A few years ago, everyone was talking aboutthe real-time enterpriseand how it
was destined to be the future. The idea was that companies would be
instantaneously aware of any important business events taking place
so that they could respond appropriately and
immediately.
But that has not happened because of both the cost and
complexity of such a proposition. As a result, the hype has moved
on to
real-time data analytics for certain sets of data rather than
for all of it.
This area is, in turn, facing its own problems. Although the
concept of being able to make snap business decisions based on
up-to-the minute information may resonate with a lot of people, too
many are confused as to what it means and how they can make it a
reality.
What real-time means
And one of the issues here is simply that of terminology. Gareth
Herschel, a research director at Gartner, says, "For most
organisations, data is not truly real-time. The term is a bit
misleading and it is a bit of a false goal for organisations to
work to.
"The question is, is there a real benefit in being able to
analyse data minute-by-minute or is day-by-day or week-by-week
enough? And the answer really depends on how fast their market
moves."
Herschel says the concept of "right time and right latency" is
more apt. "The more real time data is, the better, but the closer
and closer it gets to that, the more systems cost to put in place.
So there has to be a clear business benefit for investment," he
says.
Steve Shepherd, solutions sales manager at
IT services provider Logicalis, says that organisations are
unlikely to get much change out of £500,000 for a real-time data
analysis system once software licences and services are taken into
account.
"These things are not a small investment. If you just put a
glass sheet over your systems and introduce a notification tool to
say that 'based on these conditions, I will send you a report', it
will not cost too much. But if you are trying to profile a customer
using data from different systems to do more in-depth analysis, it
becomes much more involved and, therefore, much more expensive," he
says.
Real-time reporting systems
And this highlights another problem - of what exactly real-time
data analytics means. Herschel says there are three key categories
of system, all of which are at varying stages of maturity. The
first area comprises reporting, but this is rarely undertaken in
true real-time.
"Most enterprises are trying to lower the latency of how
frequently they update and collect data from across the
organisation. This tends to be a knock-on effect from investing in
a data warehouse and so is quite a universal trend," he says.
Business activity monitoring
The second field is
business activity monitoring (BAM) and complex event processing
(CEP), with the former being a sub-set of the latter.
With BAM, the idea is that organisations receive alerts when a
specific but unusual event takes place so they are in a position to
respond quickly. An example might be a package being delayed at
customs or a customer cheque bouncing.
CEP, however, is about companies being able to respond swiftly
to multiple anomalous events. So, for example, if a customer did
not receive their salary one month but deposited a large one-off
payment the next, their bank might deduce that they had lost their
job and just received a severance payment. This means that,
although it might not be a good time to try to sell them a
mortgage, it might be appropriate to offer them a loan.
"Every company could probably benefit from this in specific
areas of their business because it is about notifying people of
problems before they become significant so that action can be
taken. But, although some enterprises have been doing it for a
while, it is relatively early days," Herschel says.
Real-time data mining
The third category is real-time data mining. The goal here is to
provide guidance to call centre agents, for example, about what
action to take at a given point in the sales or service
process.
So, if a customer phones to pay a bill the system might
recommend that the call centre agent promote a specific product or
even nothing at all based on an up-to-the-minute profile.
"This is one of the hottest areas of customer relationship
management (CRM) and is an area that call centres in sectors such
as telecoms, financial services, retail, and travel and leisure are
looking at.
"It has been around for about 10 years as a concept, but few
companies are using it, because it requires a change of mindset on
the part of the organisation," says Herschel.
A key challenge here is that the performance of most call
centres is measured on how quickly they answer customer calls.
Cross-selling and up-selling slows this process down.
Therefore, organisations often find they need to hire additional
staff, which will need to be trained in sales. Existing personnel
may also become unhappy with their new role, which leads to
attrition.
"Real time analytics projects are very similar to CRM projects
in that most are not technical failures, but organisational ones.
There is no point having information at your fingertips if it takes
a week to react to it. So it is all about the decision that you are
trying to make," Herschel says.
Case study: Zavvi Entertainment
Zavvi Entertainment Group, which bought out Virgin Retail, went
live with a near real-time system about 14 months ago, with the aim
of enabling stores to provide customers with up-to-date information
about stock availability.
Another key goal was to enable central buyers to view, at
15-minute intervals, how well or badly individual products were
selling in order to improve purchasing decisions.
For Tony Johnson, IT director at the Zavvi Entertainment Group,
faster access to relevant information has a definite business
benefit.
"Getting early visibility of day one sales and being able to
respond does give us a distinct commercial advantage, especially
with new [computer] games, which are always released on a Friday
and quite often have a constrained supply," says Johnson.
Previously, because Zavvi used overnight batch processing to
update information, the first time company analysts had a chance to
look at the day one sales data was on Monday morning.
"But now we can look at reports at any point on Friday so that
we can make decisions rapidly. This gives our central buyers more
chance of procuring popular products, which has a direct impact on
revenues," Johnson says.
Johnson says that Zavvi decided to update its data at 15-minute
intervals rather than on a real-time basis for pragmatic
reasons.
"Initially, we were not 100% sure that the system would cope
with more regular updates, because you have to take account of the
overhead on store systems, the data warehouse, the reporting and
analysis tools and on pulling data over the network. We have since
found that we could do it significantly more frequently, but we
have also found that we do not need to," he says.
An issues to bear in mind here is that if information is updated
too regularly staff can "spend their lives in front of the screen
pressing the refresh button", particularly at first because of the
novelty factor.
But Johnson also says, "It is absolutely fundamental to be clear
about why you want to introduce real-time data analysis and the use
that the data is going to be put to. It is great for people to look
at it and say we have done this volume of sales at this time, but
if they are not acting on that data, it could end up just being a
distraction."
He also says that it is sensible to start small with one or two
clear goals in order to prove that the system works, while at the
same time building on what is in place "to exploit existing
investments".
Zavvi's had already implemented a data warehouse along with
MicroStrategy's online analytical processing tool, but subsequently
deployed Microsoft's BizTalk middleware to act as the glue between
key systems - and to provide an interface to suppliers' systems in
order to improve data exchange.
"BizTalk pulls data from the store point-of-sale systems back to
the central data warehouse and a heavy amount of work went into
developing the interfaces here. Significant work was also needed to
build the MicroStrategy reports to deliver information to the
buyers' desktops and to push that data into our merchandising
system," says Johnson.
Nonetheless, Zavvi managed to implement its new system in six
months.
Case study: Unilever
The same was true of Unilever, which introduced a near real-time
data reporting system from IRI Research at one of its Southern
European subsidiaries this year. It also has plans to roll the
offering out across several others, including the UK, in the year
ahead.
The aim, says Peter Kollecker, customer development process
manager at the company's development process office Europe, is to
analyse sales data from each of its retailers three or four times a
day "to see where things are at with on-shelf availability, how new
products are doing in the first four weeks of introduction and
whether products are selling fast, slow or not at all following
promotions".
The underlying objective is to enable Unilever to react more
quickly to market conditions and to be more proactive in its
dealings with retailers in order to meet their stock requirements
more effectively.
But a key issue here is one of building up trust between the two
parties. "There has to be trust there because you want to expose
data to each other so communication is very important.
"We have been very open about what we want to report on and the
outcomes we want, but it is also important to explain what benefits
there will be to both sides. It has to be a win-win situation,"
says Kollecker."