Blog: James Taylor Subscribe to this blog's RSS feed!

James Taylor

I will use this blog to discuss business challenges and how technologies like analytics, optimization and business rules can meet those challenges.

About the author >

James is the CEO of Decision Management Solutions and works with clients to automate and improve the decisions underpinning their business. James is the leading expert in decision management and a passionate advocate of decisioning technologies business rules, predictive analytics and data mining. James helps companies develop smarter and more agile processes and systems and has more than 20 years of experience developing software and solutions for clients. He has led decision management efforts for leading companies in insurance, banking, health management and telecommunications. James is a regular keynote speaker and trainer and he wrote Smart (Enough) Systems (Prentice Hall, 2007) with Neil Raden. James is a faculty member of the International Institute for Analytics.

March 2009 Archives

Last week I posted a couple of times about my impressions from the SAS Global Forum. In one post I said that "SAS customers talk about the great results they get when they put their predictive analytics to work in operational systems" so I thought I should expand on that a little, using the customers I heard during the one day I attended.

One panel I saw had two customers - Stephan Chase, VP Customer Knowledge at Marriott and Eric Webster, VP Marketing at State Farm - talking about putting predictive analytics to work in their businesses. Stephan, for instance, pointed out that 1% of Marriott's customers generate 20% of their revenue so putting customer knowledge to work really makes a huge difference. He also made the very valid points that more data can sometimes obscure not inform and that analytics must support both the timeless core of a business and the more innovative edges. Marriott's use of analytics drives their pricing, loyalty and marketing with models embedded in all sorts of operational systems.

Eric pointed out that Insurance is an information business - there's no physical product. He made a great point that while State Farm is a data-driven business it is also a relationship-focused business. He saw the power of analytics in its ability to help the "faces" of State Farm (the agents, claims processors etc) build a relationship. This use of predictive analytics, again deployed into front-line systems, enables people who have never met a customer or prospect interact with them as though they have known them for years. These predictive models recreate the corner store, as Richard Hackathorn likes to say, and keep loyalty where you want it - with the company. They do so by being embedded in operational systems and by delivering predictions about individual customers.

Another session had Scott Overby of Discover talking about their Teradata/SAS data infrastructure. I like Discover as a Decision Management example (see this blog post about Michele Edelman's presentation at the Decision Management Summit for example) and it was fascinating to hear Scott talk about the challenges of building the data infrastructure needed by a decision management business like Discover. The need to focus on scale, on real-time analytics and operational data, and on avoiding brittle systems caused by over-optimized analytic data design all came up. The Enterprise Data Warehouse that Discover has implemented is designed to support both BI/business analytics and Decision Management/predictive analytics and this is the way to go for those who want to maximize the value of their data. Personally I just love the way Discover talks about making hundreds of new data elements available to decision services as a minor change :-)

A final customer session was also a speaker from Marriott (whose name I failed to record) and she spoke about revenue management at Marriott. Not only is this intensely analytic, it too drives these analytics into the transactional pricing and booking systems that everyone uses. The use of adaptive control techniques to test and learn is crucial to revenue management and she had a great phrase Marriott uses "Success is never final". She also made the point that when measuring performance, of decision making say, you should consider both performance and opportunity. Just how good could the decision be?

Some great customers with great results. All done using SAS software and then putting the predictive analytics developed to work. Decision management, to use my favorite phrase.

