On Fri, 2004-06-04 at 14:26, Paul Gilbert wrote: > Marc Schwartz wrote: snip
> > > >I agree that quality and value are important, but I think that the issue > >of cost should not be discounted out of hand. Value (for both company > >and client) is directly tied to cost. > > > > > [snip ...] > > Marc > > I agree with this, and most of what you say. Cost is important in both > large and small companies, and also in government. The point is really > that total cost of ownership is a very complicate thing, and you should > not get into it without the specifics of a particular company and > situation in mind. For example, if the end users takes responsibility > for all the support, the cost implications will be very different from > the situation where the IT department needs to guarantee availability. > Even in your own company you may have a very different attitude with > respect to your research software and your accounts receivable software. > People in finance making real time market decisions will have a very > different cost structure from academics in finance. Paul, Sorry for the delay in my reply. A sudden request came from a client this afternoon and I just finished the analysis. I am in agreement with you that situational differences will bias the focus on costs and perhaps even the ability to define them well. Within the timeframe that Kevin and/or his associate has for this presentation, this topic cannot be adequately covered. That does not mean that you cannot raise it for further consideration by the audience within the context of pointing out R's strengths. I suspect that if you point out these issues, somebody with the right insight will have a light bulb go on relative to the possibility of cost savings versus their current set of tools. They will of course need to pursue that line of thinking outside of the scope of this presentation, which is fine. I do use a commercial product for accounting (which is the only reason that I still dual-boot Windows and Fedora Core 2.) My alternative is to send all of my paperwork to my accountant to do all of my ledger entries, which at $300 U.S. per hour is not inconsequential. So, yes, I make a cost based decision to purchase a commercial OTS product that enables me to do the grunt work and send an electronic file to my accountant for review. My cost per unit of time is cheaper than my accountant's. If I could do the same thing with an open source product, hallelujah. It would save me even more. Unfortunately, as is oft discussed, this is one area in which the Linux world is still lacking. I suspect that will change in time however. On the other hand, I use a payroll services company to handle that part of the business. Their costs to run the payroll, pay taxes, worker's comp and all the rest of the associated procedures are cheaper than what I could do on my own. It is not that I could not do it technically, but that use of my time would in the long run cost me more money than what I pay for the service. Each component of the process does need to be evaluated within the context of the alternatives and appropriate risk/benefit considerations. > The point was really that many people are very sensitive to arguments > about cost, and often have positions that they feel obliged to promote. > So, as soon as you mention cost you are likely to get into a very long > discussion that will not be fruitful unless you are prepared to talk > about very particular situations. For this particular audience I think > it would probably be more to the point to describe how good and reliable > R is. A particular company may decided R is just too expensive. For > example, some companies in finance are worried about real time decision > making. They may have to hire 10 more IT staff to guarantee 24/7 > availaility with no more than 5 minutes outage per week whereas, with > commercial software, they may be able to buy guarantees. (This is not a > statement about commercial software being more reliable, it is a > statement about being able to buy insurance.) But, as you say, for must > of us R is a real bargain. There is no doubt that folks are willing to spend more in some cases for a piece of paper that guarantees availability and perhaps some form of compensation for downtime, which in these situations will likely result in lost revenue. As long as clients are willing to pay for those features and/or companies determine that business imperatives warrant such expenditures... I think that your closing point relative to R's reliability is important. This goes to the old saying "Facts are negotiable, perception is reality." Why would R be more or less available or reliable than another analytic tool? More often than not, I suspect that it is not the application but the underlying infrastructure that results in reduced availability. The perception issue may be the biggest hurdle that the open source world needs to (and will) overcome in the commercial marketplace. Anyway, I think that we are more in agreement than disagreement here. I hope that Kevin has benefited from the discourse in some fashion. Best regards, Marc ______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
