Learning, Skills, Careers and Internal Mobility Analytics

Introduction

Organizations increasingly compete through what their people can learn and do. Traditional human resource management often records course attendance, job titles and promotion counts. People analytics asks more demanding questions: Which capabilities will the organization need? Where are the largest skill gaps? Does learning improve performance or retention? Which career paths are realistic? Are internal opportunities distributed fairly, and do internal moves improve organizational outcomes?

Learning, skills, careers and internal mobility analytics connects workforce data with decisions about capability development and talent allocation. It combines human capital theory, learning evaluation, workforce planning, organizational network analysis and responsible data governance. The purpose is not to reduce people to scores. Rather, it is to improve the quality, fairness and transparency of decisions affecting employees and organizational performance.

Learning objectives

By the end of this chapter, students should be able to:

1. Explain the relationship among skills, capabilities, learning and business strategy.
2. Construct a practical skills taxonomy and identify capability gaps.
3. Apply established learning-evaluation models and distinguish association from causation.
4. Estimate training impact and return on investment using appropriate assumptions.
5. Analyze career paths, succession pipelines, internal talent marketplaces and mobility outcomes.
6. Identify ethical, legal, privacy and measurement risks in people analytics.

1. From jobs to skills and capabilities

A job is a bundle of tasks performed under an organizational role. A skill is a learned capacity to perform a task or activity. A competency is a broader combination of knowledge, skill, behavior and judgment applied in context. A capability is an organizational capacity created by combining people, processes, technology and knowledge. For example, data visualization is a skill; analytical problem solving may be a competency; evidence-based commercial decision-making may be an organizational capability.

Skills can be classified in several ways:

• Technical skills: programming, accounting, statistical analysis or equipment operation.
• Human skills: communication, negotiation, collaboration and leadership.
• Domain skills: knowledge of a sector, product, customer group or regulation.
• Transferable skills: capabilities useful across multiple roles.
• Digital and data skills: data literacy, cybersecurity, automation and artificial intelligence use.

A skills taxonomy is a structured vocabulary that defines skills, groups related skills and describes proficiency levels. It is not merely a list of keywords. A useful taxonomy contains clear definitions, synonyms, relationships among skills and observable evidence for proficiency. It should distinguish a skill from a credential, job title and personality trait.

Building a skills taxonomy

A practical construction process is:

1. Clarify strategic purpose. Link the taxonomy to products, services, customer needs, regulatory obligations and future workforce scenarios.
2. Collect evidence. Review job descriptions, work outputs, occupational standards, learning curricula, project records and structured interviews with subject-matter experts.
3. Normalize terminology. Combine synonyms such as “data visualization” and “dashboard design,” while retaining meaningful distinctions.
4. Define proficiency. Describe observable performance, not vague labels such as “basic” or “advanced.”
5. Validate. Ask managers and experienced practitioners whether the taxonomy reflects real work.
6. Govern change. Assign ownership, version definitions and review the taxonomy as technology and work change.

A four-level scale may be appropriate for an undergraduate application:

Proficiency levels must not be treated as universally comparable. Level 3 in spreadsheet modeling is not automatically equivalent to level 3 in negotiation. Evidence should be role-specific, and self-assessments should be supplemented by demonstrations, work samples, validated assessments or manager observation.

Skills graphs and ontologies

A taxonomy is hierarchical; a skills graph represents relationships among skills, roles, tasks, learning resources and projects. An ontology may specify that one skill is a prerequisite for another or that two terms are equivalent. These structures support search and recommendation systems, but automated similarity is not the same as professional equivalence. A model may identify linguistic proximity between “sales forecasting” and “financial forecasting” while missing important domain differences.

2. Capability-gap analysis

A capability gap is the difference between required future capability and available current capability. The comparison may occur at employee, team, function or enterprise level.

A simple gap score for skill s and employee i is:

Here, R_is is the required proficiency and C_is is the current proficiency, measured on the same validated scale. A positive score indicates a shortfall; zero indicates alignment; a negative score indicates capacity above the stated requirement. The formula assumes that proficiency levels are sufficiently comparable for subtraction and that requirements are explicitly defined. Because ordinal ratings are not always equal-interval measurements, the result should usually be interpreted as a prioritization indicator rather than a precise quantity of missing skill.