Posted March 28, 2009 5:38 AM
Permalink | No Comments |

Madeline Albright gave a great presentation at the SAS Global Forum in Washington DC this week. Several of her bon-mots are in the title but there were many others, some of which are below. Each of them struck me as relevant to readers of this blog:

  • Facts not Fears
    Businesses all too often do things based on their fears, not on the facts. They price lower than they could because they fear customer won't buy, for instance. Use analytics to find out the facts (and to find out what the facts mean) and use business rules to act on them.
  • Confidence not Certainty
    Being confident is critical to automating decisions - you must be confident in the rules you are proposing, for instance, if you are to allow them to treat your customers on your behalf. But you should not be certain, you should test and re-test assumptions, simulate the changes you are considering before deploying them and challenge your approach using adaptive control.
  • Critical Thinking not Wishful Thinking
    Wishful thinking like "I want customer retention to improve so I will set that as a target" is not as useful as critical thinking like "I want customer retention to improve so I will identify the decisions that make a difference to customer retention and design decision making strategies that will make a difference"
  • Diligence in testing assumptions
    Decision analysis - the use of performance management tools to manage the decision making process itself - is essential to being usefully critical of your own decision strategy. Adaptive control is key to these three as it provides the approach and the infrastructure to constantly challenge the way you do things. You don't pretend to be certain or that your assumptions will always be true. You don't hope you have the best approach in use, you test and learn.
  • From information scarcity to information overload
    with so much information more and more potential consumers of data will not be able to cope. Analytics with its ability to focus in on the business implications of all this data has much to offer.
  • Any summary is dependent on the perspective of the summarizer
    And this increasingly means the math geeks (metaphorically) in the basement. Is their perspective the same as yours? Do they measure model accuracy in terms of K-squared or business results?
  • Comparative and Historical Perspective
    Decision analysis should involve both kinds of analysis.
  • Two other great phrases she used included Henry Kissinger's "constructive ambiguity" in public utterances/press announcements and Secretary Albright's description of herself as "an optimist who worries a lot". She was a fabulous speaker and both interesting and amusing, despite the seriousness of her topic.

Posted March 26, 2009 8:14 AM
Permalink | No Comments |

I spent a couple of days with thousands of SAS users this week at the SAS Global Forum 2009. There were some great sessions and, as usual with SAS, some terrific customer stories and I suspect I will write a couple of posts here or on my other blog. This post, though, is about the theme - Leading with Confidence in an Era of Uncertainty.

The idea behind the theme, as I heard it, is that SAS delivers fact-based confidence for your decisions. Solid business analytics (of which more later) replace hunches with facts and take the guesswork out of decisions. While I understand that individual SAS users want to feel confident in their decisions, I think that the companies that use SAS want much more - they want accuracy, better decisions, optimal decisions. Sure, they want to be confident in them too but the confidence is secondary to the need for decisions that are materially, measurably, practically better. Leading with accuracy rather than with confidence.

Multiple speakers brought up the need to "understand data and information quickly" as though this was a business objective in its own right. But I don't think it is. Businesses need to act on data quickly and accurately (there's that word again). Understanding it is a critical step but not the payoff.

"Deliver the right information to the right person at the right time". Well yes but why? So that the right decision gets made - that's the purpose of it all, that's what adds value to the business. So why not focus on making the right decision and if that means delivering information to the decision maker, great, make sure its the right information etc etc. But perhaps it means putting the right rules in the system or optimizing the constraints correctly or some combination of these things. Decision first, everything else only after.

Stephen Baker spoke about his book Numerati and one of his examples made this point, at least to me. He was talking about his own industry - media - and the challenges analytics are creating for it. In particular he used an example of ad pricing and the need to tie ad pricing to analytics about the impact of the ad. All true but companies just like the one he works for are automating ad pricing using rules (there are lots - color, size, scope etc) already. Using knowledge-worker focused analytics would let a few pricing analysts make analytically based decisions but the business cannot afford to go back to only having a couple of specialized pricing analysts who can calculate the price (that's why they automated it, after all). Analytics should be fed into that process to alter/influence the pricing rules so that the automated decision is correct but getting this right is going to take more than just analytics, it is going to take decision management with rules and analytics.

Over and over I hear SAS customers talk about the great results they get when they put their predictive analytics to work in operational systems. They use the integration with Teradata, batch scoring, hand-coding of predictive models, loading SAS models into rules engines and more. They understand the power of predictive analytics to improve their bottom line by improving the operational decisions in their business. I am even certain that SAS understands this. Yet somehow this never seems to come up in the core SAS pitch and that worries me.

Confidence is not the issue, accuracy is. Information is not the goal, better decisions are.


Posted March 24, 2009 1:51 AM
Permalink | No Comments |
In an article on Intelligent Enterprise titled Predictive Analytics: What Have You Done For Me Lately? Dave Stodder summarized a few points from Predictive Analytics World:
predictive analytics could enable companies to hang tough during lean times by helping them identify the most important customers, focus resources on critical business processes and uncover fraudulent behavior before it causes serious damage.
This is an interesting summary as it shows how predictive analytics can apply to strategic decisions (who are my most valuable customers), tactical ones (how do I reassign resources between projects) and operational ones (is this transaction fraudulent).

Of course the first one also requires that you focus on operational decisions, after all you must make decisions about how to treat individual customers in a way that reflects their value to you if you are to put that insight into customer value to work. Troy Powell, over at Walker, had a great post around this topic in which he discussed how important it was to understand how you were going to impact the business.

In fact I was struck throughout the event by the focus of predictive analytic users on operational decisions. It wasn't the only thing they used analytics for but it was sure the most prevalent. You can check out my blog post summarizing what I learned at Predictive Analytics World too if you like.

Posted March 17, 2009 6:06 PM
Permalink | No Comments |
In a recent piece Data Strategy Journal - DATA: YOUR ULTIMATE ASSET Thomas Redman had a nice intro on the value of data:

An exercise popular in many training courses goes something like this. The class is asked to imagine a fine antique French desk, recently purchased for $20,000. Atop the desk sits a brand new laptop computer, complete with all the bells and whistles, that cost $3,000, and a compact disk that cost thirty cents. The CD contains the only known list of the names and purchases of the organization's fifty largest customers. Now, the exercise goes, a fire has started and you can only save one of the three. Which do you save?
Of course the answer is the CD because it has the data and the data is worth far more than anything else. Thomas explains why:

data are not just unique; they are the organization's ultimate proprietary assets. 
While I don't disagree with him, it seems to me that there is more to it than this. If we don't act on what the data tells us then it is not really much of an asset. So perhaps we should consider the decisions we make with the data to be our ultimate asset. Some of this data is "important for operations and tactical decision-making" and the challenge here is similar to one he describes for data. Just as "too many organizations subordinate 'data management' inside their IT departments", so too many subordinate decisions inside their information systems. To resolve this you need to apply Decision Management.

One of the key values behind Decision Management is that you should manage decisions as a corporate asset. When customers interact with you they consider every decision you take to be a corporate one - that is, a deliberate one. Thus if your website, your call center or your agents make a poor decision, an inconsistent decision or an out of date decision, it reflects on your whole organization. Yet, every day you must make decisions faster, across more channels and product lines. This makes it harder and harder to ensure that the decisions your organization takes are the best ones and the ones you intended to take. You need to find ways to derive insight from that data and push that insight into your operations.

As you move to improve, innovate, automate and even "off-shore" your processes you must face the fact that the differentiation of your company from others increasingly comes down to how you make decisions. You can inject sustainable competitive differentiation into your processes effectively only by identifying the key decisions and using a Decision Management approach to take control of these decisions. Even if your processes look like everyone else's, making your decisions unique (and better) will ensure customers understand why they should do business with you. Many of your processes remain fairly static over time while the risk assessment, value calculations, policies and regulations that determine the outcomes of decisions within them change all the time.

You can use decision management to develop distinct Decision Services for critical operational decisions - what product to offer, how to price and underwrite a policy, etc - and so make decisions easy to manage, improve and change independent of the processes or systems that use those decisions.

A last word from Thomas. He says "efforts to put all this unique data to work must be led by 'the business.'  Better still, a Chief Data Officer" or perhaps a Chief Decision Officer.

Posted March 9, 2009 4:07 PM
Permalink | No Comments |
PREV 1 2