One interesting story emerging from the 2013 holiday
shopping season is that many of the big data analytical tools employed by
retailers failed to predict consumer behavior. A notable item is that Amazon had over a million customers sign up for its Prime expedited shipping service in the third week of December, causing the company to throttle back its sign-up acceptance.
Additionally, everything from weather patterns to
last-minute competitor offers threw a monkey wrench into analytical models,
confounding even the most data-savvy retailers. Even the lowly consumer has
learned to game the big data system to some extent, comparing pricing and
exploiting price match policies at retailers to get the best deals. In some
cases, consumers even skipped pre-Christmas offers that didn’t offer compelling
price cuts and saved their buying for post-holiday sales.
Big data isn’t
The power of big data is the massive amount of data points
and the ability to analyze structured and unstructured data, but unfortunately
the technology is not magical and, like any analytical tool, provides answers
based on the past but cannot predict the future. Just as forecasting the
weather or predicting stock market moves will always be an inexact science, so
too will the results of even the best big data tools be a hypothesis rather
than gospel truth.
At its very best, any analytical tool is the proverbial
canary in the coal mine. The singing stops when the danger is already real,
with the hope that you have enough time to react. The canary can’t preemptively
predict the future — it merely warns of an active gas leak so you can seek
shelter. Despite occasionally overzealous vendor claims, big data can identify
historical trends and alert you to a change that’s already occurred, providing
valuable reaction time.
Leveraging the canary
Once your organization understands that big data is not a
magical crystal ball, you can begin to use it appropriately. Big data was never
meant to be a Wizard of Oz-like man behind a curtain, who would opine about the
future and issue dictates; rather, it was meant to be a tool to test hypotheses
developed by intelligent humans.
In the case of a retailer facing an uncertain holiday
season, an obvious concern would be shoppers waiting until the end of the
season. Various hypotheses could be developed to gauge this, ranging from
traffic counts in various stores, to browsing digital offers but not pulling
the trigger. More sophisticated analysis might factor in environmental data
like the weather, or issuance of price matches or some other metric. With deep
big data capabilities, this retailer might see a high incidence of price
matches early on the morning on Black Friday and issue a predetermined flash sale
that evening to steer consumers to a higher-margin product. The hypotheses were
in place, generated by intelligent humans, and big data provided an alert that
led to quick action.
Big data fire drills
In the hypothetical scenario above, rather than waiting for
the big data Wizard of Oz to dictate the next tactic, the retailer had a
variety of contingency plans and let the data guide the next action. The faster
reaction time engendered by big data provided the advantage, not some property
of big data itself.
Key to this reaction time is having various hypothetical
scenarios as well as appropriate responses determined and ready in advance. Too
many organizations wait for the analysis before determining the next action.
Like a coal miner staring at a dead canary and wondering where he should go, or
if the elevator is waiting at the bottom of the mineshaft, this wastes valuable
time. You need a variety of plans to cope with the potential outcomes of your
analysis, rather than a vague idea in someone’s head, and these should be
tested and ready to execute. Just as the miner practices his escape plan, you
should rehearse responses to critical business changes. While lives may not be
at stake, competitive advantage and the associated revenues surely are.
Does your big data analytics team usually have hypothetical scenarios as well as appropriate responses determined and ready in advance? Let us know in the discussion.