At team level, a weighted average gap can be calculated as:

where n is the number of employees, w_i is the importance or exposure weight for employee i, and G_is is the individual gap. Weights should have a defensible meaning, such as proportion of critical work performed, not be chosen merely to produce a desired result. This metric can conceal concentration risk: an average gap may look acceptable even though a critical unit has no proficient employees.

A stronger diagnostic combines:

• Demand: skills required by strategy, workload and future scenarios.
• Supply: verified skills currently available.
• Criticality: consequence of the capability failing.
• Time horizon: how soon the skill is needed.
• Acquisition options: build, buy, borrow, automate or redesign work.
• Equity: whether access to development is fairly distributed.

The “build, buy, borrow and bot” language is useful as a strategic shorthand, but automation should be assessed for quality, safety, job redesign and employee impact rather than assumed to be a simple substitute for people.

3. Learning analytics and evaluation

Learning analytics uses data about learners, learning experiences and subsequent outcomes to improve learning and decisions. Typical data include enrollment, completion, assessment results, practice activity, feedback, time to competence, work quality, productivity, safety and retention. A click or completion is an activity measure, not proof of learning or business value.

Kirkpatrick’s four-level model

Kirkpatrick’s model organizes evaluation into four levels:

1. Reaction: participants’ perceived relevance, usability and engagement.
2. Learning: change in knowledge, skill or attitude, ideally measured through valid assessments or demonstrations.
3. Behavior: transfer of learning to workplace behavior.
4. Results: effects on organizational outcomes such as quality, sales, safety or customer service.

The levels are logically related but not guaranteed to occur sequentially. Positive reactions do not prove learning, and learning does not automatically transfer when managers, systems or incentives discourage application.

Phillips’ ROI extension

Phillips adds a financial return level. The basic return on investment calculation is:

where B is monetized benefit and C is total program cost. Costs may include design, instructors, technology, participant time, travel, administration and evaluation. Benefits may include incremental contribution, avoided defects, reduced accidents or reduced turnover. Benefits should be stated in a common currency and time period. ROI is not meaningful if benefits are merely revenue with no consideration of margin, or if costs omit substantial participant time.

A benefit–cost ratio is different:

A BCR of 1.5 means that each monetary unit of cost is associated with 1.5 monetary units of benefit; ROI is 50%, not 150%.

CIPP and learning-transfer perspectives

The CIPP framework evaluates Context, Input, Process and Product. It is useful when the decision is not simply whether a course “worked,” but whether the organization selected the right problem, designed an adequate intervention, implemented it well and achieved useful outcomes.

Transfer theory emphasizes that learning depends on opportunity to apply, managerial support, feedback, psychological safety, job design and reinforcement. Therefore, a weak post-training result may reflect a work system problem rather than ineffective instruction.

Experimental and quasi-experimental designs

The strongest practical evaluation compares outcomes with a credible counterfactual—the outcome that would have occurred without the intervention. A randomized design assigns eligible participants to treatment and control groups, subject to ethical and operational constraints. A pre-post design without a comparison group is easier but cannot distinguish training effects from seasonal changes, new technology or general performance trends.

A difference-in-differences estimator compares changes over time between treatment and comparison groups:

Y_T,post and Y_T,pre are treatment-group outcomes after and before learning; Y_C,post and Y_C,pre are comparison-group outcomes. The key assumption is parallel trends: absent training, both groups would have changed similarly. Analysts should examine pre-intervention trends and document selection, attrition, interference and other concurrent interventions. The estimate is an association with causal interpretation only when design assumptions are credible.

Worked example: estimating training return

A customer-service program costs $48,000, including delivery and participant time. After a credible comparison, the estimated annual benefit is $78,000, consisting of avoided rework and additional contribution margin. Then:

The BCR is 78,000 ÷ 48,000 = 1.625. The result is an estimate, not a fact. Analysts should test sensitivity to alternative attribution percentages, benefit duration, discounting, implementation costs and measurement error. If only 60% of the estimated benefit is attributable to the program, the adjusted benefit is $46,800 and ROI becomes −2.5%: (46,800 - 48,000) ÷ 48,000 × 100 = -2.5\%. This illustrates why attribution is more important than presenting a precise-looking headline number.

4. Career analytics and succession

A career is a sequence of work experiences, roles, learning opportunities and transitions. Career analytics examines movement patterns and the conditions associated with progression, lateral development, retention and exit. It should not assume that upward promotion is the only successful career outcome. Expert, project-based, flexible and lateral pathways may create value for both employees and organizations.

