Page 235 - Lean six sigma demystified

Chapter 6 Tr an S a C T iona L Six Sigm a 213

3. Expect each new application release to be error-prone. Use systematic
problem solving to identify and remedy all of the requirements, design,
and coding errors. Resolve problems at their source, not necessarily where
they show up.
4. As your new system evolves, simplify and streamline the software to prevent
the creeping complexity that will render it inflexible and unchangeable.
Having worked with software for over 30 years, I’ve noticed some patterns
of behavior that in hindsight seem obvious, but, in foresight, are largely
ignored.
First, most new application systems arrive at around 2.5 sigma—over a 15%
error rate. This is not because the IT department did a crappy job of testing, but
because it’s almost impossible to specify every condition that you’ll encounter
when developing a new application for a large company. After their cataclysmic
maiden voyage, these systems achieve equilibrium around three sigma—3% to
6% error—while still encountering the enormous costs of human error correc-
tion on the remaining fallout (autonomation).
Second, all application systems have some method for detecting errors—
input that doesn’t match expected parameters—and someplace to store these
errors until they can be examined and resolved by a living, breathing person.
When you’re doing 10,000 transactions a day like most large companies, 15%
errors translates into 1,500 errors a day to be corrected. Most company CIOs
expect their shiny new systems to be infallible, so this error rate comes as a
shock. Squads of error correctors are rounded up to fix the growing backlog of
errors that are delaying order fulfillment, billing, and payment. Customer ser-
vice call centers are pushed to their limits by customers trying to find out what
happened to their orders, bills, and payments. After much blood letting, the

error rate falls to 3% to 6%, which seems tolerable compared to the previous
level. Most large companies, whether they admit it or not, have staffs of 50 to
60 people fixing these ongoing errors created every day on each of their key
information systems. Because this error correction is done by people who are
inadequately trained and powered by the same technology that created the
original errors, as much as 15% of the errors are corrected incorrectly and have
to be fixed again, and again, and again.
Information systems usually involve ordering, production, delivery, billing,
and collection systems. Like salmon in a stream, most companies try to swim
up river from the polluted end of the process rather than correcting the prob-
lem at its source. Start with the ordering system and downstream improve-
ments will be substantial. Then move downstream, system by system, eliminating