Content Copyright © 2006 Bloor. All Rights Reserved.
I have been grappling with real-time or near real-time business
intelligence for a long time and, at last, with SeeWhy, it looks as
though someone has finally started to tackle some of the
fundamentals which will make it possible for true real-time BI to
become pervasive. Real-time for a lot of people can be the
difference between winning and losing. Customers could churn if you
fail to detect a service failure and react, fraudsters can have
taken thousands of pounds and have long gone before a fraud is
detected, clerical errors not detected and corrected whilst the
customer is still present could result in the loss of thousands of
pounds in misplaced processing.
The key to true real-time is to be able to analyse data whilst
it is still in-flight. SPSS, the data mining tool company, tell me
that the huge savings in processing time they are seeing with
in-database mining do not come from the processing in the kernel so
much as from not having to get involved in data transport. This has
been the weakness of all supposed real-time systems up to now; they
are not really real-time at all, they are still waiting for data to
be piped to a database, stored and analysed in situ. It may not be
batch but it is all too often too slow to be real-time.
SeeWhy are able to exploit a number of emergent tends which make
now the correct time to address real-time business intelligence.
Firstly we have the emergence of Service Orientated Architectures.
These appear to be being universally adopted as the response to the
ever-growing complexity of today’s applications. In a
service-orientated architecture, events lead to requests for
services being triggered and in such a model Business Intelligence
is the event seeker that triggers a request for a service.
Then there is the change to the economics of service. With the
ever-increasing pressure on time that consumers suffer they are
increasing volatile and likely to churn. A decade ago people were
far more forgiving of a service failure, now the slightest failure
can result in people looking to leave their bank, insurance
company, phone company or retailer. This volatility and the costs
involved in losing valuable customers change the economics of
building the justification for a real-time solution. There is now a
pressing business imperative; you really do need to detect and
react to all significant failures impacting your valuable
customers. Business is now working increasingly in real-time in
order to satisfy volatile customers; accordingly we have to monitor
what occurs in real-time if we are to be able to maintain the
appropriate service level.
SeeWhy automatically looks to build rules without the need for
intervention. This enables it to build rules at a very fine grain
of abstraction. Traditional systems were too reliant people to
build rules. Domain experts used in this way were so unproductive
as to limit the extent to which those rules were applicable to a
small minority of situations. And as the technology was also slow,
rules tended to only be applied to samples rather than the whole
population. Now with SeeWhy it is becoming possible to apply very
fine grained rules, identified automatically, to a sequential scan
of the whole population. This opens up many new application areas
and is a very exciting addition to the armoury of corporate
productivity and profitability. I am very excited by the
possibilities that SeeWhy presents and will be monitoring its
progress with interest.