Loading ...
Sorry, an error occurred while loading the content.

Re: [CMMi Process Improvement] Size based estimation

Expand Messages
  • Neil Potter
    Ross There is a short article on size estimation/tracking at http://www.processgroup.com/pgpostmar05.pdf (page 1) Neil ... Please feel free to email with
    Message 1 of 19 , Aug 31, 2006
    • 0 Attachment
      Ross

      There is a short article on size estimation/tracking at
      http://www.processgroup.com/pgpostmar05.pdf (page 1)

      Neil


      > To all who have responded to this today: your willingness to take the
      > time to help is appreciated beyond what mere words can express. The
      > information I have gleaned in this one day gives me hope that my
      > organization can actually achieve improvement using the CMMMi
      > framework. I wish I had found this forum sooner! Almost 2 years have
      > gone by since this effort started in earnest, and almost 1 year has
      > gone by since our SCAMPI B.
      >
      > Please continue to add to this thread. All input helps with
      > interpretation of the model.
      >
      > Thanks,
      >
      > Ross
      >
      >
      >
      >
      >
      >
      >
      >
      >
      > Yahoo! Groups Links
      >
      >
      >
      >
      >
      >

      Please feel free to email with questions.

      Thanks and regards, Neil Potter

      The Process Group
      Email: neil@...
      Tel:972-418-9541
      Fax:972-618-6283
      http://www.processgroup.com
    • Adail Retamal
      Rajendra, Some models use a (somewhat) polemic method known as Backfiring , i.e., to map a size based on some effort amount. This, as with all models, should
      Message 2 of 19 , Aug 31, 2006
      • 0 Attachment
        Rajendra,
         
        Some models use a (somewhat) polemic method known as "Backfiring", i.e., to map a size based on some effort amount.
         
        This, as with all models, should be used with care and only after a good calibration of the model with your own data.
         
        There is a product called Estimate Pro, now part of the Borland CaliberRM product, that does all this and provides helpful estimations for effort, duration, cost, staff, size, defects, etc. It uses COCOMO II, Monte Carlo simulation and Putnam (SLIM).
         
        Also, due its integration with CaliberRM, you can inform your effort directly in the requirements, and then have EstimatePro import those data and present the results. Very fast and flexible!
         
        Hope this helps!
         
        Adail

         
        On 8/31/06, Rajendra Gupta <gupta.rajendra@...> wrote:

        Hi,
         
        Based on the exchange of emails, I have understood that wide band delphi (WBD) method satisfies CMMI L2 requirements if handled well.  I support this view but have some queries based on some of the literature that I have gone through on this methodology:
         
        (a) While I have understood that it helps arriving at reasonably accurate effort estimates, I think WBD does not provide size estimation, as FP, LOC or COCOMO provides.  Am I correct in my understanding?  Here each task in WBS is given effort estimate.
         
        (b) Since it does not establish a size estimate, some metrics such as Productivity, Defect Density for the product cannot be arrived at.
         
        (c) Since WBD requires iterative sessions, as every iteration proceeds there is greater likelihood that effort estimate will continue to increase rather than decrease.  The risk is loosing customer due to higher estimate as compared to the customer.  Need to get practical experience from other users of WBD.
         
        Thanks in advance on your inputs.
         
        Regards
        Rajendra
         
         

         
      • Neil Potter
        Hi Gupta, ... Delphi can estimate any unit, size or effort. For effort, an effort amount is estimated for each task in the task list / WBS. For size, the team
        Message 3 of 19 , Aug 31, 2006
        • 0 Attachment
          Hi Gupta,

          See comments below:

          > Hi,
          >
          > Based on the exchange of emails, I have understood that wide band delphi
          > (WBD) method satisfies CMMI L2 requirements if handled well. I support this
          > view but have some queries based on some of the literature that I have gone
          > through on this methodology:
          >
          > (a) While I have understood that it helps arriving at reasonably accurate
          > effort estimates, I think WBD does not provide size estimation, as FP, LOC
          > or COCOMO provides. Am I correct in my understanding? Here each task in
          > WBS is given effort estimate.

          Delphi can estimate any unit, size or effort. For effort, an effort amount
          is estimated for each task in the task list / WBS. For size, the team
          estimates the size(es) of the project, but these might be broken down
          differently from the WBS. Eg size could be estimated by module, and effort
          by task.

          >
          > (b) Since it does not establish a size estimate, some metrics such as
          > Productivity, Defect Density for the product cannot be arrived at.

