Skip to main content
Live from Gartner Symposium ITxpo - Analytics: Action Based on Integrating Processes and Applications

I am attending Gartner Symposium and ITxpo this week and blogging as I go.

Gareth Herschel presented on Analytics: Action Based on Integrating Processes and Applications. A disappointing crowd to one of the presentations I thought looked most interesting. Maybe the 8am start put people off.

He asserted that we have more and more data from digitizing business - almost exponential growth curve. Cost for hardware and software also falling. But need to grow analytic capability in proportion to this data as the data tells us what is happening as well as an ability to act on this.

  • Knowledge gap - between collection and understanding
  • Execution gap - between understanding and acting

This is similar to the concepts Richard Hackathorn has discussed. Gareth says you cannot jump these gaps other than at once. This means using analytics and turning around our usual perspective - decide what we want to know and figure out how to get the data needed rather than capture data and then see how we could analyze it, present it etc.

His recommendations were to focus on those decision areas where you have support from managers and to balance investments between process flexibility (rules) and analytic insight (predictive analytics). EDM, of course, does both. You might also enjoy this post on recreating the corner store and on the future of CRM as well as my Predictive Analytics FAQ and EDM FAQ

BTW Fair Isaac is at booth 305

Drivers for the increase in use of analytics

  • Analytics can test assumptions in our core business processes - we think we know and we take decisions appropriately. Does not respond well to change and can have "group think" problems where everyone assumes the same thing is true.
  • We all have more segments, more products, more competitors than 5 years ago and this complexity requires new approaches - analytics or, I would say, Enterprise Decision Management.
  • Gartner is trying to define an integrated business intelligence and performance management framework. Within this the analytics are separated into process-driven analytics (I would call this EDM as it is about bringing insight into a business process) and analyst-driven analytics (BI)
  • Gareth identified three values that could be added by analytics:
    • Analytics can add value to a process (up-sell or cross-sell in a call center) but this will often be different for each customer, each segment and that's why analytics adds unique value.
    • Analytics can also improve the process to make it run more effectively through better understanding, clarity of implications and more appropriate reactions.
    • Analytics can also cut steps
  • Me I would say ALL of these are things that analytics can do ONLY if it is used in an Enterprise Decision Management way.
  • Examples:
    • Insurance claims, remove process steps using analytics to identify those unlikely to be fraudulent and eliminate steps to settle quickly. Classic EDM as the analytics is only part, need rules about customers, regulations etc. Claims is a classic EDM area.
    • Equipment rental company added a survey to see who would recommend them. Presenting this in a dashboard did not help but used rules to these to drive triggers to follow up on good, bad or indifferent responses. Gareth discussed this as analytics but I would say it is more about mining data for business rules.
    • Another company did customer satisfaction surveys, took the 20% who responded and data mined this to identify kind of customers most likely to be happy or unhappy and then fed all happy ones to cross-sell and all unhappy ones to retention, not just the 20% who responded. Classic data mining/analytics for customer segmentation.

Success factors for analytics

  • Must balance analysis with ability to respond to it - EDM is about managing this correctly. If we over analyze we end up with what Gareth called "Knowledgeable impotence" a classic driver for EDM. If we have very flexible processes but don't understand them then we end up with random differentiation, another one.
  • Poll very much focused on both (56%), 29% felt needed insight but had flexibility only 15% felt had insight and needed flexibility
  • 10 types of analysis - optimization, simulation, predictive data mining, descriptive data mining, forecasting, statistics, ad-hoc, visualization, reporting/dashboards, events/alerts. Some of these come in BI, the others are really key for EDM
  • Ability to respond, flexibility, is also important. Gareth did not make this distinction but I think both process and decision agility are called for and that EDM is ideal for decision flexibility.
  • Tying insight to execution means good business logic (rules) and processes.
  • Need to match analysis to assumptions and business strategy otherwise will get confusion.
  • Need to create a learning cycle process that creates value from analytics e.g. if care about churning then not just who might churn but do we care, can we do anything about it etc.
  • Deployment was next. Today most analytics are included by jumping out of a process, doing some analysis, coming back. This works well for very ad-hoc analysis e.g. in a strategic process, or where there is a lot of nuance to the decision. Process-driven or EDM-like solutions feed analytic insight into the process directly. The second approach works well with unskilled consumers of the data (call center reps for instance) or there is no-one as it is an automated process. Lastly he differentiated between re-use of decisions across processes and localized ones in an application suite. Me I think the first is better as it allows for reuse and consistency but using rules and analytics in a single application is still better than nothing.

Market evolution

  • He put up a four quadrant chart with an axis from domain-focused to enterprise breath, and integration to insight. He had Fair Isaac in the Enterprise Analytic category which is probably right but I would note that Fair Isaac also has a strong domain expertise in Banking and risk.
  • The polling was interesting with each spot
    • Enterprise application suites and domain process specialists came top
    • Enterprise analytic products came a close second
    • Domain analytics came a poor last

He got an interesting question about real-time analytics. He said, and I would agree, that real-time delivery of analytics works well and that it is not clear that building analytic results real-time offers more value. He felt that real-time was most appropriate where there is risk involved in the decision. I think he means that you need analysis of the most up to date data.

Great session.

Technorati Tags: , , , , , , , , , , , , ,

related posts