Workforce Supply, Demand and Gap Analysis: From Skills Inventories and Markov Chains to the Real Choice Between Hiring, Training and Restructuring

It is tempting to think of workforce planning as one combined estimate: how many people will we need. In practice, the useful version of the exercise forces two independent forecasts that are only compared at the end. Supply analysis asks what the organization will have available, counting its current workforce, its internal movement patterns, and whatever the external labor market can plausibly add. Demand analysis asks what the organization will need, driven by its business plan, its productivity assumptions, and its own managers' judgment about where work is growing or shrinking. Keeping the two separate matters for a specific reason: collapsing them into a single number hides which half of the estimate was weak. A gap can exist because supply was overestimated, because demand was overestimated, or because both forecasts were individually reasonable but built on assumptions that will not hold at the same time. Comparing two explicit forecasts, each with its own stated method and its own stated uncertainty, is the only way to know which assumption to revisit when the plan turns out to be wrong, which it usually will be to some degree.

This lesson works through internal supply, external supply, demand forecasting, and gap analysis in that order, because that is roughly the order in which the information becomes available to a planning team: what the organization already has is the easiest question to answer with reasonable confidence, and what the gap implies for action is the hardest.

A skills inventory is a structured record of what the current workforce can do, not merely what job title each person holds. A well-maintained inventory captures certifications and licenses, language ability, equipment or software proficiency, performance history, geographic mobility, and often a stated interest in future roles. The distinction matters because job titles and organization charts describe positions, while a skills inventory describes people, and the two do not always line up. A technician hired five years ago as a general mechanic may since have earned the exact certification a new fleet requires, a fact that a title-based headcount report would never surface. The inventory's value is almost entirely a function of how current it is kept; a skills inventory updated once a year during an annual review cycle is already a historical document by the time a planning team needs it for a decision six months later.

The honest limitation of a skills inventory is that it answers "what do we have" far more reliably than "what will we have." People leave, get promoted, change roles, or simply decline to use a skill they are technically certified in. Treating an inventory as a static warehouse catalog, rather than a snapshot of a population that is constantly moving, is the single most common misuse of the tool, and it is the reason the next two tools in this lesson, replacement charts and Markov chain models, exist: both are attempts to describe how that movement happens, not just what exists at one point in time.

A replacement chart takes a small set of critical positions, usually senior or highly specialized ones, and lists against each one the current incumbent and one or two named people who could step into the role, typically tagged by readiness: ready now, ready in one to two years, or ready in three or more years with specific development. It is deliberately narrow. Where a skills inventory tries to describe the whole workforce, a replacement chart concentrates on the roles whose vacancy would actually hurt the organization if it happened without warning, which is a different design choice, not a smaller version of the same thing.

Replacement charts are easy to criticize as a thin planning tool, because they usually rest on a manager's subjective judgment about who is "ready," and that judgment can be shaped by visibility and favoritism as much as by demonstrated capability. The criticism is fair but incomplete: a replacement chart is not pretending to be a statistical forecast. It is a deliberately qualitative tool for a small number of positions where the cost of being wrong is high enough that a manager's informed judgment, made explicit and reviewed rather than left implicit, is worth more than no answer at all. Lesson 69 returns to the broader succession-pipeline question this tool only partially addresses; here, a replacement chart is treated as one input into internal supply analysis, not as the organization's full answer to leadership continuity.

A skills inventory and a replacement chart both describe people as they are right now, or as a manager judges they might become. Markov chain analysis tries to do something more specific: use an organization's own recent history of internal movement, promotions, lateral transfers, and exits, to estimate the probability that a person in a given job category this period will be in each possible category, including "no longer employed here," next period. David Bartholomew's 1967 book Stochastic Models for Social Processes developed this family of models for manpower and educational systems generally, treating an organization's job levels or departments as a set of "states" through which people flow according to transition probabilities estimated from counting how many people actually moved between each pair of states in a past period. Victor Vroom and Kenneth MacCrimmon applied a closely related stochastic approach specifically to managerial careers in their 1968 Administrative Science Quarterly article, modeling how individuals moved through a single organization's management hierarchy over time and finding the approach useful for describing real career patterns rather than treating each promotion as an unpredictable, case-by-case event.

The practical reading of Heneman and Sandver's critique is not that Markov analysis is useless, but that it answers a narrower question than it looks like it answers. It describes aggregate flow for a category of people reasonably well over a short horizon, particularly in large, stable categories with a long run of consistent history behind them. It describes what any specific individual will do considerably less well, and it describes the near future of a small, fast-changing, or newly created category barely at all. A planning team that uses it to estimate how many of this period's 200 junior analysts will likely be senior analysts next year is on reasonably solid ground. A planning team that uses the same transition matrix to decide which specific analyst gets promoted, or expects the same matrix to still be valid after a strategic reorganization changes who gets promoted and why, is extending the tool past where its own assumptions hold.

