Peter Cappelli’s “Talent On Demand”

There is a compelling scene in HBO’s quasi-fictional Western, Deadwood, which qualifies as an instant lesson in the essentials of talent management. The 1870s mining boom town of Deadwood, which is just emerging from Wild West state-of-nature conditions, has attracted the attention of the robber baron George Hearst. Al Swearengen, the town’s incumbent strongman, adapts nimbly and switches from a governance-by-murder strategy to a more artful one, and in the process very effectively shifts power between his two key reports to reflect the priorities of the new situation. Dan Doherty, his dim and violent second from the lawless past, is gracefully shunted aside, while the more suave, but restless and underutilized Silas Adams is handed the tricky and critical “stretch” task of managing the relationship with George Hearst. Somehow, Al keeps both men motivated and loyal. Modern executives will recognize all sorts of very current themes in this little vignette. So how do you master these timeless elements of talent management, while operating in modern business conditions? You read Talent On Demand by Wharton’s Peter Cappelli, a book that completely validates my belief that talent management is the issue of the next decade (the purchasing function ranks a close second).

The Problem of Talent Management Defined

What makes talent management possibly the hardest, and most important problem in management is that it involves entities — people — who must at once be treated as unique one-of-a-kind items that need individualized handling, and as standardized parts in a supply chain feeding an abstract organizational structure. Consider the issues highlighted in the Deadwood vignette above:

  • There is the need to focus talent development efforts on current strategic priorities
  • There is the need to treat a structural position (second-in-command) as a plug-in spot for an abstract “part” of a machine, to be replaced as necessary by a waiting inventoried item in the supply chain
  • There is the need to play to people’s strengths
  • There is the need to maintain motivation and loyalty while playing chessmaster
  • There is the need to keep an eye on retention issues
  • There is the need to grow employees through developmental experiences
  • There is the need to assume some risk and control costs by allowing mission-critical tasks to be used for development purposes (otherwise you lose too much money on expensive “pure development” assignments)

The Book

This isn’t an easy set of challenges. The solution, in my mind, comprises three parts — individually-tailored local management practices, which I briefly covered in my review of the strengths movement, analytics-driven impersonal and global practices, which this book covers, and methods to balance the imperatives of strategy (the “pull”) and the capabilities of your talent (the “push”), which I’ll cover in my next article in this series.

The book is in two parts. The stand-out excellent Part I covers the historical development of talent management, from the robber baron era to modern times, and concludes with a careful and pragmatic framing of the problem. Part II, which covers four competencies that comprise Cappelli’s “On-Demand” model, is more patchy. The first two chapters of Part II, covering the build-vs.-buy issue and the problem of uncertain demand, are excellent. The last two chapters, on managing talent ROI and internal markets, are between weak to competent.

The Highlights

Maybe I am biased, because I am partial to history, but the best part of the book is the history bit, which manages the impossible task of being both magisterial and succinct. We get a whistle-stop tour starting from the early days, when railroads were the best-practice leaders and everybody else poached from the Pennsylvania Railroad. The story proceeds to the diffusion of inspired ad hoc models, largely invented by famous names ranging from Dupont and Carnegie to Scwhab and J. P. Morgan, and carried throughout the economy by their proteges. Next we get a careful treatment of the growth and professionalization of talent management. We learn about the evolution of training models, rotational assignments, high-potential programs, succession planning, 360 degree reviews. and prediction models models at places like GE, Dupont, AT&T and GM. The story concludes with the great collapse of traditional talent management in the eighties. Though the story is US-focused, international vignettes are thrown in occasionally.

What is really interesting about the way the story is told, is the careful inter-twining of the natural logic of a maturing organizational function and the realities of the backdrop of world history. We get quick but comprehensive treatments of the impact of the great depression, World War II, deregulation and outsourcing. There is also a layer of abstraction, where we learn how the long-term cyclical dynamics of the economy drove talent management practices back and forth on various critical issues, such as inside-vs.-outside hiring, centralization vs. decentralization and functional vs. generalist training. And that only scratches the surface — there are a dozen cyclical and monotonic trends in this story.

