U.S. ATLAS HL-LHC Upgrade Level 2 and Deputies-NSF only Management Mailing List

[Gustaaf Brooijmans <gusbroo AT nevis.columbia.edu>]

Hi,


> 1. Presenting BCP and Actuals tables in the L2 talks
> A very interesting cross-check on our BCP strategy is to compare costs 
> for BCP-1033 (excluding escalation) with the COVID-actuals we observed 
> in the period 01-Nov-20 - 30-Apr-21. If our BCPs are accurate, these two 
> numbers should agree reasonably well. I have compiled these comparisons 
> at L2 in the attached table. As you can see, the NSF totals agree within 
> 6%. However, individual L2 systems have much larger deviations.
> It might be useful for Gustaaf to show the rolled-up comparison in his 
> talk. However, given the spread in L2 results, I'd suggest that we *do 
> not* show the actuals tables in the L2 talks, but rather show only 
> BCP-1033 and BCP-1038 diff's for Direct, Fringe, Overhead, Escalation.
> Let me know what the rest of you think.

Thanks for this.


> 2. Harmonizing statements in variance reports with numbers in covid 
> BCPs/Actuals
> We discussed at length the recent example of inconsistencies between 
> reported covid actuals and statements in variance reports. The 
> conclusion that we reached is that these inconsistencies are inevitable 
> because of the way we report covid-related actuals. The problem is that 
> the COVID-related fraction for each month is entered in the accrual 
> sheet before the actual cost variance (CV = ACWP-BCWP) is known. Since 
> the COVID fraction is a rough estimate of the fraction of that month's 
> ACWP that is caused by COVID, it is very unlikely that it will turn out 
> to be exactly equal to the CV that's actually calculated once the 
> statusing is finished. Thus the numbers that we have will almost never 
> agree with our explanations in the variance reports.
> This problem could be fixed if instead of reporting the COVID fraction 
> as a fraction of the ACWP for that month, the anticipated fraction of 
> the eventual CV were reported instead. Thus we would enter, for example, 
> that 100% of the actual CV (whatever it turns out to be numerically) is 
> due to COVID. Then it would be easier for the variance report 
> explanations to reflect the actual variances. However, this method also 
> has flaws in that it's entirely top-down.
> In any case, we should bring up these two options to the NSF and ask 
> which they prefer. Neither is perfect but both have some attractive 
> features.

Yowza!  Please do not ever mention this idea any further!  Basically 
this says "we'll look at the numbers and then we'll write a story that 
fits."

For my example from LAr, the explanation is presumably that something 
else cost less.  There is for example the missing engineer at UT Austin, 
a change in personnel at Columbia.  The point is that the story has to 
be complete.  Not to invent one!

Best,

Gustaaf