Features
Feature
From Insights to Actions
Decision management helps unlock the potential
of predictive analytics in operations.
by James Taylor
Someone once said, “He who
knows first wins.” The implication
is clear—collecting and analyzing
data faster than your competitors will
enable success. Yet gaining insight before
the competition helps only if you can also
act on that insight.
A company that sees a new consumer
trend first but takes too long to respond
could fall behind. After all, customers only see
how you act, not what you know. Knowing,
then, is not enough. You must act on this
knowledge—you must make decisions.
Micro-Decisions Stack Up
Organizations make myriad decisions. A few
are strategic, focusing on where and how to
operate, or tactical, regarding campaigns and
projects. But the vast majority of decisions
are operational ones required by a specific
transaction, such as determining:
- Whether a claim is fraudulent
- What offer to extend to a customer to
profitably retain him or her
- Which order to delay because parts are
in short supply
Predictive analytics turns uncertainty about
the future into a usable probability.
While taking timely action always matters,
it is particularly important in these
operational decisions. The time available
to react is typically shorter, and responses
can be required by automated systems like
ATMs or websites. Although individual
choices have limited impact, their cumulative
effect is massive. For instance, they
represent the front line in managing risk.
After all, risk is often acquired one bad loan,
or one fraudulent claim, at a time.
Such “micro-decisions” relate to a single
customer, order or claim. Focusing on them
allows each individual or transaction to be
handled uniquely while also allowing for
automation. The key is the ability to apply
analytics—especially predictive analytics—
to micro-decisions. If models can be built
to determine how likely it is that a claim
is fraudulent or that a specific offer will
retain a customer, then even an automated
decision can deliver a personalized result.
But to do this, advanced analytics must be
embedded in operational systems.
Challenges Ahead
When it comes to applying advanced, predictive
analytics to operational micro-decisions,
three main issues present themselves:
- Linking analytic insight to action
- Bringing different organizational
perspectives together
- Focusing analytic efforts on a more
“industrial” approach
Analytic models deliver insights from data
that you would not otherwise have noticed.
Analytics simplifies data to amplify its meaning, turning large volumes of it into a
small amount of valuable insight. Predictive
analytics does more, turning widespread
uncertainty about the future into a usable
probability. You don’t know which claims
are fraudulent, but you can use predictive
analytics to estimate how likely a specific
claim will be false.
It’s easy to forget that only actionable insight
is valuable. Sadly, a majority of analytic models
never make it into production—resulting in
insight without action. Analytic models must be
combined with the rules that determine what to
do next. This is a challenge particularly when it
comes to operational decisions.
Additionally, the different groups and
roles involved pose another persistent
problem. Often referred to as a three-legged
stool, operational analytics requires the
business, IT and analytic teams to cooperate
to net positive results. Only when all three
work together will the stool stand, after all:
- The business understands the decision
being made and its process context.
- IT knows the systems that support the
operational environment.
- Analysts leverage the data to derive
the insight.
Finally, operational analytics
frequently challenges the tenet that
analyst teams should decide which tools
to use and how to use them. While this
flexibility is important in supporting
analytic model development, it can
result in models that are hard to implement
in operations. A more industrial
mind-set is required, one in which the
only good analytic result is a business
result, and the only good model is one
that changes how the business works.
Decision Management
Boosts Value
To successfully apply analytics in operations,
a new approach is needed, one that
builds on agile IT methods, business process
management and analytic methodologies.
Known as decision management,
this approach has three phases:
- Decision discovery involves an ongoing
effort to find, document and understand
the choices that matter to your
business. As processes are analyzed,
the decisions that underpin them are
identified and documented. Only with
an explicit understanding of these
decisions can sensible choices be made
about the use of analytic technologies,
business intelligence (BI), business
rules, etc. And only with an understanding
of how these decisions relate to your
objectives and critical measures can
you differentiate between good and bad
ones, or establish changes to move your
metrics in the right direction.
- Decision services focus on delivering
coherent IT components that make
decisions. Built using business rules
and predictive analytic models, they
link the business and analytic environments
with the service-oriented IT
ecosystem that supports operations.
It’s critical to use the right technologies
to build these services and to
ensure that decisions are not hidden
in systems or processes.
- Decision analysis entails the ongoing
monitoring and improvement of
decision making. No choice is made
the same way forever. What makes
a decision good or bad shifts as
competitors, markets and customers
evolve. Only by collecting and
analyzing data about the effectiveness
of your decision making can you
keep improving.
Limitless Potential
Predictive analytics applied to operational
decision making is the next wave
of business innovation—the next major
source of competitive advantage. The
most successful practitioners are using
decision management.
The potential is limitless for
embedding analytics in operations
and improving the quality
of countless micro-decisions
that involve customers and their
interactions with your company.
Making it work requires an explicit
focus on choice and actions and on
bringing together IT, business and
analytics teams.
James Taylor is CEO of Decision
Management Solutions.