Internal supply analysis describes people the organization already employs. External supply analysis asks a different and, in most respects, a harder question: how many people with the needed skills exist in the labor market the organization actually recruits from, and how is that number likely to change. The United States Bureau of Labor Statistics publishes one of the more methodologically explicit public examples of this kind of forecasting through its Employment Projections program, which builds labor force projections by applying projected labor force participation rates, broken out by age, sex, race and ethnicity, to population projections produced separately by the Census Bureau, and then works that labor force estimate through the rest of an economy-wide occupational projection model. The method is systematic and transparent about its own steps, which is precisely what makes its limitations instructive rather than embarrassing: a ten-year projection built this way is only as good as its participation-rate and population assumptions, and those assumptions are reviewed and revised in later editions precisely because they do not hold perfectly over a decade.

An individual organization planning its own workforce rarely has the resources to build anything close to that level of national modeling, and usually does not need to. What a planning team can reasonably do is track the demographic and labor market trends relevant to its own specific talent pool: the output of the universities or training institutes that supply its specialist roles, the regional unemployment rate in the specific occupation it hires for rather than the headline rate, migration patterns if it recruits across borders, and the age structure of the specific occupational group it depends on, since an occupation with a visibly aging incumbent population poses a different external-supply risk than one with a young, growing pipeline. The honest caveat applies here as much as it does to the national-level data: external supply estimates for a specific skill in a specific region are frequently thinner and less current than an organization would like, and a workforce plan that treats a two-year-old regional labor market estimate as current fact is building on weaker ground than it appears to be.

Demand forecasting asks how many people, in which roles, the organization's own plans will require. James Walker's 1980 text Human Resource Planning, still cited as one of the field's foundational treatments of forecasting methodology, describes ratio-trend analysis as the most common quantitative starting point: take a historical ratio between a business activity measure, such as units produced, revenue, or client accounts, and the number of employees in a given job category required to support it, and project that ratio forward against the planned level of future activity. If the organization has historically needed one production supervisor for every fifteen production workers, and plans to add sixty production workers, ratio-trend analysis implies four additional supervisors, provided the ratio itself is assumed to hold.

That last clause is where the technique's real weakness sits, and it is a weakness of assumption rather than arithmetic. A ratio-trend forecast is only as good as the stability of the underlying ratio, and the ratio changes whenever productivity changes, whether through new equipment, process redesign, or a change in the skill mix of the workforce itself. A closely related technique, productivity-ratio forecasting, makes this dependency explicit rather than implicit: it forecasts the total output required, estimates output per worker based on current or planned productivity, and divides the first by the second to get a headcount. Making the productivity assumption explicit is an improvement over quietly assuming last year's ratio still holds, but it does not remove the underlying judgment call; it only forces the planner to state it plainly, which is itself worth doing, since a stated assumption can be challenged and a hidden one usually is not.

Both techniques share a structural advantage and a structural limitation. The advantage is that they are fast, require only data the organization already has, and produce a number that is easy to explain to a finance department asking for a headcount justification. The limitation is that both techniques are essentially extrapolations of the past; they forecast demand well for activities that are growing or shrinking smoothly along an established pattern, and they forecast badly for a genuinely new role, a discontinuous change in strategy, or a skill the organization has never needed before, because there is no historical ratio to extrapolate from in the first place.

The illustrative chart above is a constructed teaching example, not a real company's data, and it is built to make a specific point visible: a gap does not have to appear suddenly to be serious. In this hypothetical case, the shortfall in Year 1 is modest, four technicians, and an organization that only checks its numbers once a year might reasonably decide it is not worth acting on yet. By Year 3, on the same hypothetical assumptions, the gap has grown to twenty-four, a shortfall that cannot plausibly be closed through hiring alone in the time remaining if it is only noticed that late. The lesson is about the forecasting horizon, not about these specific figures: a widening gap trend, caught early through a ratio-trend or productivity-ratio projection run every year rather than only when a shortage is already being felt, leaves more response options available than the same gap caught late.