But this isn’t a story told for the sake of story-telling. It is a story told to frame the problem a particular way. The framing is this: traditional models of talent management are based on assumptions of long-term market predictability and capacity planning. Yes, we all recognize that these conditions no longer hold, but what Cappelli really manages to underline is that we need to unlearn and relearn the assumptions as encoded in operating practices. It is not sufficient to unlearn and relearn as individuals, which we’ve all already done (okay, the ones among us who actually bother to think). An example that really highlights this is AT&T — decades after the antitrust legislation broke Ma Bell apart, the talent management function still operated the same way, because nobody had bothered to tease apart fundamentally true assumptions from situational ones.

From Part II, I’ll pick out only one highlight, but it is an important one. Cappelli notes that talent management models basically fell from a peak of sophistication in the 60s to “nothing” today (or worse, a 60s hangover that is worse than nothing). He correctly notes that you don’t have to really solve the whole issue at once. You just have to do “better than nothing” to show marked improvements. An example of this is his pointer that simply changing capacity planning to recognize the asymmetric relative costs of undershooting versus overshooting demand (i.e. hiring too many versus too few) will result in significant improvement. Currently, he recommends attempting to undershoot, due to the lower cost of buy vs. build. This is a simple shift from point thinking to “weighted spread” thinking, and he is right. All you need is a few such roughly correct moves in the right direction. You don’t need a huge overhaul that attempts to fix the whole problem at once.


Part I, like I said, is basically a history, and it is impossible to summarize, since it is extraordinarily dense (though a crisp and easy read). Part II has 4 chapters that together constitute what Cappelli calls an “On-Demand” model, which is based largely on stochastic supply-chain thinking:

  • Chapter 4 covers the make-versus-buy decision (internal development versus external hiring) and concludes with the succinct and intelligent first-step recommendation I mentioned: move from point thinking to weighted spread thinking. For example, supposes it is cheaper to hire than to develop talent internally, and you need a 100 programmers. Let’s say hiring, retention and development uncertainties lead you to believe that your system will produce between 80 to 120 programmers. Cappelli’s model would suggest actually targeting an undershooting policy (i.e. set your system up to produce 90 programmers). This is similar in principle to the sort of reasoning that leads airlines to overbook — it is more expensive to fly an empty seat than to bump a passenger. But don’t mistake this model for pure left-brained thinking: issues of cultural resistance to outsider experienced hires and such are smartly treated.
  • Chapter 5 covers demand uncertainty management. This is basically the problem of not knowing clearly how much of what kind of talent you will need, and when. Again Cappelli zooms in on a roughly-right first-step recommendation: portfolio methods. This really only applies to larger, functionally diversified enterprises and works like this: you set in motion a plan to develop a mix of generalist and specialist-functional talent in all your organizations. You then proceed on the assumption that each of your organizations will either undershoot, overshoot or get it right, and be faced with the respective consequences. You hope enough get it right to make the whole portfolio pay off, and also mitigate the situation by moving your generalist talent around. This of course, requires a certain amount of dual career-track management, where centralized corporate control is strong enough to fight silos for influence, and can actually move people around based on corporate rather than local needs, and resist local urges to hide valuable talent.
  • Chapter 6 is all about ROI and is a dense but not very original coverage of everything you can do, ranging from tuition assistance programs to employee-contributions. It has one serious weakness though — while it recognizes that companies that invest in developing an employee may not reap the benefits due to poaching, it does not suggest an actively-managed poaching strategy as a way to offset retention failures. This may have a moral reason behind it — Cappelli notes in the history section that at one time corporations viewed poaching as unethical. This is actually still true in some ways, since rampant poaching hollows out development investment in the economy at large, leading to a tragedy of the commons (Cappelli alludes to this, but we don’t get the full prisoner’s dilemma treatment). But still, big corporations subsidizing the talent development costs of little ones is what creates vibrant entrepreneurial cultures like the Bay Area (recall the famous “Sons of eBay” phenomenon). Whatever the morality, the optimal amount of poaching in the economy is not zero poaching, and ROI models need to factor in an active poaching strategy. At least some zero-sum thinking is necessary. You need to poach passive candidates from other companies at least as much as you are being poached on. Contribute to the stone soup with some development though.
  • Chapter 7 is about internal markets. This is the weakest chapter, especially since it completely ignores the role of technology and the growing phenomenon of semi-permeable boundaries between internal and external talent marketplaces by such game-changers as online marketplaces.

