As a decision support tool, analytics should clarify the right, best, or optimal actions to take in various circumstances
Analytics is an overarching concept: In the lexicon of
normative adjectives applied to the noun "analytics", prescriptive"
is a bit closer to the imperative end of the normative spectrum in terms of
meaning. Does that mean prescriptive analytics is akin to actionable?
BI/Analytics
Evolutionary Journey: In last thirty years, BI/Analytics is well on its evolutionary journey: Descriptive analytics - vast majority of traditional
BI - provides insight into what has happened. Predictive analytics helps model
and forecast what might happen in the future. Prescriptive analytics goes
beyond Predictive analytics and seeks to determine the optimal solution among
various choices, given the known constraints and parameters. The
descriptive/prescriptive/predictive categories are another attempt to try and
differentiate advanced/strong analytics from general BI - and that is certainly
worthwhile.
Prescriptive is about
finding the optimal solutions under a set of conditions or varies circumstances. Prescriptive
analytics go far beyond descriptive analytics (what has happened) and
predictive analytics (what might happen) to what action you should take based
on all the available facts past, present and future. Think about it as finding
a route on a GPS device with
traffic information. The system knows where you are and where you want to go.
It knows the likelihood you will run into traffic and how long it will delay
you. And then depending on what your preference is (most use of highways,
shortest time, and least use of highways) it recommends a route. That is
prescriptive...finding the optimal answer under a set of conditions.
Prescriptive
analytics is akin to "actionable". ALL
analytics should enable action: Without the capability for action, it's a
worthless activity. In that sense, all analytics should aim to be prescriptive.
Further, predictive analytics needs to convey some sort of qualitative value in
conjunction with the results. There remains a fine line, then, between
'predictive' & 'prescriptive' analytics. Select the optimal qualitative
value from predictive analytics, and you have the 'prescriptive' equivalent.
As a decision support
tool, analytics should clarify the right, best, or optimal actions to take in
various circumstances, as borne out by objective descriptions about what
has happened in the past and predictive models about what is likely under
various future scenarios. In terms of guidance, you can distinguish between
"prescriptive" (here's what you should do) vs. "proscriptive"
or "interdictive" (here's what you should avoid doing).
 |
| Add caption |
Prescriptive
Analytics may include performance management/score-carding: Some say
prescriptive analytics is a re-branding of operations research, in terms of
focusing on constraint-based optimization. But prescriptive analytics has also
been stretched to include performance management, scorecarding, and other
guided analytics in the BI context. In the final phase of analytics, the
predictive models are applied and business decisions are made. Then company
performance is tracked so that actual outcomes can be compared to predicted
outcomes to improve the predictive quality of the model.
Bottom line: All
analytics should enable action. Whether you need descriptive, predictive or
prescriptive analytics is a question of picking the right tool for the job well
done.
0 comments:
Post a Comment