          See above. If the team estimates size, they can make some assumptions about
          Productivity

          >
          > (c) Since WBD requires iterative sessions, as every iteration proceeds there
          > is greater likelihood that effort estimate will continue to increase rather
          > than decrease. The risk is loosing customer due to higher estimate as
          > compared to the customer. Need to get practical experience from other users
          > of WBD.
          >
          Hmm. That's almost a level 1 thought (ie accurate data loses customers)!

          As the delphi goes through its 3-4 rounds (over a 2hr period) the estimates
          do become more robust, and if things have been left out in the task list,
          the number does get larger, but more accurate. What delphi does do is give
          the project team data to support its negotiations with the customer.

          Neil



          > Thanks in advance on your inputs.
          >
          > Regards
          > Rajendra
          >
          >
          > On 8/31/06, Mike Bandor <mbandor@...> wrote:
          >>
          >> Ross,
          >>
          >> I'm surprised by this as Wideband Delphi is an accepted estimation
          >> method as long as you are truly doing it and not a "wild guess" (or a WAG)
          >> instead. The mathematical models would be better, but the CMMI does not
          >> dictate which method should be used. That is an organizational decision.
          >> There are other "size" options besides FP or LOC within some of the other
          >> models available on the market. Personally, I would question the guidance
          >> your "former" LA provided. If you look at the model (Project Planning), it
          >> states:
          >>
          >> SP 1.2-1 Establish Estimates of Work Product and Task Attributes
          >> Establish and maintain estimates of the attributes of the work products
          >> and tasks. [PA163.IG101.SP102]
          >>
          >> Size is the primary input to many models used to estimate effort, cost,
          >> and schedule. The models may also be based on inputs such as connectivity,
          >> complexity, and structure. [PA163.IG101.SP102.N102]
          >> Examples of types of work products for which size estimates are made
          >> include the following: [PA163.IG101.SP102.N103]
          >> · Deliverable and nondeliverable work products
          >> · Documents
          >> · Operational and support software
          >>
          >> Examples of size measures include the following: [PA163.IG101.SP102.N104]
          >> · Number of functions
          >> · Function points
          >> · Source lines of code
          >> · Number of classes and objects
          >> · Number of requirements
          >> · Number of interfaces
          >> · Number of pages
          >> · Number of inputs and outputs
          >> · Number of technical risk items
          >> · Volume of data
          >> The estimates should be consistent with project requirements to determine
          >> the project's effort, cost, and schedule. A relative level of difficulty or
          >> complexity should be assigned for each size attribute.
          >>
          >>
          >> So you see, there are other "size" options and methods to determine the
          >> effort. It sounds like the LA is taking the wording for subpractice 2 too
          >> literally:
          >>
          >> 2. Use appropriate methods to determine the attributes of the work
          >> products and tasks that will be used to estimate the resource requirements.
          >> [PA163.IG101.SP102.SubP102]
          >> Methods for determining size and complexity should be based on validated
          >> models or historical data. [PA163.IG101.SP102.SubP102.N101]
          >> The methods for determining attributes evolve as our understanding of the
          >> relationship of product characteristics to attributes increases.[
          >> PA163.IG101.SP102.SubP102.N102]
          >> Examples of current methods include the following: [
          >> PA163.IG101.SP102.SubP102.N103]
          >>
          >> · Number of logic gates for integrated circuit design
          >> · Lines of code or function points for software
          >> · Number/complexity of requirements for systems engineering
          >> · Number of square feet for standard-specified residential homes
          >>
          >>
          >> Like I said, if you are truly using Wideband Delphi (and keeping the
          >> historical measurements as part of the inputs for the next estimation
          >> effort) there shouldn't be a problem.
          >>
          >> Mike
          >>
          >> Michael S. Bandor
          >> Acquisition Support Program (ASP)
          >> Software Engineering Institute - Carnegie Mellon University
          >> mbandor@...
          >> http://www.sei.cmu.edu/programs/acquisition-support/
          >>
          >> NOTE: The information or views represented in this posting do not
          >> necessarily reflect those of the SEI or Carnegie Mellon University.
          >>
          >> cmmi_user_at_jci wrote:
          >>
          >> I work for a tier 1 Automotive supplier working on CMMi L2 for our
          >> software organization. We have had 1 SCAMPI B appraisal so far, and
          >> are still wandering in the wilderness trying to solve some of our
          >> bigger gaps, estimation being a large one.
          >>
          >> During our appraisal, we were informed that Wideband Delphi was not
          >> an acceptable estimation method, and that we must have some type of
          >> mathmatical model for estimation based on historical information.
          >> This is an issue for us because we do not have systems in place to
          >> generate historical information. We were also told that the only
          >> options for size were either LOC or FP. FP is not an option for us
          >> since we will not make the investment to train anyone in how to do
          >> it. Also, our sponsor made it clear that FP would never be
          >> considered.
          >>
          >> After reading some of posts to this board, it seems to me that we
          >> should be able to find something that works for us as long as it is
          >> some type of size attribute, use Wideband Delphi to estimate it, and
          >> start there with the intent of refining the process over time as
          >> historical data is gathered. This was absolutely denied as an option
          >> by our "former" Lead Appraiser.
          >>
          >> Any thoughts on this from any of you that have been through this
          >> pain? I'm optimistic because I see names that are familiar to me
          >> from past SEPG conferences.
          >>
          >> Thanks in advance,
          >>
          >> Ross
          >>
          >>
          >>
          >
          >

