Accelerating Time to Solution and Project Risk Reduction with Data Virtualization

Michael Linhares
I wanted to also thank Bob and Jim and the Composite team for inviting me to speak today.  I do enjoy doing this, and I’ve also got the opportunity to speak about something completely different.  So this is something new for me to talk about, things that not necessarily I would have talked about before, technical things, implementation.  So I’m going to kind of do a mix of stuff, and kind of talking a little bit about how to bring other ideas of process, improvement, so continuous improvement techniques into your agile or iterative way of developing your software and your implementation of doing data integration. Maybe using a little bit of innovation and introducing a culture of innovation into what you do and how you do it, and thinking about that a little bit while reducing the risk of deployment; decreasing the time that it actually takes to deploy things, and decreasing the cost. So kind of putting these all together, thinking about how to mitigate risk and being honest with yourself and your customer about what the biggest risks are in the projects that you’re running, and how you’re running them. So kind of bringing a whole bunch of ideas together, and I’ll try to put it in the context of some of the projects that we worked on and that we’re working on now, actively involved with, which is a pretty big one and pretty challenging.

So I’ll give a little bit of perspective of myself, where I sit in the organization, and a little bit of my history. I’m at Pfizer, it’s a pharmaceutical research and development company, and I work in the R & D space, so I actually work in pharmaceutical sciences, and we’re the part of the organization that actually manufactures the drug, forms it into a formulation, a tablet or a liquid, and then delivers it to all the clinical supplies for studies, and then actually transfer the technology to our manufacturing, eventually. So I’ve been a Composite user since 2005, and my first eight years at Pfizer were doing science, pure science, and then I switched to doing information architecture, and I’ve pretty much been doing that ever since, in a couple of different roles. Right now I lead a business information service team, and we really focus on portfolio planning, investment decision making around a pharm/sci portfolio, and research utilization.  I also own the strategy for pharm/sci, so we’re probably, if you think about it, a mid-size company.  We have a budget of about $700 million.

So in the pharmaceutical process, which is pretty long actually, we get involved in what we call exploratory development, where we’ve identified a compound that we believe will be a good drug for a specific target, and we start making material to be able to go to the clinic and be able to do safety studies.  If everything works out well, we continue to go on to full development and do large clinical trials.  We develop the actual material for doing the transfer to the commercial organization, then actually interface with the commercial organization for doing that.  So one of the things that is important to understand is that this process is about ten/fifteen years long, so it takes quite a long time. We’re involved for a long period of time, but we also interface with research on one side; so we have one set of customers of research on the left hand, which has a set of data and a kind of philosophy of how to do things, and on the other hand we have our commercial customers, and our PGM manufacturing, which is Pfizer Global manufacturing, and they have a certain perspective on what information is and now to operate, and they’re quite different. So we get to interface with both.

So one of the things that I’ve been trying to do for at least the last five years, if not before that, was try to articulate to our organization and our customers how to get more value out of the information that we have.  So originally, when I was a scientist, I would try to convince people that really what you need to do is publish your information.  So as a scientist, you understand the concept of publishing papers, you understand the concept of actually publishing your data so that other people can use it, and use it in different ways than you ever imagined.  So instead of everybody sticking with Excel, and I know everybody loves doing Excel and everybody’s business is pretty much run off of Excel, and transfer all this information around in spreadsheet format, really trying to get people to think about things in much more online, integrated ways so that everyone has access to the same information at the same time, no matter which tool you want.  So this is a really important kind of philosophy piece where I want to provide information to you and everybody else, and you can use any tool you want to look at it.  You can use any extraction tool, any analysis tool, I really don’t care, but I want to give everybody the same data, and everybody to have access to it all the time.  So I think that’s kind of a philosophy thing that I’ve been pushing for.

So I wanted to talk a little bit about innovation and value and data quality. I know data quality is going to come up a lot, and I kind of want to put it in context of things, so I’m going to talk about a New York innovation, which is a hamburger. 

