Page 292 - Deep Learning

Error Correction in Context 275

This view gains support from naturalistic corpora of collective errors.
Safety scientists have gathered such corpora for several types of collectives,
including businesses, clinics and hospitals, military organizations and socio-
technical systems, especially transportation systems (e.g., airlines). The errors
described in such corpora exhibit a bewildering variety. However, the variabil-
ity resides to a large extent in the specifics of the particular events. Narratives
of the burning of the Hindenburg airship, the collapse of the satellite phone
industry and the military losses during Operation Market Garden read like
horror stories by very different authors. However, the scripts, as opposed to
the events themselves, are not entirely different: In most stories the relevant
collective moved in spite of the presence of counterindicators that should have
constrained the move. If we abstract from the domain specifics, there is con-
siderable similarity among an airplane crew that takes off in bad weather with-
out de-icing the wings, a corporate management team that becomes gripped
by deal fever and completes a financially disastrous merger without adequate
review of the financial benefits and a combat force that initiates an attack in the
face of intelligence that implies that the battle cannot be won.
For example, consider the corpus of business errors described by Paul B.
Carroll and Chunka Mui in their 2008 book about the Billion Dollar Lessons
that one can draw from “the most inexcusable business failures of the last 25
years.” They analyzed hundreds of cases in which the execution of a com-
20
mon business strategy caused a major failure because the strategy was not the
right one under the circumstances. They classified their cases into seven types.
For brevity, I collapse their classification into four types of strategic business
errors. The first is inappropriate extension of an existing business. The move
in this case is to expand the business, but the chosen expansion is inappropri-
ate in some way and the result undermines rather than strengthens the busi-
ness. According to Carroll and Mui, this can happen by overestimation of the
benefits of a merger with another company; by overlooking the downsides of
rollups and consolidations, that is, attempts to buy up other businesses; and
through adjacencies, moves in which a management team expands into an
adjacent market for which their business is ill suited or that does not provide
any advantage. The lesson taught by these cases is obviously that an expan-
sion of any sort, once thought of and proposed, should be studied carefully
for pitfalls and counterindications before the organization charges ahead. The
conditions under which different types of expansions are likely to work need
to be specified.
The second and third strategy errors are each others’ mirror images: To
stay the course even though technical inventions are changing the relevant