In April 1860 a company began carrying mail from St. Joseph, Missouri, to Sacramento, California, by relay: a rider covered roughly ten miles at a gallop, swapped horses at one of nearly two hundred stations, and after several stages handed the mail to the next rider. A letter crossed the continent in about ten days. When the transcontinental telegraph was completed in October 1861, the Pony Express, after eighteen months of operation, closed within days. The job it did, moving information across a continent, is done today more than ever, but the solution was obsolete the week a better one appeared. The horses, the stations, the riders and the brand were worth nothing against the telegraph, which needed none of them.
A flywheel keeps a machine running when its power is interrupted, and the flywheels of chapter 5 do the same for a firm. From roughly 2009 to 2015 Volkswagen sold diesel cars whose software switched on the emissions controls only during a test. Executives knew. The company paid tens of billions of dollars in fines and settlements, and its market value fell by a third in a week. Today it is still among the two or three largest car makers in the world. Incumbents with spun-up flywheels are nearly impossible to kill. What, short of a telegraph, does kill them, or, more often, kills the advantage and leaves the firm alive?
Imitation, and the Friction That Slows It
Most advantage ends the ordinary way: a rival copies it. In the 1989 Tour de France, Greg LeMond's aerodynamic helmet and clip-on bars were alpha for one afternoon; by 1995 they were standard equipment. The mobile ordering app of the Chinese coffee chain Luckin Coffee was alpha for a few years. Patents run out, trade secrets leave with the engineers who know them, and the proprietary process becomes the subject of a conference talk. An alpha asset lives only as long as it stays hard to acquire.
What sets the length of that life? Friction, of which there is more than economists' models assume and less than incumbents hope. Customers switch slowly. Electronic toll collection was demonstrably better than a coin basket for a decade before most drivers had a transponder, and LED lighting was cheaper over its life than the incandescent bulb for years before it was in most sockets. Everett Rogers spent a career cataloguing why good ideas diffuse slowly, from the British navy's two-century delay in adopting citrus against scurvy onward; his answers in Diffusion of Innovations (relative advantage, compatibility, complexity, trialability, observability) amount to a list of frictions. Know-how moves slowly even between units of the same firm, as Gabriel Szulanski found (chapter 5), and more slowly between rivals. So an advantage with a theoretical life of a season often has a practical life of years, and a firm that understands its own frictions can estimate how long it has.
Friction is also what a language model reduces. A product that took a year to reverse-engineer can be approximated in a week. A process that lived in the heads of a few engineers can be described, by them, to a tool that documents it. Advantages that rest on a better solution now have shorter lives than they used to, and the assets that never depended on friction for protection (relationships, installed bases, brands) gain in relative value.
Two Conditions for Disruption
Some advantages do not erode; they end outright, either because the solution to the job changes, as it did with the telegraph, or because the job itself changes, as it did with the decline of church bells. Executives fear the first kind most, so let's start there.
A new solution technology destroys an incumbent only when two conditions both hold. First, the new solution must actually prevail: a substantial fraction of the market must come to prefer it. Second, the incumbent must be unable to adopt it profitably, perhaps because the value is redistributed to the consumer and nothing is left to capture, or because someone else controls the new solution, or, most often, because the incumbent's alpha assets are irrelevant to it. Consider Kodak. Digital imaging prevailed, and nothing Kodak owned (its chemistry, its film plants, its relationships with photo finishers) had any use in it. The job survived; the firm's assets did not. The entire value of consumer photography did not move to a successor firm but to the consumer, who now takes pictures for nothing.
Most threats fail one condition or the other. Although electric cars have taken share for fifteen years, the incumbents that make combustion engines also make electric cars, since the assets that matter in the sustaining phase of the automobile business (brands, dealer networks, scale, manufacturing capability) are mostly relevant to the new solution. Electric cars fail the second condition; e-books fail the first. They exist and are convenient, and their sales plateaued at a fraction of the market because a substantial fraction of book buyers prefer paper.
The threat that passes both conditions is rare. I once tallied, informally and from a bankruptcy database, the public companies that failed in the decade from 2005 to 2015 because a new technology made their solution obsolete. I found about sixteen, out of several hundred bankruptcies among several thousand firms: Blockbuster, Borders, Kodak, Circuit City, a handful of newspaper and printing companies, Nortel. That is a base rate for bankruptcy from technological disruption on the order of a few hundredths of a percent per firm per year.
That base rate comes with two cautions. First, bankruptcy is a narrow measure, since margins, share and independence usually go before the firm does, and a company can survive while its advantage disappears entirely, as the newspaper companies that did not go bankrupt have. Second, a tally of failures is not a theory. Clayton Christensen's account, in The Innovator's Dilemma, is that a cheaper, worse product improves until it displaces the incumbent from below. In a 2015 article, Andrew King and Baljir Baatartogtokh went through seventy-seven cases that the disruption literature uses and found that most did not fit the theory's own conditions. Disruption by technology is real, rare, and poorly predicted by any theory I know, including my own two conditions.