So in the 1800’s a lot of Germans were working in New York City as dock hands, and the street vendors came up with an innovation to feed them for lunch, and that was the hamburger.  Okay?  So how is this relevant to what we’re doing?  I kind of want to put it into the context of a challenge for what I’m going to call information brokers. So I think Ted called us data people, well, I’m going to call us all information brokers.  Really, we’re moving information around, we’re making it available to other people, and then I’m going to look at the relationship between the cost of something and the quality of it. So we’re going to look at burgers, and the cost and quality of burgers in Manhattan, to start with.

So we can go down the street and go to Burger King, and you can probably get a Jr. Whopper for about a buck 29, and you’re going to have a certain amount of quality to that.  It’s processed beef that’s probably been shipped in, they flipped it on the grill, and you get pretty much what you expect out of a Jr. Whopper. You could also go down to Blue Smoke, which is downtown a little bit further, and for about $12 you can get a burger which is hand ground, prime beef, really good, great service, but it costs a lot more, right? Or at least an order of magnitude more. You could also go down to the Wall Street Burger Shoppe and get their most expensive burger, and it’s actually about $175, and it’s made out of Kobe beef, it has black truffles and gold leaf on it, and you know, the quality is probably significantly better. So you actually look at this, and you say “Okay, so I’m actually paying more for better quality.” Right? So now if you think about pharmaceuticals, and you go to the doctor, and the doctor gives you a scrip, you go to the pharmacy to fill the scrip, and he gives you something from a branded pharmaceutical. It’s expensive, but your insurance is hopefully going to pay for it, but one of the things in a branded pharmaceutical is you expect it to be 100% in terms of the drug that’s there.  At least plus or minus a certain percentage, and it’s turns out to be it’s 100%, plus or minus 5%, so that’s what the FDA requires the standard to be. So then if you actually go and you say “No, I don’t want the name brand, I want a branded generic.” It’s a lot less money, it’s actually cheaper, probably by an order of magnitude, however the quality is the same.  In order to get that branded generic, the company had to prove that it had the same bioequivalence as the original one, and its potency had to be 100%, plus or minus 5%. Now you could also go to Wal-Mart and get the $3 version, and again your quality is exactly the same. 

So why I bring this up is that I believe that as information delivery, as information brokers, our situation is more like the drugs then the burgers. The expectation of us, in delivering the data is that it has high quality, and that quality is defined based on the business case. Each business is going to define their quality requirements a little bit different, whether you’re in banking or whether you’re in pharmaceuticals, if you’re in research, or you’re in manufacturing.  Everybody is going to have a little bit different twist on what their measure of quality is, but really the expectation is that you can’t relate cost to quality.  I can’t go to my customer and say “you want good quality?  Well, that’s going to cost a million bucks.” I can’t basically overcharge for quality when I’m doing data integration. So I’m going to do a little build here. 

How do we speed up our projects while maintaining this really high quality?  If you look at Lean Six Sigma approaches, they have this kind of triangle of cost, time and quality.  Lean Six Sigma in manufacturing, you’re looking at processes and deviations of processes, and patterns to try to reduce and avoid and remove processes that don’t add value.  You’re trying to look at how to improve quality measures, and a lot of it has to do with physics or engineering types of aspects, and they really try to decrease both time and quality to save money and get better products. What I’m going to suggest is, you actually turn this on its head, and we put quality at the top, and it always stays high.  We’re not going to change quality, we’re not going to look at trying to change or effect quality, it’s always going to be what is expected, and be high.

So what we’re really going to focus on is how do we decrease the cost of implementing solutions and decrease the time of implementing solutions?  So I’m going to try and squeeze this triangle by doing different approaches.  By taking a different approach to actually doing the project and using different kind of philosophies and applying things to the project, so that I can actually keep my quality high, decrease cost and decrease the time.

