Career Development, Mentoring, Coaching and Learning ROI: Why Growing People and Proving the Payoff Are Two Different Problems

Career development, mentoring, coaching and ROI measurement look like four separate subjects, and in most textbooks they occupy separate chapters. They belong together here because each one concerns what happens to the relationship between a person's development and an organization's measured outcomes once the time horizon extends past a single course or workshop. A training needs assessment (Lesson 42) and an instructional design (Lesson 43) are built around a specific, bounded gap. Career development is not bounded in the same way; it is the accumulation of many such gaps, relationships and choices across years, and the tools organizations use to shape it, mentoring and coaching, work through mechanisms that have little to do with a classroom. ROI measurement, meanwhile, is the attempt to answer in financial terms whether any of this longer-term investment was worth making, which turns out to be a much harder question than evaluating whether a single three-day workshop improved a specific skill.

Donald Super's (1980) paper in the Journal of Vocational Behavior, "A life-span, life-space approach to career development," proposed that a career is not a single choice made once but a sequence of developmental stages tied to the evolving self-concept a person forms about who they are and what work they are suited for. The model has two dimensions. The life-span dimension lays out five stages, growth, exploration, establishment, maintenance and disengagement, each with a characteristic developmental task. The life-space dimension, visualized in Super's "Life-Career Rainbow," recognizes that a person holds several roles at once, worker, student, parent, citizen, and that these roles interact: strain or success in one affects capacity in the others. The combination was an advance over earlier trait-and-factor models that treated career choice as a single matching exercise between a person's traits and a job's requirements, because it treated career as a developmental process unfolding over decades rather than a one-time decision made in early adulthood.

The honest limitation of the model is in its age bands, not in its underlying insight that career tasks recur and evolve. Super's stages describe, reasonably well, the developmental tasks a career tends to present, exploring options, establishing competence, maintaining a position, winding down. What they describe less well today is the assumption, built into the fixed age ranges, of a relatively linear, often single-employer working life in which someone moves through each stage once, roughly on schedule. That assumption matched the mid-twentieth-century labor market Super was writing within reasonably closely. It matches a shrinking share of actual careers now, where people re-enter exploration in their forties after a layoff or a deliberate change of field, where establishment and maintenance can interleave repeatedly across several employers, and where a single person can sit in more than one stage at once depending on which role or project is examined. This is exactly the kind of organizational and economic shift that motivated the alternative model below, not a reason to discard Super's stages outright, but a reason to treat the ages as illustrative rather than diagnostic.

Douglas Hall introduced the term "protean career" in his 1976 book Careers in Organizations, borrowing the name from Proteus, the shape-shifting sea god of Greek mythology, to describe a career driven by the person rather than by the organization. The timing matters: Hall was writing as organizations began restructuring and downsizing in ways that made the stable, sequential, upward path implicit in stage models like Super's an increasingly unreliable promise for the people living inside those organizations. Hall revisited and developed the idea twice more, in a 1996 Academy of Management Executive piece naming it directly as a 21st-century phenomenon, and in a 2004 Journal of Vocational Behavior retrospective marking a quarter-century since the concept's introduction. Across these three pieces, the protean career settled into a specific claim: that a career's success criteria are defined by the individual, using their own values, rather than by an organization's promotion ladder, and that two psychological competencies make this self-direction workable rather than merely aspirational.

It is tempting to treat Hall's protean career as the corrected, modern replacement for Super's outdated stage model, and some practitioner writing does exactly that. A more accurate reading is that the two models describe different axes rather than competing answers to the same question. Super's stages describe a temporal sequence of developmental tasks, what a person is typically working through at a given point, regardless of how they feel about whose criteria define success. Hall's protean orientation describes a psychological stance toward that question of whose criteria matter, largely independent of age or tenure. A forty-five-year-old employee sitting squarely in Super's maintenance band by age can still hold a strongly protean orientation, actively redefining what counts as success on her own terms rather than settling into the role the organization assigned her. A twenty-eight-year-old in Super's exploration band can, just as plausibly, hold a strongly organization-dependent orientation, waiting for a manager to define the next step rather than driving it herself. Treating the two as commensurate alternatives to pick between misses that they are answering different questions, and treating either one as a universal prescription for how every employee's career should work, which Rule 14 of this series flags directly, ignores how differently career structures vary across industries, seniority levels and the specific labor market an organization operates in. The practical use of both models for HR is as a vocabulary for a career conversation: where is this person in the developmental task sequence, and how much is this person driving that sequence themselves versus expecting the organization to drive it for them. Neither question has a single correct answer that applies to everyone in a given role.

Kathy Kram's (1983) article "Phases of the Mentor Relationship," published in the Academy of Management Journal, and her 1985 book Mentoring at Work: Developmental Relationships in Organizational Life, gave the field its most durable framework for what a mentor actually does. Kram's research, based on interviews with pairs of managers in organizational settings, found that mentoring relationships deliver two distinct kinds of benefit to the person being mentored, which she grouped as career functions and psychosocial functions. The distinction matters because the two function sets serve different needs and can exist independently of each other: a mentor can provide strong psychosocial support with weak career functions, or vice versa, and a mentee experiencing both is getting something meaningfully larger than either alone.