Useful measures include:

• time in role and time to proficiency;
• promotion, lateral-move and internal-transfer rates;
• internal fill rate for vacancies;
• retention after a move;
• pay and level changes after mobility;
• skill acquisition and application;
• diversity of candidate slates and outcomes;
• succession coverage for critical roles.

A basic internal fill rate is:

The numerator and denominator must refer to the same population and period. A high rate is not automatically positive if it reflects restricted external hiring, poor quality matches or excessive movement from already understaffed teams.

Career-path analysis

Career paths can be modeled as a transition matrix. If there are roles A, B and C, the probability of moving from role i to role j during a defined period is:

where N_ij is the number of observed transitions from i to j, and N_i is the number of people at risk of transitioning from i. The denominator must exclude people not observable for the full risk period when appropriate. Each row of a complete transition matrix sums to 1 only if all destinations—including remaining in role, exit and other states—are included.

Transition analysis can reveal bottlenecks, common feeder roles and inequitable access to opportunity. However, historical transitions describe what happened, not what should happen. A model trained on historical promotions can reproduce biased manager preferences, unequal sponsorship and occupational segregation.

Succession pipelines

Succession analytics identifies potential readiness for critical roles and the development actions needed to improve coverage. A succession pipeline should include readiness definitions such as “ready now,” “ready within one to two years” and “longer-term potential,” supported by evidence. It should also monitor pipeline diversity, single-incumbent risk, development activity and replacement time.

Potential ratings should not be confused with predictions of an employee’s worth. They are judgments under uncertainty and should be calibrated, documented and reviewed for consistency. Employees need meaningful development conversations, not opaque labels that limit future opportunities.

5. Internal talent marketplaces and mobility outcomes

An internal talent marketplace matches employees to projects, mentoring, learning, gigs and vacancies using skills, interests, availability and eligibility. It can make hidden skills visible and reduce dependence on informal networks. Effective marketplaces combine algorithmic recommendations with human review, employee choice and transparent criteria.

A mobility outcome should be assessed at several levels:

• Employee: skill growth, engagement, career progression, pay, workload and retention.
• Receiving team: time to contribution, quality, manager satisfaction and delivery.
• Releasing team: continuity, backfill burden and knowledge transfer.
• Organization: internal fill, vacancy duration, capability coverage and performance.
• Equity: access, selection rates and outcomes across relevant groups.

A movement is not necessarily beneficial merely because it is internal. Poorly managed transfers can create skill loss, workload inequity and “talent hoarding,” in which managers obstruct mobility. Governance should establish release principles, service-level expectations, transparent eligibility and shared accountability.

Illustrative case: Northstar Services

Background

Northstar Services, a fictional regional business-process company, planned to introduce automation and analytics across customer operations. It had 1,200 employees, rising external hiring costs and uneven access to development. Executives proposed a data-literacy academy and an internal project marketplace.

Problem

The first proposal measured success by enrollment and completion. Analysts identified three weaknesses: the skills vocabulary mixed job titles with skills; self-ratings overstated proficiency; and no comparison group had been planned. Women and part-time employees were also less likely to be nominated for development.

Analysis

A cross-functional team built a taxonomy covering data interpretation, process mapping, automation supervision, communication and domain knowledge. Each skill had four behaviorally anchored levels. Job and project requirements were mapped to the taxonomy. A baseline assessment combined a practical task, structured manager evidence and employee self-report.

The team then identified high-priority gaps by considering strategic importance, current supply, time to need and operational risk. Employees were invited to apply for academy places rather than relying solely on manager nomination. A matched comparison group was established for evaluation, with attention to pre-program performance and work type.

Approach and solution

The academy used short lessons, practice on real operational data, peer feedback and manager-supported application. The marketplace advertised time-bounded projects with explicit skills, workload and selection criteria. Recommendations were generated from skills and interests, but employees could inspect and correct their profiles. Managers retained responsibility for work allocation and safety.

Evaluation covered reaction, assessment performance, workplace application and operational results. The analysts tracked completion, practical assessment gains, quality defects, time to competence, internal moves and retention. They also examined participation and outcomes by gender, employment status, function and accessibility needs, while applying minimum-cell rules to protect confidentiality.