So I’m going to go back a few years, when I was a data architect and this was before virtualization; so this is 2003 we had to build a research information factory.  So we had to – everybody wanted to put all the information in research into one location.  We wanted “I can see all my data in one location” situation, right?  So our center of excellence and our toolbox was pretty limited at the time.  We were an ETL shop, and we had the Inmon philosophy of data warehousing with a little bit of Kimball marketing involved.  So we’re going to do everything.  We’re going to take the data, and we’re going to move it into a persistent stage, we’re going to move it into an ODS, we’re going to copy it into a warehouse, we’re going to copy it also to mart that’s a star schema. So we’re going to copy the data once, just straight copy it; we’re going to transform it once into the third normal form, and we’re going to transform it again into a fact based star schema for access.  We’re going to do this for all the data, right?  So when we started doing this, and for example in one case, we have in research high thru-bit screening data, which we take our entire sample library, of all active compounds, and we screen it against a new target, and we produce a lot of data with this.  So we had about a billion records to move, so we spent all of our time optimizing our ETL processes to move – I think we were moving about 20 million records a day, and we would move it to stage, then we’d move it to the ODS and then we’d move it to the mart, and we were just starting to think “why are we doing this?  Why are we spending so much time and money moving this data three or four times?  What’s the value in this?  How is this actually going to really augment the business and help the business solve their problems?  If we just moved it once, then they could query that and it wouldn’t impact the transactional system where they were collecting it in the first place.”

So what’s the real value of this?  As we were doing this, we started evolving this thinking about why are we doing this?  And then we started coming to the point of “Okay, we want a new mart.” And we’d go to the shared services team, and they’d say “well, okay, that’s going to cost – it’s going to take us about a year and cost about a million bucks.” So we’re kind of in this $1 million/one year time frame situation, and we’re going “Okay, that’s not going to work for us anymore.  We need something different. We need something new.  Where are we going to go from that?” So I kind of – after transferring this to a shared service group, and they’ve been supporting it since – it’s our research and information factory at Pfizer, and it works quite well, but it now is – in my current role, would I ever want to do that again? Would I ever want to spend all that money and time, this took us about three and a half years to build, and three to five million dollars, would I want to do that again? 

So I’m going to say “No, I won’t do that again.”  I’m going to go and do a continuous improvement approach on doing things.  So as I said before, I’m now going to have my quality up high, and I’m going to start shrinking my cost and time.

So I’m going to start thinking about different things that I can do, so one of the first questions I’m asking is why move all the data? In this case, I’ve got a billion records, I’ve got a whole bunch of things, I actually need to move it to be able to optimize queries.  I need to actually move it out of the transactional systems.  Now, I was lucky in research, in that I was able to, as part of that project the first thing that we did was se what all the global reference data was, and told all the delivering applications they had to use the global reference data to deliver their data.  So I had a very good standardized set of reference data that I was already using.  So everybody had the same list of projects, everybody had the same list of compounds; everybody used the same primary keys.  It was almost an ideal situation. So what we were really now thinking about is “as I move forward, how to I keep replicating that ideal situation, and reduce the amount of times that I move the data?  If I move it once to kind of stage it, and get it out of the transaction system, then can I put a different type of mart layer on top of it?  A different type of technology to use?”

The other thing is in the past, in that past situation, everybody has Oracle, right?  So when you look at that situation and you say “I have everything in Oracle.  Do I need any other technology than Oracle to build views?”  It’s got materialized views, and no, I don’t.  In that situation, that works perfectly fine and I’m good. I can stick with my technology, but it’s when you start bringing in things and we do another – we bought Pharmacia, Upjohn, Searle Company, and we started seeing non-Oracle sources. We started seeing XML sources; we started seeing people wanting to add things from Excel.  Then we buy Wyeth, and now we’re seeing SAP coming along.  I think there was also a situation in Wyeth where you have – we have one team with the secret list of projects that we’re going to transfer from Wyeth into Pfizer and we’ve got another team that has a spreadsheet with the ones that are important and attributes about them, how do I put that data together now, really fast? And they want it yesterday, kind of situation. They want to have a dashboard with statuses on it, so I need new technologies to do that.

