A Monthly Article from our Speakers
Current Article of the month
The Software-Defined Enterprise: Microservices, Modern Architecture and Business Agility
by Frank Greco
Integrating Business Intelligence into the Enterprise (Part II)
BI Web Services
The fourth approach to integrating BI into the enterprise is similar to the third except in this case there is loose coupling of BI tools and analytic applications with operational systems whereby integration can be achieved via the introduction of BI web services. Figure 3 shows the concept of web services. Web services are the latest craze in the industry. The key advantage they offer is the ability of any application to simply dynamically bind to a web service ‘on the fly’ at run time. There is no need to build any ‘tight’ interface between enterprise applications and BI tools or analytic applications. In fact the enterprise operational application doesn’t need to know how to interoperate with the BI web service. Here analytical enterprise java beans, an OLAP product, a data mining product or an analytic application can be published as one or more web services using the emerging industry standard Web Service Description Language (WSDL). Web services are published in a UDDI directory that is similar to an electronic ‘Yellow pages’. UDDI directories can exist inside the enterprise on an intranet, or out side the enterprise on the public Internet. To make use of any web service (OLAP web service, data mining web service, printing web service, security web service etc), operational applications or even database triggers and stored procedures simply issue an XML query to the UDDI directory to search it to find the web service needed. Sufficient information exists in the web service description to tell the operational enterprise application how to automatically and dynamically connect to that BI web service at run time. This connection occurs by the operational application sending an XML message over the web to the BI web service using a protocol called Simple Object Access Protocol (SOAP). In this architecture there is no need to tightly integrate the applications as in the third option discussed above. This strategy is purely based on loose coupling and whilst very new, is already supported by some BI vendors e.g. Microsoft has introduced it in its Analysis Services OLAP technology as part of its .net (pronounced “dot net”) architecture. Hence, operational applications can request reports, OLAP cubes, and mining on the fly at run time. This can happen within the enterprise and between enterprises. Therefore suppliers for example could dynamically link to a retailer’s BI web service and request intelligence on demand so that the supplier can more accurately optimise their manufacturing and distribution to meet that demand. In addition, a front office operational application such as Siebel CRM could call a BI web service to request customer intelligence about a customer that a customer service representative is dealing with right there and then in the call centre. BI web services can be invoked on an event driven basis. An example would be a simple operational database update that fires a database trigger that in turn dynamically links to a BI web service to get up-to-date customer intelligence.
Intelligence can of course be passed back to the requesting application in XML form. In this case it could include XML cubes so that intelligence can be absorbed into operational applications easily. Alternatively a user could then be presented with an XML cube on their device.
The Portal Option
The recent surge in portal implementation presents another option of integrating BI and operational applications into the same secure, personalised web based user interface without the need for tight integration of operational and analytic applications. Corporate and e-business portal products allow users to see both related operational and analytic application content within the same user interface. Some portal products go further than that e.g. SAP Portals Top Tier with its very powerful ‘drag and relate’ concept shown in Figure 4. Here for example, a user could drag and drop a customer number onto an analytic application to cause the intelligence about that customer to appear. Hence the portal changes the business process to allow the user to leverage intelligence as and when needed.
The portal option is very powerful in that it removes the need for the user to have to learn and struggle with multiple user interfaces. Here the operational application and the BI system are integrated into the portal common look and feel. The user no longer ‘has to know’ which system to use to get the information they want. Even though multiple systems are being used, it is presented to the user like it is all one system. In addition, the portal allows users to organise content and relate BI to other types of content so that the user can easily find what they need. Many portals also handle access from multiple devices (e.g. mobile devices) and automatically adapt content to fit the device being used. Hence users can access BI on the road. Some BI tool vendors have their own BI portals e.g. Brio Portal, Business Objects InfoView, Cognos UpFront etc. However these portals do not provide access to operational applications. What most companies are doing is integrating BI portals into portal platform such as SAP Portals, IBM WebSphere Portal Server, Plumtree, Epicentric Portal Server, and Microsoft Sharepoint. To accommodate this, many BI tools vendors have now announced pre-built XML interfaces to these portal platforms. These interfaces go by various names depending on the portal platform e.g.SAP Portals iViews, Plumtree Gadget Web Services (not to be confused with web services as discussed earlier), IBM WebSphere Portal Server portlets, Microsoft Web Parts etc.
Real Time Analytics
In this final approach all the steps in the extended analytic process are provided by technology that runs automatically. Front office, back office, and e-business applications place a real-time call to a decision engine requesting a decision or recommendation. The decision engine then invokes an analytic engine (typically a BI tool) on demand to analyse the data in the BI system to the produce intelligence needed. The intelligence is then passed back to the decision engine (typically in XML format) that then uses business models or rules to make a real-time recommendation or take a decision. This is shown in Figure 5. Example recommendations might be product recommendations passed back to a customer service representative or a prospect on a web site. It may be used to personalise a users web experience etc. The key point here of course is that there are no people involved in a real time analytics process. It is event driven, real time automatic analysis, automatic interpretation of BI and automatic action/recommendation all in the midst of real time business operations. Note also that the data being analysed can be a day old, a week old or near real time. To get near real time data to analyse, interpret and act on, requires ETL tools to integrate with enterprise application integration (EAI) technology to get changes as they happen in business operations. Alternatively EAI XML messaging technology could simply pass the data directly into the warehouse.