          Please feel free to email with questions.

          Thanks and regards, Neil Potter

          The Process Group
          Email: neil@...
          Tel:972-418-9541
          Fax:972-618-6283
          http://www.processgroup.com
        • Rajendra Gupta
          Thanks Neil and Adail. Refering to Neil s response, WBD (Wideband Delphi) can be used to derive either effort or size. If I use WBD to derive size, then I
          Message 4 of 19 , Sep 2, 2006
          • 0 Attachment
            Thanks Neil and Adail.
             
            Refering to Neil's response, WBD (Wideband Delphi) can be used to derive either effort or size.  If I use WBD to derive size, then I will be able to work out metrics such as Productivity.  I have follow-up questions:
             
            (a) If I need to know Productivity, can I still use WBD even when I am using it to derive effort and not size.
             
            (b) While CMMI talks about 'attributes...', the older version CMM clearly mentioned size estimation.  Does it mean that WBD (when used to derive effort and not size) will not work in CMM even though it works fine in CMMI.
             
            Regards
            Rajendra
             
            On 8/31/06, Adail Retamal <adail.retamal@...> wrote:

            Rajendra,
             
            Some models use a (somewhat) polemic method known as "Backfiring", i.e., to map a size based on some effort amount.
             
            This, as with all models, should be used with care and only after a good calibration of the model with your own data.
             
            There is a product called Estimate Pro, now part of the Borland CaliberRM product, that does all this and provides helpful estimations for effort, duration, cost, staff, size, defects, etc. It uses COCOMO II, Monte Carlo simulation and Putnam (SLIM).
             
            Also, due its integration with CaliberRM, you can inform your effort directly in the requirements, and then have EstimatePro import those data and present the results. Very fast and flexible!
             
            Hope this helps!
             
            Adail

             
            On 8/31/06, Rajendra Gupta <gupta.rajendra@... > wrote:

            Hi,
             
            Based on the exchange of emails, I have understood that wide band delphi (WBD) method satisfies CMMI L2 requirements if handled well.  I support this view but have some queries based on some of the literature that I have gone through on this methodology:
             
            (a) While I have understood that it helps arriving at reasonably accurate effort estimates, I think WBD does not provide size estimation, as FP, LOC or COCOMO provides.  Am I correct in my understanding?  Here each task in WBS is given effort estimate.
             
            (b) Since it does not establish a size estimate, some metrics such as Productivity, Defect Density for the product cannot be arrived at.
             
            (c) Since WBD requires iterative sessions, as every iteration proceeds there is greater likelihood that effort estimate will continue to increase rather than decrease.  The risk is loosing customer due to higher estimate as compared to the customer.  Need to get practical experience from other users of WBD.
             
            Thanks in advance on your inputs.
             
            Regards
            Rajendra
             

             

             




            --
            Regards
            Rajendra
            1-401-827-3036 (Office)
            1-401-487-4628 (Cell)
          • Neil Potter
            Hi Rajendra ... You can use WBD just to estimate effort and not size. ... Assume that cmm and cmmi are identical on this issue. They just used different words
            Message 5 of 19 , Sep 4, 2006
            • 0 Attachment
              Hi Rajendra

              > Thanks Neil and Adail.
              >
              > Refering to Neil's response, WBD (Wideband Delphi) can be used to derive
              > either effort or size. If I use WBD to derive size, then I will be able to
              > work out metrics such as Productivity. I have follow-up questions:
              >
              > (a) If I need to know Productivity, can I still use WBD even when I am using
              > it to derive effort and not size.

              You can use WBD just to estimate effort and not size.
              >
              > (b) While CMMI talks about 'attributes...', the older version CMM clearly
              > mentioned size estimation. Does it mean that WBD (when used to derive
              > effort and not size) will not work in CMM even though it works fine in CMMI.