So what am I going to do?  So I come up with this little framework, and this is a modified framework of what we’ve done in the past. Really what it is, if you look at it from the standpoint, I think everyone who does virtual data integration or data integration has a similar framework, a different way of drawing the picture.  But we’re basically using Composite to help us integrate the data, this isn’t not necessarily just a pure virtual world, we are using a hybrid world where we are, actually having physical data stores and physical databases.  We have multiple sources; I just have some of the sources here.  Some of the sources happen to be Composite sources, other sources are databases, anything ranging from SQL to Excel to XML, and then we have some data entry tools that we’ve written to actually – it’s more about annotating the information.  So we have a list of projects and we want to annotate them and add more information, but then we have one single source of data.  So we have people called it OneSource, single source, you know, and all these different names, but I think we’re all trying to achieve the same thing. So we have a single source of the data to use, but then I’m going to allow people to use whatever tool they want on top of that. If you want to use BusinessObjects, fantastic, if you want to use Spotfire, great.  If you want to write you own Java or .net applications to sit on top of it, fantastic. I’m happy with all of them, it’s a good solution, and it works for everybody.

Now, the other thing about this is, what’s nice about this is, it’s reusable.  So I can keep reusing it for different purposes. So when the Wyeth integration comes along, I just grab the data from two or three sources that I want to do, pop them together, give my client and my management the dashboard they wanted, and this is like a two to three day process, versus a two to three week process or any other timeframe that I might have to do.  So I can do this very fast and very agile.

So, if I think about then, adding on what I’ll call a culture innovation, and really this is some – as being a scientist and thinking in a very scientific method way over the years, and kind of just the way that my thought process works over being an engineer, for example, is that thinking about what is innovation and how does innovation apply, so Jay Paap at MIT did some really nice work with his group there, and published some papers on disruptive innovation and these kind of ideas of where really thinking about what is innovation and how to apply it.  So if you come up with an idea, a need, a business need and you come up with technologies to solve that business need, that’s a good idea.  But if you can actually implement that and then reuse that idea multiple times and then actually get value out of it, now you have some innovation.  I think it’s that difference between being able to reuse and get value out of something.  So some of the things that I look at in doing virtualization and doing integration is that really you have to think about designing for reuse, and not necessarily always jumping at the one technical solution for solving a problem.  I think a lot of us, sometimes we’ll have a customer come to us for a need, and we’ll go “Oh, I know how to solve that.” And dive right into it and just start solving the problem without taking a step back and thinking about it a little bit, and actually abstracting up. So you know, if you ever go to an architecture conference, hear John Zachman talk about the Zachman framework, there’s seven levels of abstraction, really abstracting out at least one or two levels if hugely valuable to understand where you are and where you sit within the solving the problem, and are you trying to solve a problem that somebody else already has?  So I think there’s a few other things here that are really important, and it’s a rigorous evaluation of what your customer needs are.  So I think a lot of time as information delivery or information brokers, we’re asked for something and we go “okay, we’ll do it”, instead of asking “why do you need this, and what is your business purpose?”  I’ll give you some of the questions you can ask to try and flush out what is the person really looking for, and how it should be applied.

I think it’s also fair to look at all the different options. Virtualization is a good solution, ETL’s a good solution sometimes, and there’s a lot of different options that you can throw at things, so always keep that tool box that you have open, and really look at the different options that you have, and don’t just stick to one – if somebody comes and says, and it happens all the time to us, where a business customer comes and says “I want to do this.” And they basically in their statement is the tool, or is the technology, and sometimes I think we have to take a step back and say “okay, maybe what they want is – they know what they want but they don’t necessarily know that that’s the right tool.” They’re just trying to give you a hint, or that’s the tool they know. I think also coming up with really realistic solutions, and this has to do a lot with the risk, so if you think about the risks that you’re going to take in deploying a new technology to an organization or a technology that is unfamiliar or is introducing change to the organization, understand where you may fail, and understand to not bet the whole farm on something.  So kind of take these concepts of Wharton school has a great class around failure and the analogy kayaking.  Know when to roll over, know what happens when you roll over, and if you do roll over, know how to get out or flip back, right?  So it’s really understanding a lot around your potential for failure, when you’re going to fail, and what your most likely point of failure is. If you can understand that for yourself and do a little bit of risk management around it, it’ll get you to be much more successful.