Outcome

After nine months, assessment results improved more in the academy group than in the comparison group, and project participants showed improved process-quality indicators. The organization reported lower external hiring pressure for selected roles. However, participation remained lower among employees with inflexible schedules. Northstar responded with paid learning time, repeated cohorts and improved manager guidance. The evaluation concluded that the program contributed to outcomes but could not attribute all operational change to training because automation implementation occurred concurrently.

Learning points

The case demonstrates that a taxonomy is a governance asset, not merely a technology feature; access design affects measured impact; marketplace recommendations require human accountability; and credible evaluation includes both benefits and limitations.

6. Technology, AI and data architecture

A mature architecture links human-resource, learning, work and outcome data through governed identifiers. Common components include a skills repository, learning-management system, human-resource information system, performance or quality systems, project marketplace and survey platform. Data dictionaries should define fields, ownership, refresh rates, permissible uses and retention periods.

AI can assist with skill extraction from job documents, recommendation, assessment feedback and detection of unusual mobility patterns. Risks include proxy discrimination, feedback loops, hallucinated skill inferences, surveillance, opacity and automation bias. Natural-language similarity should not be treated as proof of competence. High-impact employment decisions require explainability appropriate to the decision, human review, contestability and monitoring after deployment.

7. Ethics, law and governance

Learning and mobility data can affect access to work, promotion and income. Good governance therefore requires:

• purpose limitation and data minimization;
• informed notice and, where required, a lawful basis for processing;
• accuracy, correction and access mechanisms;
• role-based access and security;
• retention limits and secure deletion;
• impact assessments for high-risk analytics;
• fairness testing across relevant groups;
• human review and appeal;
• documentation of models, assumptions and decisions.

Legal requirements differ by jurisdiction. Analysts must consult applicable privacy, employment-discrimination, accessibility, labor and records laws rather than assuming that an employee’s access to a system implies unrestricted permission for secondary analysis. Small-group reporting can reveal identity; aggregation and suppression are often necessary.

Fairness analysis should examine the full process: who is invited, who applies, who is selected, who completes, who receives a project and who benefits. A parity statistic alone cannot establish fairness because legitimate differences in eligibility, experience or availability may exist. Conversely, apparently neutral criteria can reproduce historical disadvantage. Stakeholder consultation and qualitative evidence are essential complements to metrics.

Best-practice implementation sequence

1. Start with a strategic decision, not a dashboard.
2. Define skills and proficiency through observable evidence.
3. Combine multiple data sources and assess measurement validity.
4. Separate descriptive, predictive and causal claims.
5. Design a counterfactual before launching a program.
6. Measure transfer conditions, not only learner behavior.
7. Report uncertainty, sensitivity and limitations.
8. Build employee voice, correction and appeal into systems.
9. Monitor equity and unintended consequences.
10. Review taxonomies, models and outcomes as work changes.

Chapter Summary

Learning, skills, careers and internal mobility analytics helps organizations align human capability with strategic demand. Skills taxonomies provide a shared vocabulary, while capability-gap analysis compares required and available proficiency. Learning evaluation should move beyond attendance and satisfaction to learning, transfer and organizational results. Kirkpatrick, Phillips and CIPP offer complementary perspectives, while randomized and quasi-experimental designs improve causal credibility. ROI requires complete costs, defensible benefit estimates and explicit attribution.

Career and mobility analytics examines transitions, pathways, succession readiness and internal talent marketplaces. Measures such as internal fill rate and transition probabilities are useful only when their populations, time periods and assumptions are clear. Historical data can reproduce inequity, so responsible systems require validation, transparency, privacy, human oversight and employee appeal. The central professional principle is to use analytics to expand informed opportunity and organizational learning—not to create opaque rankings that constrain people.

Key Terms

Skills taxonomy: Structured vocabulary that defines and organizes skills.

Competency: Combination of knowledge, skills, behaviors and judgment applied in context.

Capability: Organizational capacity created through people, processes, technology and knowledge.

Proficiency: Demonstrated level of performance in a skill.

Capability gap: Difference between required and available capability.

Learning analytics: Analysis of data about learning activities, learners and outcomes.

Transfer: Application of learning in the workplace.

Kirkpatrick model: Four-level framework covering reaction, learning, behavior and results.

CIPP: Evaluation framework covering context, input, process and product.