There are a number of technology options for decision engine technology e.g.
- Deployed mining models such as SAS C*Store and J*Store, or IBM DB2 for scoring. These models are produced by mining analysts using data mining tools and then deployed for automatic use by other applications to make recommendations.
- Rules engine e.g. Brokat Blaze Advisor, Black Pearl, Versata, Corticon
- BI tools with intelligent agents e.g. Informatica Analytics Server, Business Objects Broadcast Agent, Microstrategy Narrowcast Server
Combining The Options
Of course a number of these options can be combined to achieve dramatic business benefit. Consider the use of pre-computed intelligence and real time analytics to get high performance live recommendations. Business intelligence can be written back into the data warehouse and accessed on demand by a decision engine to make a live recommendation to a sales person or a customer service representative who can then use the intelligence to increase the value of a customer for example – see Figure 6.
Figure 6 – Pre-Computed BI and Real Time Decision Engine
However for me, it is the use of real time analytics and portal technology that really blows BI systems wide open and delivers enormous competitive advantage. In all my years in warehousing, this is the most exciting thing I have seen and is particularly leading edge. Consider the extended analytical process mentioned earlier. Take the steps analyse, interpret, decide, act. Remember that the “action” in this process could itself be a whole process that is triggered by BI being automatically interpreted. We have already discussed the significance of workflow here. What if the steps in this extended analytical process could be joined together in a workflow and given a name? For example, what if the process was called
|“Best prospect to go and see next”|
|or||“Cash flow warning early alerter”|
These are all examples of real time intelligent business processes. To decide the best prospect to go and see next we simply need to automatically analyse the data, automatically interpret intelligence, automatically decide, and automatically recommend who to go and see next.
Now publish each named real-time analytic process in a portal as a URL and then you get something like what can be seen in Figure 7. This is single click business automation. Live recommendations to employees, customers and partners on any device in real time via a portal anywhere in within or outside the enterprise – all via a single click. It is the click that triggers the execution of the whole process to produce a specific recommendation. A fully integrated BI system that is, for all intensive purposes, totally transparent to the user. No BI tools are visible to the user; all they see is a URL. This is an incredibly powerful concept being pioneered by companies like thinkAnalytics and IBM, Macromedia etc. It is this for me that really embeds BI into the heart of business operations and into being a real business driver.
I would ask how many intelligent business automation processes you can think of for your enterprise. Of course many of these actions can be non-visual. Automatic re-ordering of inventory if demand is exceeding supply would be an example. Here the action is simply a transaction invocation to re-order inventory. Portals offer even more however. For example, alerts can be flashed on a portal and we can create personalised dashboards across and beyond the enterprise.
Summary - Which approach to use when?
Given that there are a number of approaches to integrating intelligence into operational applications, a key question is which approach to use and when to use it. Operational reporting should be used for every day basic information needs. Accessing pre-computed business intelligence is a high performance option and is especially effective when the intelligence is attached to customer records. Key pieces of intelligence can then be accessed from browsers and mobile devices. This is particularly effective in front office business operations. For example sales force, customer service and even field service employees could access pre-computed customer intelligence. Tightly integrating analytic components into operational applications is clearly for focused specific use of intelligence in certain business areas. It is likely that new applications will take advantage of this option although this is static up-front integration that may require maintenance down the line. BI web services on the other hand offer dynamic integration of BI into operational applications without the costs of expensive systems integration. Increasingly, BI web services will be dynamically integrated with and shared across multiple operational applications as companies adopt so-called ‘private’ (internal to the enterprise) UDDI directories. BI web services offer a cheaper more flexible application integration option and distribution of BI via XML cubes. In addition BI web services will allow quick dynamic integration of BI tools and analytic applications into portal products thereby allowing companies to push intelligence out to the masses. Portals have already become the fast, economical way to deliver BI to a mass user base. If you have multiple BI portals however, these will need to be plugged into corporate and e-business portal platforms to tie them together with other content and to allow users to access BI produced by multiple BI tools. Portal can also integrate BI systems and tools with other operational applications and content at the presentation level. Finally real-time analytics is again mainly for use with customer intelligence in driving web personalization, real-time Internet marketing and to deliver BI in the form of live recommendations to all front office customer touch-points. Over time this technology will drive business automation and be used more widely across the enterprise. Combining real-time analytics with portals will see greater uptake in the next few years.
Upcoming events by this speaker
- From 05/02/17 to 05/03/17 - Building an Enterprise Data Lake for Enterprise Data-as-a-Service
- From 05/04/17 to 05/05/17 - Predictive and Advanced Analytics for BI Professionals and Business Analysts
From 06/05/17 to 06/06/17 -
Big Data and Analytics:
from Strategy to Implementation
- From 06/07/17 to 06/08/17 - Enterprise Data Governance and Master Data Management
- From 06/22/17 to 06/23/17 - INTERNATIONAL DATA, BUSINESS INTELLIGENCE, AND ANALYTICS CONFERENCE: Building the Data Driven Smart Enterprise