So last spring, we started thinking about something we called the Scientific Workbench, in pharmaceutical sciences, and this was taking a large set of our data that included our scientists work in a laboratory notebook, which is all XML based, and they have laboratory information management systems, which is usually called LIMS, data capture systems from all their instruments, we have a portfolio database, we have our supply chain management databases, we have our research information factory that has some of that research data that I showed before.  They want to put it all together and they want to be able to do predictions and simulations off of it, they want to be able to really ask a question like “for this compound, has anybody worked on it?  What were the external parameters that they did on it?”  And they want to find all that.  So right now, there’s no integration between these, and we started looking at this and said “Okay, here’s one way of doing it.”  We started playing around with Composite’s discovery tool, started discovering the relationships between the databases, we started modeling some of it in studio, and we came to a point where we said “Okay, stop. We’ve got serious data quality issues, actually.” And we actually needed to stop and kind of take a couple of steps back. And this is not a technical issue, it’s a business issue, where we actually have – in each one of these systems, each one of these systems was developed kind of in a silo project, developing their system, they knew what a compound was, in their perspective. They knew what a batch was, in their perspective.  They knew what a project was, in their perspective, but they didn’t look at it across the other systems, so now we’ve had to kind of take a step back and really start asking questions about the situation that we’re in, so we want to integrate this, we have the tools to integrate it.

We have the tools to access it, but our data actually in the systems is different enough that we can’t. And we actually need to go through a process to fix it, so we’re starting to ask these questions. When does it happen?  Where does it happen, around the data? What’s the business significance?  Say, for example, a compound number; some people are putting commas in them, some people are not.  Some people are putting dashes, some people are not.  Some people are dropping leading zeroes, and it’s very frustrating for the data integrator to look and say “I should be able to join a compound number across all these systems, and I can’t.  I should be able to search on a compound number, and I can’t.”

So what we’re really trying to do now is we’ve taken a couple of steps back, and we actually are going to the business and asking them, with the help of some consultants, to sit behind them, to get those SMEs in the business to actually start looking at the information that they have.  How they put it in, where they need the information from. I think Ted brought this up this morning, it was kind of coincidental, but it is a business problem, and we’ve basically said we have a solution – a technical solution to a problem, but we can’t apply it because of the data problem that we have and the quality problem. 

So now what we’ve d one is we’ve kind of backed off, right?  So we’ve got all these systems on the bottom, we came up with this common BusinessObject model, we’re calling it, so okay, it’s a common model and we have all these different ideas that we want, but now we’re actually taking that step back and looking at the data issues that we have, looking at the processes of how the data flows in, where the data comes from, and trying to come up with a set of data standards for our organization. Then we can collaborate with our research group on one side, and our manufacturing group on the other side, and satisfy both.

One of the nice things is that we already have these very large physical additions of the rif, and development information factory built in kind of the same format, and they have all the data that we need to collaborate with, and we’re just putting some virtualization right on top of it.  So we basically built a nice virtual situation that we now can access through this one tool that we decided to go with through our research scientists, called Tripos D360, which is kind of beside the point.  But one of the things we also realized was each one of these systems contains a lot of information, and actually by going through this process and asking these questions again and again, we found out that actually the scientist only want about 10% of the data that’s in all these systems. But they want the possibility to access the entire set of data at some point, if needed.  But that’s a second tier query, so it’s not a primary query in going in and saying “I want to search everything that’s in every one of those particular systems.” So we’re trying to think about ways and solutions to providing the scientists to get back at all the data in the system, but not necessarily replicating or presenting all the data in the first layer of the queries; only that 10% of the information that they are really asking for.  Then through a process of iteration, if they ask for new data, we’ll be able to provide it to them then very quickly.

So what are some of the lessons learned that we’ve had?  I think I’m just going to build this out. I think really for us, over the last five years, and this might even be speaking more toward all of us at Pfizer, in the research and development, is really looking at and abstracting the requests that you get at least one or two levels up to understand what the big picture is, and I think there was a point made earlier today about communication.  I think once you do that and you communicate with architects and other information architects in your organization, and if you have a center of excellence, communicating with other people about what problem you are trying to solve, asking them if they’ve tried to solve it before, or if they have already solved it, you can reduce the amount of work you’re doing, significantly.  I think also challenging processes that aren’t value add is very useful.  So we ended up at one point, having a process for promoting BusinessObject processes that took about three weeks, and we challenged the process with our shared services organization, and we asked them, we got the right people in the room, it took only an hour to get all the right people in the room, and everybody agreed that we had overemphasized the risk of promoting a universe and reports in BusinessObjects. We know have a process that takes one day. So you know, really by challenging some of the processes that you might have and really thinking about the risk that people are assuming about certain processes that you do; you can decrease the time to delivery.