Ratio-trend and productivity-ratio forecasting both assume a stable, extrapolable past. When that assumption is weak, either because the role is new, the strategy is changing discontinuously, or the organization simply lacks reliable historical data, demand forecasting falls back on structured human judgment instead of a formula. The simplest version, often called a jury of executive opinion, gathers relevant managers to discuss and converge on a demand estimate together. It is fast and draws on real operational knowledge that a formula cannot capture, but it carries a specific, well-known risk: a face-to-face group discussion is vulnerable to whoever is most senior, most confident, or most recently vocal in the room, and the resulting number can reflect social dynamics as much as considered judgment.

Norman Dalkey and Olaf Helmer's 1963 Management Science paper, "An Experimental Application of the Delphi Method to the Use of Experts," describes a structured alternative designed specifically to reduce that risk. In a Delphi process, a panel of experts answers a forecasting question independently and anonymously, the facilitator aggregates the responses and shares the aggregate, typically with the range of views and the reasoning behind outlying estimates, back to the same panel, and the panel answers again, now informed by the group's spread of opinion but without knowing which individual said what. Repeating this for a few rounds, Dalkey and Helmer found, tends to narrow the spread of expert estimates without the social pressure of a face-to-face argument pushing the group toward whoever spoke first or loudest. The method was developed originally for technological forecasting, not human resource planning specifically, and applying it to a demand estimate for next year's headcount is an adaptation of the original method to a different domain, one that preserves its core structural feature, anonymous and iterative judgment with controlled feedback, while changing the question being asked.

The honest trade-off between the two judgmental methods is speed against structure. A jury of executive opinion can produce a number in one meeting; a proper Delphi process takes multiple rounds over days or weeks. Neither method eliminates the fact that human judgment, however structured, is still an estimate rather than a measurement, and both remain vulnerable to the same forecasting bias, collectively underestimating or overestimating future need, if the panel itself shares a common blind spot about where the business is heading.

Once supply and demand have each been forecast for the same future period and the same job category, gap analysis is the comparison between them. A shortage exists where forecast demand exceeds forecast supply; a surplus exists where forecast supply exceeds forecast demand. The distinction sounds almost too simple to dwell on, but it matters because the two conditions call for largely non-overlapping responses, and a planning process that produces a single organization-wide headcount number, rather than a gap broken out by job family, department or location, can easily hide the fact that both conditions exist simultaneously in different parts of the same organization. A logistics company adding certified technicians while its customer-service headcount shrinks because of a new self-service app is not one workforce problem; it is a shortage and a surplus sitting inside the same company at the same time, and solving the first by hiring externally while ignoring the second as a possible internal source wastes an obvious option.

None of the four demand-forecasting techniques in this lesson, and neither of the internal-supply tools, settles the question of which response an organization should actually choose once a gap is confirmed. They produce the number; the choice between hiring, training, restructuring and managing attrition is a separate judgment that has to weigh variables the forecasting methods themselves do not capture. How scarce is the skill in the external market, a question external supply analysis speaks to directly. How learnable is the skill internally within the time the forecast horizon allows, a question only the organization's own training capability can answer. How confident is the organization in its own forecast, which depends on whether it came from a stable ratio-trend extrapolation or from a jury of executive opinion guessing at a genuinely new kind of role. And how reversible is each response: a hiring freeze can be lifted quickly if the forecast turns out wrong, while a round of layoffs cannot be undone by rehiring the same people on the same terms a year later.

This is also where gap analysis connects back to organizational design, Lesson 33's subject: a persistent, structural gap, one that keeps reappearing even after hiring and training responses, is sometimes a signal that the job itself, or the way work is divided across roles, needs to change rather than simply be staffed differently. Restructuring is listed in the response table above as a surplus response, but it belongs on the shortage side as well, in the less obvious form of redesigning a scarce role so that more of its work can be done by people who are not scarce, which is a design decision, not a staffing one.

Two limitations deserve to be stated directly rather than smoothed over. First, the evidence that formal workforce planning actually improves organizational performance is weaker and more mixed than the confident tone of most practitioner writing on the subject suggests. Stella Nkomo's 1987 study in the Strategic Management Journal compared firms using formal human resource planning against firms that did not, and found no statistically significant difference in several performance measures between the two groups cross-sectionally, alongside some indication of a positive change in performance within firms after they adopted formal planning, measured before and after adoption. That is a genuinely mixed result, not a clean endorsement, and it is a useful corrective to any claim that the techniques in this lesson are proven to pay for themselves; they are tools for making a specific kind of decision more informed, not a demonstrated source of competitive advantage in their own right.

