Y2K Failure and Spending Projections

I took a chart from the GartnerGroup reports (March 1999 Report to Senate, October 1998 Report) that shows their estimation of probable Y2K failures over time -ramping up to a peak in early 2000 and trailing out for several years.  I wanted to have a way to build this chart based on the independent probabilities of different kinds of failures so these could be checked and compared.

My version of the original chart, adjusted to be a probability distribution - to extend across relevant years, and to, as well as possible, meet their constraints that they expect 55% of the failures to come in 2000, 25% before and 15% after.  Their curve actually does not do this, but I adjusted this as close as possible.

Then I took the independent probabilities for different failures (state fiscal year, federal fiscal year, general year 2000 system failures peaked in early 2000, embedded chip failures, embedded system failures) and mapped them individually.

And recombined them to get this curve, which is similar to the first.  The idea I have is that these curves might be used to represent the probability of having to spend money on Y2K failures.  Both the preparations and failure costs are included.

It is interesting to compare these large Y2K system failures curve with charts from an analysis of the Gartner status information.  I put a normal curve on each of their ranges - interpreting the bars as 90% confidence intervals.  To be able to add the information from different countries or sectors, I used (for the country data) several different weightings.  In this one, I used GNP weights - so that the curve is a total of the individual country curves each curve integrating to be equal to the GNP.

Then I used a report that Rubin (Dr. Howard Rubin, Chair of the Computer Science Department at Hunter College and Research Fellow for the Meta Group. The Robbins/Rubin Y2K Schedule Indicator, from the International Part of Senate Status Report) produced which narrows in on Y2K spending more particularly.  In this way, I can combine the status information to get a reasonable picture of the course of Y2K spending over the next two years.  One things I have assumed for these last two charts is that it really does take 30 months for larger organizations.  I have yet to correct for non-standard distributions and for sectors or countries who take less time.

It looks like a lot of organizations will not make it on time.

This next Chart has a lot of assumptions.  It is best viewed as a race of US Firms from the left to the right on the chart.  The "Completion Date" is the date an organization fully completes its Y2K project (in Gartner terms, it reaches 6.0)    I guessed at the numbers of firms and the project costs.  My plan was to set up the form and ask for help in refining the numbers.  Right now, I have assumed :

Further, I am using the Gartner phases above as 90% confidence intervals for large and medium, but 80% confidence for Small and Sole Proprietorships.  The units on the Y-axis are millions of dollars per day.  This would represent preparative spending and not spending on Y2K repairs after breakage. The sole proprietors do not start until 1999-06-01.  The curves each add to the total cost.  What I need is Gartner phases and Y2K spending by size of firm to do this right.

Note that if GNP is used for weighting, then the smaller companies will be more important.   "Sole proprietor" probably has some fairly large organizations.  Perhaps someone has GDP produced by size of firm.  Or Y2K project length by size of firm.