What the Incumbent Can Do
Executives are told that only the paranoid survive and that a firm must disrupt itself before someone else does. I disagree. For most incumbents, most of the time, the right response to a technological threat is to track it, deliberately do nothing about it, and keep spinning the flywheels every day. Most threats fail one of the two conditions, and chasing each one costs more than the rare one that materializes would have.
When a threat does look real, I apply the two tests of chapter 2 by asking whether our alpha assets will remain alpha against the new solution, and if not, whether we can acquire the newly relevant ones by building or buying. A third question is easier to forget: for what other jobs might our assets, no longer alpha for this one, still be alpha?
The menu of responses is short. A firm can commit fully to the new solution, which works when the solution does prevail and the firm has the capabilities to win in it; Bill Gates's 1995 memo turning Microsoft toward the internet is the example everyone cites. It can hedge with a project or an autonomous unit, as Netflix did with streaming while the DVD business still paid the bills, although the legacy business usually starves the hedge that threatens it. It can fight, by acquiring and killing, by suing, by lobbying. (Paris taxi drivers fought ride-hailing for years and bought themselves years.) It can concede and manage the old business for cash. That is the right call more often than managers admit, because the new solution often does not dominate, or takes decades, or leaves a profitable remainder. Or it can push its existing assets into adjacent jobs where they remain alpha, the subject of chapter 6 and the response most often neglected in favor of the more dramatic ones.
The test I use is the one a venture investor would apply: knowing the incumbent's assets, would you fund its plan to launch the rival solution? In 2000 the book retailer Borders could not have raised money to beat Amazon at internet retail from any investor who looked at its assets, and it should not have tried. Its shareholders could buy Amazon themselves. Management's job was to maximize the cash flow from the assets the firm had. That pointed toward the college bookstore business its rival Barnes and Noble eventually built, not toward the fight with Amazon that Borders lost.
In November 2021 I put a question to executive audiences. Toyota and Tesla had revenues and market values that invited comparison: Toyota's market value was about ten times its earnings, Tesla's about seven hundred times, on revenue a fifth of Toyota's. How would you bet? The question was a test of the two conditions. Electric drive was clearly going to take a large share. Toyota's alpha assets (its brand, its dealer network, its manufacturing capability and its scale) were mostly relevant to electric cars, while Tesla's were a head start in software and batteries and a brand built on being first. For Tesla's valuation to be right, Toyota's assets had to stop being alpha against the new solution; for Toyota's to be right, they had to remain alpha.
Anyone can look up the two companies today, and the figures will have moved again since I asked. A firm whose assets remain alpha against a new solution is not disrupted by it, however loud the solution is. A firm whose assets are irrelevant to the new solution is disrupted by it, however large the firm. Which of the two Toyota turned out to be is a fact about its assets, knowable in outline in 2021 from the worksheet in the appendix, without a forecast of either company's share price.
When Your Own Advantage Stops
The same two conditions apply to a person, where they are easier to see because the firm around them supplies the complements.
Start with a skill at executing a particular kind of work, which lasts as long as friction keeps others from acquiring it. Friction has fallen. The skills most exposed are the ones whose output a model can verify and therefore produce: a clean translation, a working function, a standard analysis. The least exposed are the skills around the execution: knowing which problem to solve, judging whether the solution is right, getting it adopted. If your asset was the execution, its life is getting shorter; if it was the judgment around the execution, its life is getting longer, since the execution you judge is cheaper than it was.
Assets that come with a position stop differently: they belong to a setting and end when the setting does. The analyst's rank was the bank's platform; the professor's inbound mail was the institution's name; the salesperson's number was the territory. Chapter 4 asked whether an advantage belongs to the person or to the position, which is this chapter's question asked in advance. If you know which of your assets are yours and which are the setting's, you can see the end of a position coming and build, before it comes, the assets that travel. If you don't, you learn the difference on the day you leave.
Accumulated advantage, for a person as for a firm, also makes changing direction costly. Ten years invested in an asset that is alpha in one setting are ten years of real reasons to stay, even when the setting is dying. The deliberate builder of chapter 7 wrote down, before starting, what would make them stop. The holder of an advantage should do the same: name, while the asset is still alpha, the conditions under which it will not be, and watch for them.
Notes
Everett Rogers spent a career cataloguing why good ideas diffuse slowly Everett M. Rogers, Diffusion of Innovations, 5th ed. (New York: Free Press, 2003). The scurvy case (James Lind's 1747 trial; the British navy's adoption of citrus in 1795) opens the book's discussion of the slow diffusion of a demonstrably effective innovation, and the five attributes of innovations that affect their rate of adoption are developed in chapter 6.
Andrew King and Baljir Baatartogtokh went through seventy-seven cases Andrew A. King and Baljir Baatartogtokh, "How Useful Is the Theory of Disruptive Innovation?" MIT Sloan Management Review 57, no. 1 (Fall 2015): 77–90.
Clayton Christensen's account Clayton M. Christensen, The Innovator's Dilemma: When New Technologies Cause Great Firms to Fail (Boston: Harvard Business School Press, 1997).
Bill Gates's 1995 memo turning Microsoft toward the internet Bill Gates, "The Internet Tidal Wave," internal Microsoft memorandum, May 26, 1995, made public in the United States v. Microsoft proceedings and reproduced at Letters of Note.