RE: [XP-NewYorkCity] Agile Techniques and Methods, 'Risk Management' and Metrics
- Ken,Thank you so much. This is excellent feedback. I really respect the work you have done very much with SCRUM.You raise very valid points that are at the heart of what I am struggling with. How can these be adapted (or not) to fit the agile world.The study results (Appendix B) of the 'Five Core Metrics' book is interesting, but these are mostly non-agile projects I would assume as they are from 1983-2000 (6,300 projects are involved).Mr. Cockburn has added to this discussion as has Mr. Jeffries (both who I admire greatly).Yeah the 'Waltzing with Bears' book got a few heated reviews, especially from a certain person from Cleveland. Ouch. I'd like to talk to him. He is obviously a risk management expert of some sort, probably with a strong academic background.Kind Regards,Damon Carr, CTO
STATEMENT OF CONFIDENTIALITY
The format and content of this material and its associated intellectual property rights are the exclusive property of Monetaire, Incorporated. All rights are reserved. No part of this material, nor summaries or derivative works, may be used in any way without the prior written authorization of Monetaire, Incorporated.
This electronic message may contain privileged or confidential information. If you are not the intended recipient of this e-mail, please delete it from your system and advise the sender.
Copyright 2004 by Monetaire, Incorporated.
From: Ken Schwaber [mailto:ken.schwaber@...]
Sent: Sunday, March 14, 2004 3:52 PM
To: email@example.com; Damon Carr
Subject: RE: [XP-NewYorkCity] Agile Techniquies and Methods, 'Risk Mangagement' and MetricsDamonI've forwarded this message to the best group of people I know who can give you feedback.The problem I have with statistical prediction of Agile processes is that, as a prerequisite to Agile, I've pretty much accepted that day to day attention is needed because the development projects in question are so complex as to not be amenable to statistical analysis. That is, the subject not only changes unpredictably, but the unpredictable changes have no pattern and are unpredictable. The "Waltzing with Bears" got a really pointed review on Amazon about its inappropriate use of statistical techniques. I also remember a discussion in an article on quantum mechanics that referred to stastics as the reading glasses for someone imbued in deterministic, Newtonian physics.I hope the group gives you valuable feedback.Ken Schwaber-----Original Message-----
From: Damon Carr [mailto:dcarr@...]
Sent: Sunday, March 14, 2004 3:36 PM
Subject: [XP-NewYorkCity] Agile Techniquies and Methods, 'Risk Mangagement' and Metrics
There are 2 books that were recently released that have influenced my
thinking around risk and metrics:
'Waltzing with Bears' - DeMarco and Lister and 'Five Core Metrics' -
Putnam and Myers, both up for a Jolt award this year.
I am an Agile advocate and user, having used XP with success both
professionally and academically.
I'm interested in this group's opinion on how these often
quantitative statistical techniques and metrics can be optimally
incorporated into an Agile process. Alternatively, should they be
included at all? For example, the DeMarco book has a probability
Distribution curve for confidence in a release date (which for us is
an iteration date of course) that looks 'lognormal' (not a
normal `bell curve') with a high steep peak to the left and a
tail to the right. For non-statistic types it is the area under the
curve (x-axis is date) that determines the confidence you have in the
date. You find the date on the X-Axis you are targeting and draw a
vertical line up to intersect the curve. The area under the curve to
the left of the line is your 'confidence' (see 'Quantifying
Uncertainty' CH8 in the DeMarco book). The sum of the entire curve is
always 1 – 100%.
You might think 'just use the date on the far right'. It is true that
would theoretically provide a 100% confidence but the date is always
so far off from where a customer wants to be. Often by 100-200% (if
not more). We are almost always forced to accept a date that is of a
far lower confidence (using this technique, which is only one of
many). For example, the customer may say a 70% confidence is the
standard. If we are forced into more aggressive dates we always make
sure the customer knows 'OK but that has a confidence of 17% - Are
you OK with that?).
My company is working on an 'Agile' process document that is unique
from any other 'line in the sand' methods I have seen (XP, Scrum,
etc.). We are interjecting Metrics and Risk Management as core
competencies and critical success factors in the process.
So does that make this a 'non-Agile process'? Or just a modified
Agile process? Have others done anything similar to what I am
describing and I am just not educated on it? We are almost done and I
am happy to post the results (in Word format) for this group if there
is interest. It is called 'Masdev' (at least for now).
Damon W. Carr, CTO
Master's Candidate in Computer Science
If you care about .NET you can watch a shameless self-promotional
video of me lecturing on .NET at Columbia to Master's Students in
Computer Science here: