BRForum 2007 Expert Panel: Emerging Trends in Enterprise Decisioning
- Ronald G. Ross, Business Rule Solutions LLC
- Don Ames, Precedent Insurance Company
- James Taylor, Smart (enough) Systems LLC
- John F. Elder IV, Elder Research, Inc.
- Mark Proctor, Red Hat
- Kirsten Seer, Business Rule Solutions LLC
- What does 'Enterprise Decision Management' mean?
- What industries are ripe for EDM and don't know it yet?
- What do you recommend for getting started?
- When will we see a robust methodology for the 'EDM approach'?
- How can we get these diverse disciplines to converge ... to work together?
- Do you have a closing message?
Good morning everyone. My name is Kristen Seer. I'm with Business Rule Solutions, and I'm going to be your moderator for this morning's panel on Emerging Trends in Enterprise Decisioning. I work with Ron Ross and Gladys Lam, who most of you have either seen or met by now.
When they invited me a couple of months ago to moderate this panel, my first response was, "Oh great! That sounds like a lot of fun. I'll do that." My second response was, "What the heck is 'Enterprise Decision Management'?" I had never really heard of that before. So, Ron was good enough to get me an advanced copy of James Taylor and Neil Raden's book, Smart (Enough) Systems — an excellent book on the topic ... really covers it from beginning to end.
As I started reading, my first response was, "Hmmm ... making better operational decisions. Now there's a good idea! Who wouldn't want to do that?" And then, as I got through the book, and learned about predictive analysis and modeling and having that in line, and how business rules fit in, and having the adaptive control and all the technology that went with that, and all the processes and infrastructure that went with that, my second response was, "This is going to be really, really hard to do." So I'm going to challenge the panel today to enlighten us with: how this is possible, why we should care about it, and what the emerging trends are in this area.
As far as logistics are concerned, being the third panel at the Forum, I'm going to reuse the rules of two of the other panels. The first rule I'm going to reuse is Gladys' rule on cell phone usage and buying the new outfit — revising the rule, of course, so that I'm the one who gets the new outfit. <laughter>
The second set of rules I'd like to reuse are Steve Hendrick's rules for the panelists — particularly the rules of conduct on no filibustering. I think that's a really good one.
As far as logistics are concerned we'll be using the same approach where I'll ask our illustrious panelists to introduce themselves — talk about who they are and what their role is and some of what their experience is within the EDM space. I will have an initial question for them, and then I want to turn it over to all of you because I know you're brimming with great questions. We've had really good questions the last two panels so I'm sure you've got great ones for these guys. And really challenge them ... give them a hard time.
Let's start off. First, if you could start introducing yourselves with your name, company, and role, and your experience with EDM so far. Then the question is: Why has EDM become such a hot topic all of a sudden? What are the most important trends you see in this area? Mark, if you wouldn't mind starting?
Yes, my name is Mark Proctor. I'm from a company that was originally called JBoss, which was acquired by Red Hat. I work on the Drools open-source rules project, which is the biggest open-source rules project — it's what we call a production rule system. It's helping take rules mainstream. We got 20,000 download attempts last month and we're growing rapidly. We've focused mostly on declarative programming.
Why EDM? EDM is nothing special — it's been around a long time. What's different is ... how do people communicate? They have to have the terms and the language. What's happened is ... everyone has heard of Web 2 and Ajax now? Ajax, the term, has been around for what ... 2 or 3 yrs now? However, developers were doing it for six years before that. The difference is there was no language for people to communicate to say what they're doing. They know they're doing Ajax but they don't know the words ... so there's no way for people to get some hype going around things ... to make sense of it.
So what really has happened is mostly ... James, he's responsible for this ... the term 'enterprise decision management' has occurred and has allowed people to communicate about a particular problem — because before that there was no way to categorize this ... there was no way to build hype around a single term.
So in reality ... nothing's new; it's quite simple. The difference is now it's being put in a language and in a term that people can understand, which then allows the hype to build around it and for it to be deployed in the enterprise.
That's probably not the answer people wanted to hear but that's what I think is the truth. It's simply that hype around the term has become prevalent.
I'm John Elder. I run a data mining company. Data mining can be thought of as the entry level ... the first step, in a lot of ways, where you discover the rules from the data. We've been in business for 12 years and have about 15 people — very applied-research oriented. So, we run into problems that are mostly statistical, where statistics and business intersect.
You have the real world ... a fuzzy, squishy real world that you have to somehow translate into a crisp technical problem that can be solved and optimized. It's an art form to try to translate what the business is really after, into something that the machine can crunch on.
Realizing that it's a complete system and that you're going from building and discovering rules, to managing them and monitoring them, is relatively new in the circles that I'm in, but it's extremely important. Let's say you're a closed system. The fact that you're discovering rules — say fraud detection, for instance — and now applying those rules means that you're shining light on an area. The cockroaches are going to go somewhere else.
So the problem is going to change out from under you. And the overall economy and what your competitors are doing will change under you as well.
Models need constant monitoring, update, and maintenance. That's a little less exciting than the model discovery phase, where you're discovering new laws of nature, new relationships between data. But it's extremely important. So finding a great framework for doing that ... it's good to see the tools that are emerging, that allow you to control that entire life cycle. I see it as a very positive development.
I'm James Taylor. Right now my description of myself is a little vague because I've just ceased working at Fair Isaac, after being there many years. I'm probably mostly to blame for the book that Mark has up here. I notice that Mark has it in crib form and Don has Google up so that he can search for whatever terms he needs. I'm being out-referenced!! <laughter>
I'm actually going to probably surprise Mark by agreeing with him completely that when we came up with the phrase 'enterprise decision management' it was really to try to label something we saw working. If you look back the last couple of decades of how people have addressed (particularly) risk-centered decisions: Should I give this person credit? Is this a fraudulent claim? ... that kind of stuff. You see what we talk about in the book being applied very successfully.
But what there wasn't was a way to link these things. People talked about, "Oh, I have an origination solution or a fraud solution." And they didn't really categorize it. So what we try to do is to come up with a way to categorize it and then generalize it and say, "The challenge here is that, yes, you have these operational decisions, and in the past you only tried to solve the problem of automating those decisions where there was a big payoff."
When you think about credit risk. If I make a good decision to offer you a credit card, I might make a $100 profit from you over the lifetime of the card. But if I make a bad decision I lose $10,000 tomorrow. That makes me really focused on how good a decision I make.
And when the decision is less black and white than that, you need more of a process; you need more experience. Those problems are not going to get addressed immediately.
That I think is what you see now — these techniques that have proven themselves now have a label. They have a sort of home. And now they're starting to get applied to problems that, in the past, would have been considered not a high enough ROI. Now the technology is more mature; the techniques are more mature. So you see it being ready to apply now to a much broader array of problems.
Good morning. I'm Don Ames. I'm Chief Technology Officer for Precedent Insurance Company, a direct-to-consumer, internet-based health insurance company.
As the CTO for that company I was responsible for the selection and implementation of a rules platform to support our online underwriting and reflexive application process, in the website, and a number of other back-office practices that are being implemented.
The hot topic — why is it a 'hot topic'? My theory here is that, over the life of a given company, certain business processes are not just automated, they are automated again, and again, and again, and again. The new CIO comes in and needs an 18-month honeymoon so he says, "We're going to switch platforms." <chuckles> So it gets automated again.
But the fact is that everybody's for sale. Every property on the block is for sale. Industry consolidation — whatever industry you want to pick — is always something worth thinking about. Well, if you're always for sale, you might as well be ready to move. And if you're going to be ready to move, you've got to expose your business rules ... you've got to expose the way you make decisions. We spend all this time in a traditional SDLC type of environment ... we spend half of our time simply trying to understand what the heck we want. And then we spend the other half of the time encrypting, encoding, and embedding it in an application.
If we're going to be ready to move we've got to not do that anymore. That's why this is a hot topic.
For me, the emerging trends that I'm seeing are really associated with open source. I think while you've got the large, monolithic specter of consolidation on one side, you've also got the ability that it is cheaper, quicker, easier to start a company because you have great, mature, robust tools available to you, without the kind of entry cost we used to have. And you've got a great pool of talented people to support those tools so that they are commercially viable.
Last, but not least ... Ron.
Ron Ross ... co-chair of the conference. Also, principal and co-founder of Business Rule Solutions. Basically, my role is to do what Gladys tells me to do, when she tells me to do it. But that's often extremely productive and a good way to go about doing things.
Business Rule Solutions was founded ten years ago. Gladys and I founded it with the explicit objective of (1) connecting better with business problems and communicating better with business people in the area of business analysis and true business modeling, and (2) using business rules to address the problem of encoding knowledge and deploying that more effectively in the organizations. Over those ten years we've pioneered, I think, some very interesting techniques and truly helped make some progress in some companies in certain areas.
Now, before that life, I was editor and publisher of the Database Newsletter ... for some twenty years. Any of you remember the Database Newsletter? <pause> Ah, that's nice — at least two or three hands. That's good. As editor of the Database Newsletter, one of the things I learned was the following: In real estate they say "location, location, location" — right? And what I learned in publishing and editing is it's "positioning, positioning, positioning." What's required for things to happen ... people need core ideas. They need a center around which new developments, new activity, initiatives, explanations, discussions can be focused.
So, what I see has happened — and whether it's a center of hype or whether it's a center of interest or a center of excellence — I don't really care what kind of center it is as long as it is a 'center'. So that's fine with me.
I do, myself, think that it is a new center that has emerged — the pieces have existed before but 'EDM' represents a new positioning of that. I think it's a new dimension altogether with respect to business rules which, of course, has been our focus for a long time. So I had to reach out and understand more about that dimension, but the more I see, the better I like what I see. I think James and Neil are to be congratulated for introducing that center effectively.
As I said in my keynote, I think it's extremely important to keep everything in balance. And Business Process Management needed a counter-balance, which I think EDM provides very effectively. Technical initiatives such as SOA need a counter-balance, which is EDM. There are a lot of reasons why EDM provides some balance that hasn't been there before.
And, finally, I feel very strongly ... even though we don't recognize it, since we're in the midst of a major change in society, that in fact, finally, we are moving toward a knowledge economy. And that the ability to encode the knowledge of businesses, in order to do better decisioning, is in fact the major enabling technology or technique in moving toward knowledge economy. And that's just exciting! As an observer of things happening in the world (and as a lifetime activity) that would be my point on the question.
Okay — thanks very much! So let's turn it over to all of you now. If you raise your hand I'll come over with the microphone.
|[from the audience]: I have a two-part question. First is the assumption that we all know what 'enterprise decision management' is. What does it mean? I'd really like to hear each one of you give your definition of what 'enterprise decision management' is. Then the other question is ... because you're using the word 'enterprise' that must imply that there are some decisions that are not 'enterprise'. How do we know which is which?|
Should I go first? I thought about leaving it as a test for the rest of the panelists. <laughter>
The book has a formal definition that I'll start with: It's an approach — this is not a technology stack — it's an approach to automating and improving operational decisions. That's what it boils down to. So taking control of the day-to-day decisions that run your business and automating them in a way that enables you to manage them (as Ron said) separately from your processes, separately from your systems, so that you get your maximum amount of agility, you get consistency across your enterprise, and you get an ability to inject precision by learning from the data that you have.
One of the problems that the name causes is that people sometimes think that 'enterprise' means "you must do this across the enterprise." And, in fact, it's enterprise decision management because the idea is that you should manage decisions as an enterprise asset. I think David Straus made this point yesterday. We manage so many things as assets — we think about them; we think about reusing them; we think about controlling them. And yet we often have decisions that are made almost incidentally. They are made based on code written by a programmer twenty years ago, or they're made based on a judgment call by some employee who's been with us two hours and is answering the phone. That's an inappropriate response for an enterprise as it moves (to Ron's point) into a knowledge-based economy.
I'll add something to that. Does anybody else want to take a shot? <panel: Go ahead.>
It's hard to disagree with the author of the book but ... what the heck. James puts more of a technology twist to it — or emphasis — than I probably would. I look upon it more as a business strategy for addressing the knowledge problem in the organization as a business proposition. And, of course, you need technologies to do that. But I think I would look at it more as a business initiative.
I think what James and Neil did excellently in the book was to explore and explain the concept of 'full loop decisioning' and that is to execute your business processes and activities, to capture the resulting data, to analyze that, and then the fourth thing (not being the least from a business rules perspective) is the deployment. It's all about turning insight into action. I think that's a very nice way to summarize their approach. So, just very minor, slightly-different positioning, but I think in agreement.
Now, with respect to the 'enterprise' question? I think that's an excellent question and I've even had the discussions with James offline, and he's said he wishes that he hadn't put the word 'enterprise' in there. He probably won't admit this...
No, I admit that. It's very open to misinterpretation.
Was that an admission of guilt? He's nodding his head, for the record.
But here's the way I look at it. John Zachman, who will be here in a bit for lunch, labeled his Framework the "Enterprise Architecture Framework." But what we have done for many years is to apply the Framework — his Framework — as a thinking tool that can be used for problems of any size. And I don't think that John has a problem with that, although he may be too nice to say so.
I think the same would be true of James and Neil, which is that they have provided a thinking tool on 'full loop decisioning' that can be applied to problems of relatively any size, with respect to enterprise scoping. Is that correct, James? <yes> So I think that's the way to look at it. 'EDM' may be the unfortunate handle that it gets, but if we use it correctly I think it can be applied at a lot of different scopes.
Looking at it from the technologist's side of view, one of the enablers of this has been SOA. Now, SOA is just a reinvention of a term, CORBA (and so forth). But SOA has got the hype; it's pulled everyone together, to try to understand what is SOA.
Now, what is the next stage? What EDM does, it gives you an application of SOA ... because no one can actually define what SOA is and what they do with it, so it's kind of a very floppy term for 'services' and stuff. What EDM does, it takes it a little bit further and asks us to get more concrete definitions of how we apply SOA.
So, as always, when you come up with a way to define something, and you come up with an understanding of that that allows further and new understandings. So, everyone came up with 'SOA' — SOA has enabled what James has done — it allowed him a way to build up a way to explain these things. And then the next stage from this — once we have terms to talk about things that we can all understand — is to try to identify technologies that enable EDM. If you look at the way you typically have process companies — and workflow companies and rule companies — now you'll find they're moving together because you need your rules and your processes to define Enterprise Decision Management.
The next stage from that is 'Which other things?' ... because it's all about declarativeness — describing the systems. That's CEP; which is the biggest hype of the moment. (CEP, for those who don't know, is 'complex event processing'.) There's going to be a lot of market consolidation there. I think you're going to see all the rules and process companies starting to have CEP offerings. BEA, for example, have just acquired something for that.
And this will continue. Now that we have an understanding of what Enterprise Decision Management is we can start to identify the 'chunks' that complete that. In that way you build up a platform, and an understanding, which enables the next stage. And so it will continue.
|[from the audience]: This is a question for everyone except Don. (Don is off the hook because he's doing some fine work as CTO of a health insurer.) What industries are ripe for EDM and don't know it yet?|
And don't know it yet?
Do you want me to go first?
I would say the two that strike me.... One's not really an 'industry' but I'll get to that. The first one is governments. Governments have a huge amount of data; they're often very reluctant to use it, or expose it too much, because it's often very personal data. And then they have all these regulations they have to follow.
So what EDM does is gives an opportunity to say, "We can improve the way our systems work" without having to reengineer them all. <aside> Is anyone here from the IRS who likes big reengineering projects?? <laughter> It's really hard to reengineer these big systems. It's just really difficult. You've got millions of people using the outputs.
So I think it's an opportunity for government agencies to have a way to bring more automation in, to up the quality of service they provide to their citizens, to use the data they have without having to expose it to people in offices, just by data-mining it and by analyzing it to see what they can learn from it.
So I think there's a huge opportunity in the government section. And that might not be as obvious to people in that sector. I think most of the other sectors pretty much have noticed. Insurance companies, banks, and so on.
The other one ... there is a great Gartner phrase; they called it a 'vertizontal' ... which is to say that it's not really a vertical (an industry) but it has characteristics of one — this whole focus around 'customer experience' and CRM. I think if you are really trying to manage the customer experience, one of your challenges as a modern organization is that many of the interactions that your customers have with you are not with you — they're with your systems. They're with your web site, your IVR system, with the system that prints the letters you send them, the system that prints the statements, the emails that you send them.
And so if you really want to improve the customer experience you have to separate the decisions about how you're going to treat your customers next, from the delivery vehicles. So I think there's going to have to be a realization that, "If I really want to treat my customers right I've got to separate that from the 'systems' that I deliver that experience through."
You see a few people starting to 'get it'. I was talking with David, who talks to a lot of people in this space, and it's amazing how many people there are who want to do like the people who've 'really got' customer-centric right, but they think they can do it by buying a new package, or changing their web site. I tell them, "No, ... you've have to actually think about your customer treatment decisions as a sort of corporate asset."
I'd be curious what John sees as the growth in data mining too ... because I think these are very similar.
In data mining the 'hot topics' are fraud detection and investment management decisioning. You look especially for situations like insurance, like stock market, like fraud, where a small edge in intelligence leverages up to a very big return. If you said the stock market is going up you'd be right 52% of the time. If you can get that up to 55% of the time you can make a great living.
So sometimes the edge of intelligence over chance that you need is very slight. And so these systems can have very non-linear effects.
For example, with fraud.... With sales you have to actually make something and sell it and there's a small margin. If you discover fraud, it all drops to the bottom line. And it also discourages more fraud and saves you from the inevitable death of the exponential growth of the disease that's inside your system. So, you need these systems.
I think another opportunity is ... most companies are getting huge amounts of efficiency and they've about wrung it out. I don't know if you've ever talked to the robots on the airlines. When you want to do an open jaw trip you'll feel you are in an endless loop because there's no way to do that. And they won't let you talk to a person. It's very, very frustrating, as a consumer, to be shuttled into automaton world, to be handled 'efficiently'. Nobody likes that.
But if you can do things that way and then allow an outlet, you can use the outlet as a sales opportunity. The point of contact with the customer cannot just be a cost to you as you're handling that problem and paying that person. If you had a decisioning system that showed you what that customer was like and had some cross-sell/up-sell opportunities available for the person handling the immediate matter you have a sales opportunity. So you could turn a loss into a profit center with the right on, fast-moving decision systems.
Rather than focusing on individual organizations and what has been identified, I think what's more interesting is inter-organizational problems that are being solved. SOA and services are allowing companies to publish information and these can be combined in new ways that have never been thought of.
There's a term 'mashups'. Google are getting into mashups ... there's a Google mashup tool; Yahoo is introducing a mashup tool. The idea is that Google is putting all their maps online, Amazon is putting all their books online — all this information is available as services. You can use these mashup tools to combine the information to create totally new services and ways of viewing things.
It's really cool things, like people taking their Google maps, combining them with the information from traffic surveys, and producing into a single map, showing the two correlated together. And then combining that with information from somewhere else that shows pollution. They get these to work together in an orchestrated way to try and help to get some decisioning.
So it's enabling a whole new level of understanding and thinking of things in an inter-organizational way that was never possible before. The barriers to inter-organizational decisioning have dramatically reduced in the last two years.
Just to add to that ... I think it's correct to focus on the kind of problem, necessarily, rather than the kind of industry. As I was thinking about this I was thinking that anybody, any company that has a complex logistics problem, or has a complex supply chain problem, or has a complex product/service configuration problem — which is mashup, in a different way of saying it — is ripe for this kind of idea.
There are very few industries that don't have at least one — and often several — of those kinds of things. So as I was listing industries I decided to shift around to the commonalities between the different kinds of companies.
I think we should let Don answer anyway.
Thank you very much. We'll talk afterwards. <laughter>
Actually, I'm going to follow along with that, but I'm going to back one more level out from that and I'm going to say ... basically, if you work for a company whose requirements never change, you probably don't need EDM. Everybody else does.
|[from the audience]: I'm on the business side of IT, and I'd like to direct my question first to you, Don. There's a bewildering array of ideas and methodologies and technologies that we've heard about. In your opinion, how do we know that we're ready to begin undertaking Enterprise Decision Management? And what's your advice about where to start on all of the ideas and processes we've heard? We're babes in the woods, compared to the insurance industry, in internal operations management!|
Granted, my experience is somewhat linear so I've got the path and, as Dilbert says, "The best solution to a problem is, by coincidence, the only one you know." First off, I made an assertion in the prior hour, when I was talking about my company's story. My assertion was that the first thing I'd look at is how valid, and how reliable, is your ability to understand the business rules and the decision processes that are embedded in your current business processes (or systems, to be more specific)? And I would venture to guess that for most people it's not reliable.
This is my first Business Rules Forum. I looked up what 'EDM' stood for on the plane on the way down (not that bad, but close). I think that, in large part, one of the things that I believe is that, first off, any business requirements — any kind of documentation that you're going to run across — is at least out of date, if certainly not a relic, a true artifact, and does not represent today's need. It may represent today's system, but it does not represent today's need.
I think that the first thing I'd say about that — with today's tools, today's open source opportunities, today's agile development techniques, and today's highly-portable solutions in the rules space — I'd be hard-pressed to believe that it makes sense to reverse engineer this stuff, to spend all of your time getting better at rules harvesting, when the skills that you have for requirements discovery and translating those into these kinds of platforms are much more readily available ... and probably a quicker path.
So I would say ... rely on those techniques. That would be my leaning.
Just a comment.... In data mining, all you need is a need. You need the desire. "We have this pain and we have to solve this problem." In consulting you learn a lot of times that the customer will present a problem and that's not their real problem. It's almost like a doctor who sees a patient because of certain symptoms and then there's the real, underlying problem. So there's a working through that.
But if you're motivated ... if you feel the pain and can somehow get the people who need to solve the problem — somehow get their interest aligned with solving that problem — then you're ready. You will have a place to hang the ideas and the technology as it comes in.
By way of analogy, Viagra has been the best thing for men's health to happen in the last several decades ... because it gets men in to see the doctor. They don't go to see the doctor otherwise; now they're there and the doc can notice a heart problem (or something else). That's a take-home image. <laughter>
I don't think anyone's going to want to follow that!
No, no one's going to want to follow that ... I did have one thought though.
John comes at it from the data mining perspective; I come at it from the opposite end of the spectrum, so to speak, with respect to business rules. But his guidance to focus in on the real business problem is exactly the right one. Find out where the real pain is and focus on how you can alleviate some of that real pain. So, the same guidance.
[from the audience]: Hi, I'd like to direct this to my buddy, James. I liked your book a lot. The way I see it, it's an enterprise architecture methodology, focused on the deployment part — the operational, day-to-day decision making, which I think is lacking in terms of robustness and in terms of focuses and (as Ron said) in terms of creating a center and its integration with business rules.
That's a great question. One of the challenges in writing a book like this, you've got to write some stuff that convinces people it's worth doing some of these things. And that has to be at a fairly high level, and it has to engage a business audience in thinking about the problem.
But, also, for every person who wants that, there's somebody else who wants to skip past the first three chapters and get a project plan. So the book ends up not doing either of these things. It is neither slim — so that you can read it over lunch — nor has it got a methodology in it.
And I think the thing for me is ... what I notice out there is that there are methodologies for doing business rules; there are several well-known ones. There are methodologies for doing data mining and predictive analytics. And what's interesting is that they have quite a lot in common. They all tend to start with the problem: What is it that you're trying to do? How do you break that down and drive through it? And they use different sources and different techniques.
So I think there's methodology 'out there' and I'm on the fence as to whether they need to be completely integrated, in the sense of one 'uber' methodology, or whether it's more a question of guidance as to how to apply this mindset in different places.
Scott and I are presenting on 'Rules and Requirements" after lunch. Is it a completely different approach to what you might call "traditional requirements gathering"? No. Would you need a different methodology for it? Well, maybe, but you probably can also just use the methodology that you're used to, with a mindset of applying the approach.