I think it’s been said over and over again, focusing on data quality and consistency from the start, so really just solving that data quality problem or assuming right up front that you know what the data quality is that you have, facing it before you do anything else, that’s going to solve a lot of your problems.  Then I think, three more, which is – I think a general philosophy is that a broad mix of technology is good for information sharing.  I think mixing this whole virtual/physical world is the ideal situation, and we’re kind of all evolving to there and all figuring it out depending on what our data size is, what our latency is, what our requirements are; and we’re kind of figuring out where virtualization works.  We’ve had problems, for example, with the consistency of availability of source databases. So if you’re in a virtualized world, you have to assume that the source database will be available all the time.  So in certain circumstances, we’ve had to really go to caching, because we didn’t trust that the database would be there all the time.

I think the other thing is that from my perspective as a person now – actually as an architect, and information architect or a broker sitting within the business and doing things, I think actually having open access to the Composite platform and promoting open access, allowing more people to use it; don’t keep it in the back office as an infrastructure tool that only the data warehousing VI team gets access to.  So you know, let more people use the tool.  Let more people understand the value of the tool I think what’s happened, I see this with my kids, and I see it with the young new hires that we have, so many people are now very, very technology savvy.  They are very comfortable with software, and if you actually provide them with a little bit of guidance and a little bit of help, and you give them a piece of software, you’d be amazed at what they can do.  I think that locking it in the backroom is not a good thing. So that’s my last plug for open sharing.  Don’t just share the data, share the tools also. And that’s all I had to say today, so any questions I’d be glad to answer.

Audience member
Your last question, or last statement about getting it out to people, were you talking about getting the studio out or just maybe the Composite information server and the access to it?

Michael Linhares
So I personally believe that you should give the right people Composite Studio.  Yes, you should give them the capability of investigating and building views, not in production, not so they can proliferate things that can’t be maintained, but give them the ability.  Even discovery, give them the ability to look for things.  Where they believe that they have a solution, and they have some time and they have the knowledge of the information and they know what they’re looking for, sometimes it’s actually more efficient to let them have it.  Now, if they try it and do it and say “Hey, Mike, I got this view and here’s how I went, can you get that into production for me?  Can you get that so that BusinessObjects can see it?”  Okay, now we’re – you know –- can you optimize it a little bit?  Make sure it’s good?

Audience member
That’s a topic of concern.  You can bring a lot of things together and you can do a lot of things incorrectly.

Michael Linhares
Oh, absolutely, yeah.  We don’t want that. We have a question back here?

Audience member
Right, my question is you talked about portfolio management related to use case or business scenario here, what kind of data volumes were you dealing with?

Michael Linhares
Our biggest tables are about a million rows, so we’re looking at say, for example, our entire portfolio for projects and tasks related to those projects, they’re about a million rows.

Audience member
I’m intrigued by your statement that as an information architect you’re kind of embedded with the business.  I would consider that maybe information architecture is a technical discipline, not a business, so how did that evolve and how does it work?

