what magnitude, is a genuine open question this lesson flags rather than answers.

Select an organization you can examine with reasonable access to its actual structure: an organization you work for or have worked for, one you can study through published organizational charts, annual reports or case materials, or a disguised composite if direct access would create a conflict of interest. The organization should have at least two distinct functions or business lines whose coordination you can describe concretely.

Prepare a report of 2,000 to 2,500 words that:

1. identifies which of the structures discussed in this lesson, functional, divisional, matrix or an explicit hybrid, the organization actually uses, with specific evidence from reporting lines and decision rights, not just its stated org chart label;
2. evaluates whether that structure fits the organization's actual degree of product, market or geographic diversification, using Chandler's (1962) structure-follows-strategy argument as the analytical lens, and states one specific coordination cost or benefit the current structure is producing;
3. assesses the organization's span of control and centralization at one specific management level, stating whether the formal span appears to match the actual decision discretion carried by that level's roles, drawing on Ouchi and Dowling's (1974) finding that formal span is a poor proxy for actual supervisory time;
4. identifies which job evaluation method, if any, the organization uses to compare job worth (ranking, classification, point-factor or factor comparison), or notes honestly if no formal method exists, and if a point-factor or similar analytical method is used, states what compensable factors it relies on and whether their selection or weighting shows any of the risks Treiman and Hartmann (1981) and Scholl and Cooper (1991) identified; and
5. recommends one specific structural or job evaluation change the evidence reviewed in this lesson would support, and states one limitation of that recommendation the lesson's evidence does not resolve.

Use at least five academic sources, including at least three from this article's reference list. State explicitly where information about the organization is unavailable and how that gap limits your conclusions, rather than filling the gap with assumption.

Assessment rubric (100 marks)

1. Benge, E. J., Burk, S. L. H., & Hay, E. N. (1941). Manual of Job Evaluation: Procedures of Job Analysis and Appraisal. Harper & Brothers.
2. Burns, T., & Stalker, G. M. (1961). The Management of Innovation. Tavistock Publications.
3. Chandler, A. D., Jr. (1962). Strategy and Structure: Chapters in the History of the American Industrial Enterprise. MIT Press.
4. Davis, S. M., & Lawrence, P. R. (1977). Matrix. Addison-Wesley.
5. Lott, M. R. (1926). Wage Scales and Job Evaluation: Scientific Determination of Wage Rates on the Basis of Services Rendered. Ronald Press.
6. Mintzberg, H. (1979). The Structuring of Organizations: A Synthesis of the Research. Prentice-Hall.
7. Ouchi, W. G., & Dowling, J. B. (1974). Defining the span of control. Administrative Science Quarterly, 19(3), 357–365.
8. Scholl, R. W., & Cooper, E. (1991). The use of job evaluation to eliminate gender based pay differentials. Public Personnel Management, 20(1), 1–18.
9. Treiman, D. J., & Hartmann, H. I. (Eds.). (1981). Women, Work, and Wages: Equal Pay for Jobs of Equal Value. National Academy Press.
10. Urwick, L. F. (1956). The manager's span of control. Harvard Business Review, 34(3), 39–47.

Verification notes: every reference above was checked against publicly indexed bibliographic records (publisher pages, journal archive pages, library catalog entries, or university-hosted summaries) via web search rather than a full-text read of each source. Findings attributed to Chandler (1962), Davis and Lawrence (1977), Ouchi and Dowling (1974), Urwick (1956), Lott (1926), Benge, Burk and Hay (1941), Treiman and Hartmann (1981) and Scholl and Cooper (1991) are reported at the level of detail available in abstracts, publisher summaries, library records and secondary academic accounts; no statistic in the article text goes beyond what surfaced in those sources. The U.S. Classification Act of 1923 and the Equal Pay Act of 1963's statutory factors are reported as historical and legal facts from public records, not as academic findings, and are described in prose rather than placed in the numbered list above. The Provati EdTech case, the tables, the decision scenario, and the synthesis connecting this lesson to Lessons 32, 34 and Module 11 are original teaching material for this series.

Lesson 34 continues Module 07 by turning from structure and job comparison to workforce supply, demand and gap analysis, working out how many people an organization actually needs inside whichever structure it has chosen.