ROI: Net monetized benefit divided by cost, expressed as a percentage.

Counterfactual: Estimated outcome that would have occurred without an intervention.

Difference-in-differences: Comparison of changes over time between treatment and comparison groups.

Career path: Sequence of roles and experiences through which work and development progress.

Succession pipeline: Population and development process supporting replacement of critical roles.

Internal fill rate: Share of defined vacancies filled by internal candidates.

Internal talent marketplace: System matching employees with internal work, learning or career opportunities.

Transition matrix: Table of probabilities or counts describing movement among states.

Algorithmic bias: Systematic disadvantage produced or amplified by data, design or deployment.

Talent hoarding: Managerial behavior that obstructs employee movement to other opportunities.

Review Questions

Short Answer Questions

1. Distinguish a skill, competency and organizational capability.
2. Why should proficiency levels be supported by observable evidence?
3. What are the four levels in the Kirkpatrick model?
4. Define a counterfactual in learning evaluation.
5. What is an internal fill rate?

Long Answer Questions

1. Explain how an organization can construct and govern a skills taxonomy.
2. Compare Kirkpatrick, Phillips and CIPP as approaches to learning evaluation.
3. Discuss why training completion is an inadequate measure of business impact.
4. Explain how transition matrices can be used in career-path analysis and identify their limitations.
5. Describe the ethical and governance requirements for an AI-enabled internal talent marketplace.

Analytical / Application Questions

1. A program costs $60,000 and produces an estimated attributable benefit of $84,000. Calculate ROI and BCR, then explain two assumptions that should be tested.
2. Design a quasi-experimental evaluation for a data-literacy academy when random assignment is not operationally feasible.
3. An organization has a high internal promotion rate but low diversity in senior roles. Propose an analytics investigation covering access, selection, mobility and outcomes.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes a competency from a skill?

A. A competency is always a formal qualification

B. A competency combines knowledge, skills, behavior and judgment in context

C. A competency is a synonym for a job title

D. A competency measures only technical ability

Correct Answer: B. A competency combines knowledge, skills, behavior and judgment in context

Question 2

If required proficiency is 4 and current proficiency is 2 on the same scale, what is the gap score G = R - C?

A. −2

B. 0

C. 2

D. 8

Correct Answer: C. 2

Question 3

Which Kirkpatrick level examines application of learning in the workplace?

A. Reaction

B. Learning

C. Behavior

D. Results

Correct Answer: C. Behavior

Question 4

A program costs $40,000 and has an attributable benefit of $50,000. What is its ROI?

A. 10%

B. 20%

C. 25%

D. 125%

Correct Answer: C. 25%

Question 5

What is the key identifying assumption in a difference-in-differences design?

A. Treatment and comparison groups have identical sizes

B. Both groups would have followed parallel trends without treatment

C. All participants complete the intervention

D. The outcome is measured only after treatment

Correct Answer: B. Both groups would have followed parallel trends without treatment

Question 6

Which measure is most directly concerned with filling vacancies from within the organization?

A. Internal fill rate

B. Completion rate

C. Reaction score

D. External turnover rate

Correct Answer: A. Internal fill rate

Question 7

Why can historical promotion data create risk in a predictive succession model?

A. Historical data never contains numerical values

B. The model may reproduce earlier unequal access and biased judgments

C. Promotion data cannot be stored securely

D. Succession models do not use employee data

Correct Answer: B. The model may reproduce earlier unequal access and biased judgments

Question 8

Which practice best supports responsible use of employee skills data?

A. Collect every available data field indefinitely

B. Allow recommendations to determine decisions automatically

C. Use purpose limitation, access controls and employee correction mechanisms

D. Hide evaluation criteria from employees

Correct Answer: C. Use purpose limitation, access controls and employee correction mechanisms

Question 9

A benefit–cost ratio of 1.5 means that:

A. ROI is 1.5%

B. Benefits equal 1.5 times costs

C. Costs equal 1.5 times benefits

D. The program lost 50% of its cost

Correct Answer: B. Benefits equal 1.5 times costs

Question 10

What is a major purpose of an internal talent marketplace?

A. To eliminate all managerial judgment

B. To match employees with internal projects, learning or roles

C. To replace skills definitions with job titles

D. To guarantee promotion after course completion

Correct Answer: B. To match employees with internal projects, learning or roles