Caroline D'Abate, Erika Eddy and Scott Tannenbaum's (2003) review in Human Resource Development Review, "What's in a name? A literature-based approach to understanding mentoring, coaching, and other constructs that describe developmental interactions," took on the overlap Kram's own framework leaves open and tried to separate mentoring, coaching and several related terms along a consistent set of dimensions rather than by definition alone. Their review is a useful corrective precisely because it does not pretend the terms are cleanly separated in practice; it states plainly that these constructs are often used interchangeably even though they are theoretically distinguishable, which is the more honest starting point than treating either term as settled vocabulary. The dimensions that do the real separating work are goal specificity, time horizon and the kind of expertise the relationship assumes. Mentoring tends toward a broad, identity-level goal, long-term professional and personal growth, over an open-ended time horizon, and typically assumes the mentor has walked a reasonably similar path. Coaching tends toward a specific, bounded performance or skill goal, over a defined time horizon with a clear endpoint, and does not require the coach to have held the coachee's exact role; a coach's expertise is often in the coaching process itself, asking questions, structuring practice, giving behavioral feedback, rather than in having already done the coachee's job.

Lesson 44 covered Kirkpatrick's four levels of training evaluation, ending at Level 4, organizational results: did the training show up in a business measure the organization actually tracks, output, quality, retention, revenue. Jack Phillips, in Return on Investment in Training and Performance Improvement Programs (first published 1997, revised in a second edition in 2003), proposed a fifth level sitting directly on top of Kirkpatrick's four. Level 5 takes the business-impact data already collected at Level 4 and converts it into a monetary benefit figure, nets that figure against the fully loaded cost of delivering the program, and expresses the result as a return-on-investment percentage, the same metric used to judge a capital investment: ROI percent equals the training benefit minus the training cost, divided by the training cost, multiplied by one hundred. Read quickly, this looks like a straightforward accounting exercise bolted onto an evaluation model that was already business-focused at Level 4. It is not, because before any business-impact figure can be converted into money and compared to cost, something has to establish how much of that business-impact figure the training program is actually responsible for, as opposed to everything else happening in the organization at the same time.

Phillips's own methodology names this step explicitly and treats it as separate from, and prior to, the monetary conversion: isolating the effects of the training from the many other factors that typically influence a business measure over the same period, a market upturn, a new incentive scheme, a change in supervision, ordinary regression toward the mean after an unusually weak quarter, or a selection effect in which the kind of employee who volunteers for a leadership program was already more likely to improve regardless of the program. This is the same correlation-versus-causation caution this series applies throughout (Rule 13): a business measure moving in the right direction after a training program ran is not, by itself, evidence the training caused the movement. Phillips's own published guidance, "Measuring ROI: The Process, Current Issues, and Trends," lists five techniques organizations use to tackle this step, each with a different reliability profile.

What is worth stating plainly, because it is easy to miss inside a confident-looking percentage, is that Phillips's own guidance already concedes the problem this table implies. One of the ROI methodology's own stated guiding principles is that estimates of improvement should be discounted for the potential error of the estimate, precisely because an estimation-based isolation is understood, by the people who built the method, to be less reliable than a controlled comparison. In practice, the overwhelming majority of published ROI case studies, including the ones Phillips's own organization circulates as evidence of the method's value, rely on participant or supervisor estimation rather than a genuine control group, because running a true control group inside an active business, holding back development from a comparable set of employees to create a clean comparison, is usually impractical. That leaves a real tension sitting inside the methodology itself: the technique that is most defensible on methodological grounds is the one least usable in practice, and the technique most used in practice is the one that carries the most uncertainty about whether it is isolating the training's actual effect or mostly capturing what the rater already believed about the program before being asked.

Reid Bates's (2004) critique of the Kirkpatrick model, published in Evaluation and Program Planning, argued that the model understates how much contextual factors, organizational culture, supervisor support, the adequacy of tools and equipment back on the job, shape whether training actually produces a result at all. That critique extends one level further up when a Level 4 result gets isolated into a Level 5 ROI figure. If workplace context partly determines whether learning transfers into a business outcome in the first place, which is exactly the transfer-climate problem Lesson 44 examined, then the "non-training" share of any observed improvement is not a fixed, cleanly separable background noise that an isolation technique can simply subtract out. It is entangled with the training itself: a supervisor who reinforces what a program taught is very often the same supervisor running other initiatives competently at the same time, which makes the two influences genuinely difficult to pull apart even with a careful estimation exercise, let alone a quick percentage on a board slide.