What’s Missing

This is an exceptionally valuable book. It is very dense though — don’t attempt to read it at Barnes and Noble; you’ll need to sit with a copy for at least a couple of weeks worth of evenings and weekends, or maybe 3-4 NY-LA flights, to digest this book. But it is well worth the trouble. It is also not an a la carte dip-in-as-you-need-to reference book. You do need to read it cover-to-cover to get the value.

Still, it stops short of being the bible of modern talent management because it doesn’t cover six things that cannot, in my opinion, be omitted:

  1. Strengths Movement: It mostly ignores/pays lip-service to the “people are unique” aspect of talent management. The enormously disruptive effect the (otherwise very valuable) strengths movement has on goal-oriented resource planning needs a better treatment. The fundamental conflict between letting your reports develop their best strengths and actually finishing projects with high sunk costs, needs to be addressed. Strengths-thinking fundamentally shifts the fungible-unique trade-off curve towards “unique,” and to that extent, paradigms based on “fungible” (such as this one, based on supply chains) are weakened.
  2. Tension with Strategic Pull: Almost all the classic big-picture management literature is based on strategic pull. Possibly the only exception is Good to Great, which proposes the talent-push based Bus theory (get the right people on, wrong people off, and then decide where to go). But this is not a trivial issue. IBM, in navigating its big move into services, laid off 250,000 people and hired 100,000. This is an extraordinarily painful way to manage talent in times of big strategic moves. But perhaps this is an impossible problem, if you think of big strategic moves as The Black Swan events.
  3. Not really on-demand: The book doesn’t deserve its title, but the idea that talent management should be on-demand is on the money. On-demand to me suggests metaphors of cloud computing and utility-based computing. If you’ve ever used Amazon AWS (my team just learned how to use the thing), you know the awesome power of being able to ramp your server fleet up or down very rapidly. That’s true “on-demand.” The labor marketplace is evolving towards that, but the book doesn’t cover this. In particular, the big trend is the building-out of a crowdsourcing infrastructure in the open economy, beyond contractors, consultants and outsourced facilities. I am talking, of course, about online marketplaces like elance, odesk, logoworks, tutorvista, liveops and Amazon mechanical turk, which truly represent an on-demand global workforce. This megatrend needs to be comprehensively treated for a book with “on-demand” in its title — it’s ecosystem management, not just hiring, development and retention. But maybe it is too early. Or maybe Jeff Howe’s upcoming Crowdsourcing will do the trick.
  4. Virtual/Home Office and Free Agent Culture: Another big omission is a treatment of the big movement towards towards sophisticated work-life blending and home/mobile/virtual offices. This clearly has huge implications for the talent management function, and the ideas first developed in Dan Pink’s Free Agent Nation: The Future of Working for Yourself need to be included in any theory of talent management.
  5. The 1:1 Talent Acquisition Market and Social Media: The book’s treatment of online resumes is decidedly dated, and is completely blind to the big development — the replacement of the traditional resume (even the always-on Website or Monster resume) by sites like LinkedIn, where all employees — happy or passively/actively hunting — maintain a dynamic (rather than static) online presence.
  6. Generational Dynamics: The book, while covering every other major thread in history, drops the ball on generational differences in attitudes towards work.

Of course, one cannot cover anything when treating such a complex issue, but still, a single big-picture, citation-dense closing chapter would have made the book much more complete.

Get Ribbonfarm in your inbox

Get new post updates by email

New post updates are sent out once a week

About Venkatesh Rao

Venkat is the founder and editor-in-chief of ribbonfarm. Follow him on Twitter


  1. HBR has a nice little review of an opinion/prediction by Drucker on talent management. Relevant excerpt:

    The growing complexity of the workforce – the need to juggle a wide variety of individuals with diverse preferences and needs, and resultant dizzying array of relationships between businesses and those who perform work – drove Peter to predict that before long, line managers would turn over responsibility for long-term talent management to some type of central staff function. It would simply become too complicated, the arrangements too varied and difficult to track, and the entire process too time consuming for individual line managers to tackle….

    …Meanwhile, traditional line managers would evolve to roles more similar to program managers or film directors – setting direction and running the team of employees who have been assigned to the task or division at that moment in time.

    I don’t agree with this, but it is interesting. I think the systemic response to increasing complexity of arrangements is going to be more technology to manage the workforce, not more centralization. But I do agree that as managerial resources shrink, ‘program’ management type functions will get priority over ‘line.’ The latter will somehow shrink.