Michael Linhares
That’s an interesting question.  At Pfizer, we do have a central architecture function, but one of the things that was noticed by the businesses was forming small teams within the business; and I have a team – there’s only three of us—is that you can basically take those connections and with a partnership with the informatics organization, we call it now business technology, BT, we partner with them closely.  So we use all the shared services, so our Composite servers, our BusinessObject servers, everything like that are all basically shared services by our BT organization. But we look at it from the standpoint of solving the problem for the business in a very agile quick way.  Using the tools that BT provides us, we actually come up with the solution, and we do it in partnership with them.  What we’re looking for eventually is, and we’re going to just start working towards it, because once you put things into production that work well and that are sustained – and what we mean by that is they’re actually being used.  So we get asked a lot of times to build stuff; we build it, we put it into production, and then when we’re monitoring it, nobody is using it.  Even though they asked for it, and things like that, so we deprecate those.  We go back to the business and say “You asked for this, and you’re not using it, so does it need to be there anymore?” we’ll get rid of it.  So what we’re doing now is we’re going to go start having those conversations to actually have shared services now support these applications moving forward, so that we don’t have to do it.  So our team is there to kind of build things quickly, and respond to the business needs very fast. The other interesting thing is, in our business, because we have our BT colleagues who support pharmaceutical sciences focus a lot on the supply chain; they focus a lot on the GNP and regulated areas of delivering drug to the clinic.  When anything – all the priorities are there.  So when you go and ask a question about “could you build this little portfolio thing” they’re like “we don’t have time for that.  We’re doing this other stuff.”  So that tends to be the problem too.  So small agile teams in the business, we now have them kind of proliferated throughout our businesses who are doing it that way, and it seems to be pretty successful.

Audience member
I’m going to hop on the last point once again, around giving access of the Composite Studio to others. I run the (inaudible 0:38:24) and we have precisely done that, unknowingly, I think it happened unknowingly, and then to the goodness of the tool that it’s so easy to use with a fairly limited amount of knowledge, but what we have seen is our group with shared services organization and we do more than just virtualization of service, so we have our entire created aspects of all of those things, so my expectation was that if you use the tool it will defuse socialized much more faster, which it did, but I don’t get back what views, what data services are available out there, it got pretty out of control, so now while the tool is socialized it has delivered its promise of accelerating or reusing the model detail.  The side effect is just becoming (inaudible 0:39:25) when there’s problem in the production, we become the first layer of getting slapped, even when the data is behind the view might be down, so that’s one side effect negative.  We’re not getting any positives back, in accounts of the enhanced knowledge about what data views are getting created, so that we can mix it with other   processing capabilities and bring newer kind of applications—

Michael Linhares
Yeah, so I think one of the things is that it’s kind of like friends, right?  So if you’re friends with somebody and you invite them over for dinner and you do things, at some point the other person can invite you over for dinner, right?  So when you give somebody a tool and you give them access to it, and you let them kind of do things, you expect them to come back to you and say “hey, here’s what I did, here’s what I’m using.”  It’s that communication going back and forth.  What we see, with scientists and researchers is, they will go find the tool on their own, or they’ll go find their own tool, on their own, and then all of a sudden there’s this tool growing out of this little place, so it’s not being managed by anybody.  They downloaded it, they bought it, now all of a sudden you have this licensing issue because they all want to use Sidebinder or – what was it, Pipeline Pilot – we had this one tool called Pipeline Pilot, they all started using it, creating massive processes, and then all of a sudden the licensing got all out of control.  So you have to kind of have a good balance, and I think if you have a defined set of tools that are supportable and maintainable and you provide people to use them and keep that communication going, keep your friends in your inner circle, then you find out what they’re doing.  You can also, the beauty of Composite it you can monitor everything that everybody’s doing.  Well, I’m totally against chargeback models, so I – but I think there’s got to be some kind of balance there.

Audience member
To these points that we just discussed, what is Pfizer doing towards setting up some other mechanism other than relying on friendships and inviting back and forth, a meta data management centralized repository process, so that you don’t rely on people sharing, but putting things out there so people are aware of what’s out there and what can be reused.

Michael Linhares
Yeah, so we’ve tried doing master data management program, and we’ve tried a meta data management program, and to be honest, we haven’t been that successful with it.  It’s always been kind of like that huge elephant, right, or we’ve always tried to boil the ocean.  I don’t know what all the analogies are, but it’s always been too big to get our hands around.  I think what we do have, is we do have a core set of architects that are documenting and sharing kind of these centralized models and some of that information, so you can— Pfizer is a – and this is a quote from a new Wyeth colleague, was Pfizer’s a you need to know company, it’s a who you know company, it’s a very people to people communication company, where some other companies tend to be more put everything in search and find it.  So it just has to do with our culture.