--On Tuesday, December 07, 2004 9:25 PM +0900 "jaehoon.yu" <[EMAIL PROTECTED]> wrote:
Dear Prof. Johnson and CDSL staff,
Greetings. My day-dream of having an automated metrics collection system eventually led me to Hackystat.
I'm David who works at the IT service division of Samsung and my duty is to research various "software estimation" methods. My initial interest was to figure out how to build a credible software metrics database so my team can create a reliable cost estimation model upon it. But this was very difficult. I saw with my own eyes how reluctant, careless, and inaccurate the developers are when they were told to collect metrics. There is a feeling within our company that metrics collection can't be trusted and it's just there only to satisfy political needs of managers.
Although I really like the idea behind Hackystat and want to apply it within Samsung, I know that I have to convince the idea to my colleagues. The thing that bothers me most is that Hackystat seems to give only indirect measures.
I want to know the amount of "effort" and the corresponding "activity" of work that developers inject(concept of activity based costing), but it seems that "Active Time" of Hackystat only covers a portion of "Effort" and it doesn't tell which "activity" that developers are in. For example, how would I measure the effort of business analysts and software architects who attends Delphi meetings and makes UML diagrams on paper? I also want the analysis results of Hackystat to be useful to managers and executive officers, but it seems to only give value to developers. Moreover, I need to know "function point(IFPUG)" values because this is the standard size measure used within our company, but Hackystat seems to only provide LOC values.
I will try to read all of published papers of CDSL, but I would still appreciate your ideas and suggestions - How I can convince my boss that Hackystat can meet the needs of "Activity Based Costing"?
Thank you. ----------------------------------------------------------- David J. Yu, CFPS, CSQE, Associate PMP
Hi David,
Thanks for your email and I'm pleased that your daydream has led you to Hackystat!
Alas, I am not optimistic that Hackystat (or any other automated measurement system) would be successfully adopted in your company as long as the current culture persists in which metrics are (a) not perceived as helpful and cost-effective to developers (which is why they don't care about them) and (b) used inappropriately by management (for political, not technical purposes).
Hackystat is not designed to support the kinds of "direct" measures you refer to (such as the effort involved in writing UML diagrams on paper in a meeting). Indeed, the main focus of Hackystat development so far has been to apply metrics to support software development in ways different than cost estimation. For a description of how our approach (Software Project Telemetry) differs from traditional cost estimation applications of metrics, please see the following paper:
<http://csdl.ics.hawaii.edu/techreports/04-11/04-11.pdf>
This will give you a sense for one way we use Hackystat data to support software development. It also speaks directly to your problems of building a credible database for cost estimation.
While I understand that you want the results to be of interest to managers and executive officers, I think you may find that the best path to that outcome is to first concentrate on establishing a metrics program that is of use to developers. If you can figure out what metrics to capture that developers find truly useful (the bottom line: the data collected can be used to save developer time either directly or indirectly), and get that developer-centered approach to metrics collection and analysis adopted successfully, then as a natural consequence you will have results of interest to managers and executive officers. (For example, that developers are now accomplishing their work in less time :-) If you ignore the needs of developers, or if you place their needs second and the needs of managers first, then it is no surprise that developers don't want to participate and that they view the metrics program as political in nature.
Finally, while it is true that Hackystat does not currently support function points as a size measure, this would not be difficult to add to the system. We can talk about this further if upon further reflection you believe there is a real opportunity to use Hackystat at Samsung.
Good luck, and keep on dreaming!
Cheers, Philip