So I think what's more likely is that you'll see more detailed guidance on how to use these different methodologies and techniques that exist, to deliver Enterprise Decision Management, rather than a new, completely different approach, from a methodology perspective. I could be wrong about that, but that's my sense. I see plenty of the right kind of material out there, and most IT departments don't need another methodology. So my inclination is towards ways to use the stuff you already have, rather than a new approach.
Ron, as a methodology person, you've got to have a response to that!
Yes. James ... first, there are business problems and then there are technical solutions. So the issue is broader, I think, than the way you frame it, which is — until we learn to undertake business modeling in terms that business people understand and can participate directly in, and be stakeholders in the solutions, we really haven't progressed too much beyond just platform-based, IT silver bullets.
So I would encourage people, rather than falling back on the old methods — whether that's a UML-based methodology or a traditional IT methodology — to look at some of the more business-oriented work that's being done by various people. Let's move to help our companies solve real business problems rather than focusing in on IT pain.
|[from the audience]: It appears that there needs to be a convergence of a number of different disciplines to support EDM, such as predictive modeling, BPM, Rules, BI, .... Do you have a recommendation for how to get everyone playing nicely in the sandbox?|
Let me jump in.
Our business survives on doing consulting. And, to do consulting in data mining, you have to have data. A number of projects die because you never get the data. Or working with the data, once you get it — understanding it — takes up 80% of your time and 120% of the budget!
So, we have learned to not assume that everyone who is on the team is 'on the team'. There are people who are threatened by looking at the data ... because it's "their" data.
A 'great' dataset is one (like recently) where 95% of the husbands were male. That's really high cross-checking. Most datasets are really in bad shape for decision making. Partly because they were built for billing (or something else) and they haven't been looked at for a long time. The data has been stored — millions of dollars has been put into building a data warehouse, most of which projects fail (by the way). It's been a data tomb. Data has only gone in; it never checks back out again.
When you do check the data out — you pull it out to do something with it — you may be setting off alarm bells on some of the people who are on your team. So you have to pay attention to that, from the very beginning, and you have to get someone from higher up to knock heads together to say, "This is important to the whole enterprise."
I would just add that.... Back when I was a consultant for Ernst and Young, I went through a whole training and organizational change methodology. EDM presents an enormous number of opportunities for organizational change: The organizational change of getting your statisticians to worry about what happens to the model, once they've finished it. Getting your IT people to let go of some of the rules and let the business manage them. Getting your business people to stop thinking, "Isn't this great! I can throw it over the wall and then blame IT when it goes wrong."
There's all sort of opportunities for very negative organizational dynamics. One thing to keep thinking about is (as Ron was saying) it is a big change to the way companies run their business. It is amazing how many companies, whom you would think would feel as though they were staring down the barrel of a gun in terms of the economy and the impact of competitors, still cannot get themselves to change their behavior.
Organizational change, building on successes, demonstrating them, finding out who's really on the team, and trying to move out the people who just can't get on the same page as everybody else — all these things will be critical success factors. There's no question about it; you're trying to change the focus of your problem solving — moving from thinking about your processes or your systems, to thinking about the decisions. Decisions are different. They change more often; they have more external constraints; they require a wide group of people if you're going to solve them well.
Let me throw in one more thing. I talked about the 'stick' — getting someone at the top; you need 'carrots' too. You need to find a way to make your partners look good. You need to go to their environment and brag to their boss about what they did on your joint project ... things like that.
You need to make the person who hired you money. You need to make them look good. That's a most important point.
I've got a two-part answer to this.
I think one enabler of this will be — again, spurred by open source technology — a unification of these technologies. You won't have a separate system for your processes and a separate system for your rules. You're going to have one piece of software to model it. That means you're not going to have the 'process guys' fighting with the 'rules guys' ... because they've got one tool, one way of modeling things. The tooling itself will help them understand when they apply it correctly.
If we look at Drools, Drools 4 already unifies processes and rules, and allows them to be leveraged much better than trying to use two different systems. We extended that for CEP as well, so what that means is you've got much more unified approach. It removes the barriers, the separation, between the different disciplines. You have a process-oriented approach, a rules-oriented approach ... and then people tend to take the two. There's a middle ground between the two. The right tooling, the right unification of the tooling and software, can help that.
The second part to that is ... you have a unification of tools, which is quite similar to rules and processes, and you have ones that aren't similar but can be leveraged together. There's a company here called Open Rules. They have probably one of the coolest things I saw here yesterday. They take standard decision tables, which is a rule metaphor for authoring, and they combine it with machine learning. This is something that's really enabled by open source ... the ability to take different technologies and have a very low barrier to entry. You just get them together and see what cool comes out.
I think open source will enable a lot more interesting ways to bring these things together. People are naturally more cooperative in moving things forward.
I have to say just a word about this. The first thing I would mention is ... a lot of times people don't even know they're not in the same sandbox. So, how do you find out you're not in the same sandbox? You examine the words you're using to talk to each other. What you need, as a part of this process, is to develop a business vocabulary — to ensure what the sandbox really means.
Secondly, if Gladys were in the room she would be saying/mouthing the word 'scope' to me — you have to define and manage within scope. That's always critical. You need a visible business sponsor. If you can't get that, there's probably going to be trouble.
Those are the main points that I would add.
|Kristen: I think we're running out of time. Before we wrap up, I want to ask the panel if there's any last message or comment they'd like to impart before we close.|
One thing is ... you want to swing for the fences, but you've gotta get that single also. Get a small pilot project to have something tangible inside your business, to show that it works and to build allies for doing more. So many moonshot programs have failed, after long expenditures. But if you can just get into the atmosphere first, and show the potential, that would be a big help.
This is the second time John and I are on exactly the same page ... because the very same thing is true in business rules. Define a problem that is of manageable scope, reasonably focused on a particular area, and show success. There's nothing that sells like success. Don't try to solve the world's problems — just solve a problem. That will help you go right along.
The thing I would close with is simply ... it's really exciting to be involved in this kind of thinking, this kind of activity at this point in time. If you could stand back and look at it, what's happening is quite exciting.
I'm sitting here on the panel and I made notes of two terms that I'm now going to have to look into. One was 'induction rules' (that John Elder used) — anything with 'rule' in it I have to go look at. I have not heard that term before. And Don, I think, said 'reflexive processing'. And (darn!) I don't know what that means exactly either. So I have homework to do.
Well, that's my goal — to use terms that sound really good that don't really have a meaning to them. <laughter>
I would say, first, agree on a goal that's larger than yourself. That doesn't mean you're trying to hit a homerun. But you are trying to achieve something that is outside of yourself — you're trying to think in a different way; you're trying to get a group of people to think in a different way.
What that introduces then is the need to embrace cultural evolution within your company. For a lot of companies it means you've got to get a culture to begin with. There are a lot of companies that are not purposeful and deliberate about their culture. That's a huge missed opportunity ... because when you see the really flashy stories about companies that are doing really exciting things, one of the things that almost all of those stories talk about is some real cultural thing that they do. They are deliberate about it.
If you're going to embrace this kind of change in the way your company behaves, you're going to have to be able to say, "We have a culture!" ... whatever that culture is. If it doesn't align with your values, it just isn't going to happen.
One of the interesting things about writing a book is that you discover these ideas that you thought you'd mention, as a kind of aside on one page, but then they turn into whole chapters. One of those was 'adaptive control'. First, it's a fairly technical topic in many respects. It plays to what has just been said. You're not going to get all of it done first off; you're not going to get it right the first time. So one of the things to invest in early is an ability to keep evaluating how you're doing ... to keep changing — to think about what it takes, in your organization and in your systems, to become more evolutionary. Because if you have a culture that really resists going back to systems and changing them or updating them ... if you have a mindset that "We'll hire a data modeler one day and then once the model is built, that's it; we're done!" Those things will get in the way, no matter what else you do. So thinking about how you can — organizationally and technically — become more adaptive is a really critical thing.
I'd like to thank our panelists for sharing their wisdom and experience with us. It's much appreciated. And thanks to all of you, as well. <applause>
# # #