Second, nearly every technique and source cited in this lesson, Bartholomew's and Vroom and MacCrimmon's Markov models, the Bureau of Labor Statistics methodology, Walker's and Bechet's forecasting frameworks, was developed in, or calibrated against, large-firm, Western, often unionized or highly formalized labor market contexts. Applying a Markov transition matrix built on stable, well-documented historical movement data is a different exercise in an organization, or a labor market, where job categories are less formally defined, where informal or seasonal employment is common, or where historical personnel records simply do not exist in the detail the method assumes. None of this means the underlying logic, compare supply to demand by category, is wrong outside those contexts; it means the specific techniques' data requirements are not automatically satisfied everywhere they might be applied, and a planning team working outside the conditions these methods were built for should treat its own numbers with a wider margin of uncertainty than the techniques' textbook presentation usually implies.

Select an organization you can examine in reasonable depth: one you work for or have worked for (disguised if needed), one documented in enough detail through a published case or annual report to analyze responsibly, or a clearly labeled hypothetical organization you construct for this assignment. Choose one job family within it that is plausibly facing either a shortage or a surplus over a one-to-three-year horizon.

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

1. describes the internal supply position for the chosen job family using at least two of the tools in this lesson, skills inventory, replacement chart, or a simplified Markov-style transition estimate, stating clearly which data you had access to and which you had to assume or estimate;
2. forecasts demand for the same job family using at least one quantitative technique (ratio-trend or productivity-ratio) and at least one judgmental technique (a described jury of executive opinion or a Delphi-style structured consultation), and states explicitly why you chose those methods for this job family rather than the alternatives;
3. quantifies the resulting gap, states whether it is a shortage or a surplus, and explains the single assumption in your supply or demand forecast that, if wrong, would change the gap's size or direction the most;
4. recommends a specific response or combination of responses (hiring, training, restructuring, attrition management, outsourcing) using the Direction/Response/Speed/Cost framework from this lesson's table, and justifies the choice against at least one alternative you did not choose; and
5. states one genuine limitation of your own analysis, drawing on Nkomo's (1987) finding or on the data-availability limitation discussed above, rather than presenting the recommendation as certain.

Use at least five academic or credible institutional sources, including at least three from this article's reference list. Where real organizational data is unavailable, label every substituted or hypothetical figure explicitly, exactly as this lesson's own illustrative chart and case are labeled, rather than presenting an estimate as a fact.

Assessment rubric (100 marks)

1. Bartholomew, D. J. (1967). Stochastic Models for Social Processes. John Wiley & Sons.
2. Bechet, T. P. (2008). Strategic Staffing: A Comprehensive System for Effective Workforce Planning (2nd ed.). AMACOM / Society for Human Resource Management.
3. Cascio, W. F., & Boudreau, J. W. (2010). Investing in People: Financial Impact of Human Resource Initiatives (2nd ed.). FT Press.
4. Dalkey, N., & Helmer, O. (1963). An experimental application of the Delphi method to the use of experts. Management Science, 9(3), 458–467.
5. Heneman, H. G., III, & Sandver, M. G. (1977). Markov analysis in human resource administration: Applications and limitations. Academy of Management Review, 2(4), 535–542.
6. Nkomo, S. M. (1987). Human resource planning and organization performance: An exploratory analysis. Strategic Management Journal, 8(4), 387–392.
7. U.S. Bureau of Labor Statistics. Employment Projections program: Labor force projections methodology. bls.gov/opub/hom/emp
8. Vroom, V. H., & MacCrimmon, K. R. (1968). Toward a stochastic model of managerial careers. Administrative Science Quarterly, 13(1), 26–46.
9. Walker, J. W. (1980). Human Resource Planning. McGraw-Hill.

Verification notes: every reference above was checked against publicly indexed bibliographic records (publisher and journal pages, Wiley Online Library, INFORMS PubsOnline, Google Books/WorldCat listings, and the Bureau of Labor Statistics' own published methodology pages) via web search rather than a full-text read of each source. Findings attributed to Bartholomew (1967), Vroom and MacCrimmon (1968), Heneman and Sandver (1977), Dalkey and Helmer (1963), Walker (1980), Bechet (2008), Cascio and Boudreau (2010), and Nkomo (1987) are reported at the level of detail available in abstracts, publisher summaries and secondary accounts. The Teesta Freight and Logistics case, the illustrative chart, the tables, and the synthesis connecting this lesson to Lessons 31 through 33 and forward to Lesson 35 are original teaching material for this series; every numeric figure in the chart, the Markov theory card's example, and the case is explicitly hypothetical and does not represent any real organization's data.

Lesson 35 picks up where this one stops: once a gap has been identified and a response chosen, how does an organization decide the right staffing level and structure to actually optimize workforce productivity going forward.