Lesson 41 established why training is designed around how adults actually learn rather than how a classroom happens to be organized. Lesson 42 determined whether a business problem is really a training problem at all, which matters directly to the material in this lesson: an ROI isolation estimate can mistakenly credit a training program for an improvement that was never a training-solvable gap in the first place, if the needs assessment that justified the program was weak or skipped. Lesson 43 built the program itself against stated objectives, the thing Level 2 learning measures are checked against. Lesson 44 covered whether that learning actually transfers back to the job, which is the mechanism, if one exists at all, through which a Level 4 business result and therefore any Level 5 ROI claim could legitimately occur. Career development, mentoring, coaching and ROI, this lesson's four subjects, sit at the far end of that chain. Career development and the distinct work mentoring and coaching do are about sustaining and directing an employee's growth over a horizon far longer than any single needs-assessment cycle. ROI is the attempt to put one number on whether any of it, the whole chain, was worth what it cost.

That leaves a genuine tension this module does not resolve, and it is worth stating directly rather than smoothing it into reassurance. A board allocating next year's learning and development budget will keep asking for a single confident ROI number, because budgets get allocated against numbers that sound settled, not against honestly hedged ranges. The most defensible answer the measurement literature reviewed in this lesson can currently offer is that isolating training's specific causal contribution from everything else happening in a business at the same time is a hard, often only partially solvable problem, one Phillips's own methodology acknowledges through its estimation-discounting principle rather than pretends away. An appropriately hedged ROI estimate, stated with its isolation technique and that technique's specific limitation named plainly, is the more defensible report on the evidence. It is also the one that looks less confident next to a competitor's dashboard showing a clean, unqualified percentage. Nothing in this lesson resolves which an individual L&D function should choose to present; it only establishes, as plainly as the evidence allows, which choice the evidence itself actually supports.

Select a real or carefully researched role inside an organization you can examine in reasonable depth, your own employer, a former employer, or a company studied through public filings, news coverage and employee reviews, disguised if needed to protect those involved.

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

1. reads the role's incumbent using both Super's stage framing and Hall's protean-orientation framing, stated explicitly as two different lenses answering different questions rather than one chosen as the correct model;
2. identifies a specific development need for that person and recommends mentoring, coaching, or both, justified against Kram's (1983) career-versus-psychosocial functions and the D'Abate, Eddy and Tannenbaum (2003) distinctions, not a default choice made without that analysis;
3. proposes an ROI evaluation plan for a training program relevant to that role, naming which isolation technique from this lesson's table you would use, why it fits this specific organizational setting, and what that technique cannot rule out even if applied well;
4. states explicitly, in a closing section, how your report would present uncertainty to a board that wants a single confident number, and defends that choice against the alternative.

Use at least five academic or documented sources, including at least three from this article's reference list. State explicitly where organizational information is unavailable or disguised, and how that limits the specificity of your plan.

Assessment rubric (100 marks)

1. Bates, R. (2004). A critical analysis of evaluation practice: The Kirkpatrick model and the principle of beneficence. Evaluation and Program Planning, 27(3), 341–347.
2. D'Abate, C. P., Eddy, E. R., & Tannenbaum, S. I. (2003). What's in a name? A literature-based approach to understanding mentoring, coaching, and other constructs that describe developmental interactions. Human Resource Development Review, 2(4), 360–384.
3. Hall, D. T. (1976). Careers in Organizations. Goodyear Publishing.
4. Hall, D. T. (1996). Protean careers of the 21st century. Academy of Management Executive, 10(4), 8–16.
5. Hall, D. T. (2004). The protean career: A quarter-century journey. Journal of Vocational Behavior, 65(1), 1–13.
6. Kram, K. E. (1983). Phases of the mentor relationship. Academy of Management Journal, 26(4), 608–625.
7. Kram, K. E. (1985). Mentoring at Work: Developmental Relationships in Organizational Life. Scott, Foresman.
8. Phillips, J. J. (2003). Return on Investment in Training and Performance Improvement Programs (2nd ed.). Butterworth-Heinemann.
9. Phillips, J. J. (2007). Measuring ROI: The Process, Current Issues, and Trends [Updated article, condensed from Phillips, 2003]. ROI Institute.
10. Super, D. E. (1980). A life-span, life-space approach to career development. Journal of Vocational Behavior, 16(3), 282–298.

Verification notes: Super (1980), Hall (1976, 1996, 2004), Kram (1983, 1985) and D'Abate, Eddy and Tannenbaum (2003) were checked against publisher, journal-index or ERIC bibliographic records via web search. Phillips's ROI methodology and its isolation-of-effects techniques, including the discount-for-estimation-error guiding principle, were verified directly against the full text of Phillips's own "Measuring ROI: The Process, Current Issues, and Trends," fetched and read in full rather than taken from a secondary summary. Bates (2004) was checked against its EconPapers and Semantic Scholar bibliographic record. The Protik Software Services case, the tables, the decision scenario, and the synthesis connecting this lesson to Lessons 41 to 44 are original teaching material for this series.

This closes Module 09 of The Human Capital MBA. Module 10 turns to Performance Management, beginning with performance management systems: concepts, models and design.