              Assume that cmm and cmmi are identical on this issue. They just used
              different words in cmmi to explain the same thing. My guess is that 'size'
              was read as 'Lines of code' by many readers, so 'work product attributes' is
              a more generic term. But they are the same. So WBD works for any estimation
              practice in any model

              Neil


              >
              > Regards
              > Rajendra
              >
              > On 8/31/06, Adail Retamal <adail.retamal@...> wrote:
              >>
              >> Rajendra,
              >>
              >> Some models use a (somewhat) polemic method known as "Backfiring", i.e.,
              >> to map a size based on some effort amount.
              >>
              >> This, as with all models, should be used with care and only after a good
              >> calibration of the model with your own data.
              >>
              >> There is a product called Estimate Pro, now part of the Borland CaliberRM
              >> product, that does all this and provides helpful estimations for effort,
              >> duration, cost, staff, size, defects, etc. It uses COCOMO II, Monte Carlo
              >> simulation and Putnam (SLIM).
              >>
              >> Also, due its integration with CaliberRM, you can inform your effort
              >> directly in the requirements, and then have EstimatePro import those data
              >> and present the results. Very fast and flexible!
              >>
              >> Hope this helps!
              >>
              >> Adail
              >>
              >>
              >> On 8/31/06, Rajendra Gupta <gupta.rajendra@...> wrote:
              >>>
              >>> Hi,
              >>>
              >>> Based on the exchange of emails, I have understood that wide band delphi
              >>> (WBD) method satisfies CMMI L2 requirements if handled well. I support this
              >>> view but have some queries based on some of the literature that I have gone
              >>> through on this methodology:
              >>>
              >>> (a) While I have understood that it helps arriving at reasonably
              >>> accurate effort estimates, I think WBD does not provide size estimation, as
              >>> FP, LOC or COCOMO provides. Am I correct in my understanding? Here each
              >>> task in WBS is given effort estimate.
              >>>
              >>> (b) Since it does not establish a size estimate, some metrics such as
              >>> Productivity, Defect Density for the product cannot be arrived at.
              >>>
              >>> (c) Since WBD requires iterative sessions, as every iteration proceeds
              >>> there is greater likelihood that effort estimate will continue to increase
              >>> rather than decrease. The risk is loosing customer due to higher estimate
              >>> as compared to the customer. Need to get practical experience from other
              >>> users of WBD.
              >>>
              >>> Thanks in advance on your inputs.
              >>>
              >>> Regards
              >>> Rajendra
              >>>
              >>>
              >>>
              >>>
              >>>
              >>>
              >>
              >>
              >
              >

              Please feel free to email with questions.

              Thanks and regards, Neil Potter

              The Process Group
              Email: neil@...
              Tel:972-418-9541
              Fax:972-618-6283
              http://www.processgroup.com
            • Patrick OToole
              Rajendra/Neil, Stepping back from this a bit, the way that I typically see Wideband Delphi used is when the project is estimating effort. However, if done
              Message 6 of 19 , Sep 5, 2006
              • 0 Attachment
                 
                Rajendra/Neil,
                 
                Stepping back from this a bit, the way that I typically see Wideband Delphi used is when the project is estimating effort.  However, if done properly, with each iteration, the team will be documenting their "estimation assumptions" - which REALLY turn out to be the project attributes that serve as the basis of the effort estimates.  So, while trying to get their arms around the project's effort estimate, they wind up discussing "size-like" parameters.
                 
                That, to me, is one of the primary strengths of the Wideband Delphi method - it pools the collective knowledge of the estimators, discusses a much broader array of work and work product attributes, and (hopefully) results in these being documented.
                 
                If the organization is on their toes, they will develop a good understanding of which attributes lead to more/less effort being required, and will use that information to get better estimates on future projects.
                 
                I have always considered the Wideband Delphi method perfectly appropriate for ML2 provided that the "basis of estimate" assumptions are documented.
                 
                Regards,
                 
                Pat
                 
                 
                 
                 
                ----- Original Message -----
                Sent: Monday, September 04, 2006 8:28 PM
                Subject: Re: [CMMi Process Improvement] Size based estimation

                Hi Rajendra

                > Thanks Neil and Adail.
                >
                > Refering to Neil's response, WBD (Wideband Delphi) can be used to derive
                > either effort or size. If I use WBD to derive size, then I will be able to
                > work out metrics such as Productivity. I have follow-up questions:
                >
                > (a) If I need to know Productivity, can I still use WBD even when I am using
                > it to derive effort and not size.

                You can use WBD just to estimate effort and not size.
                >
                > (b) While CMMI talks about 'attributes. ..', the older version CMM clearly
                > mentioned size estimation. Does it mean that WBD (when used to derive
                > effort and not size) will not work in CMM even though it works fine in CMMI.

                Assume that cmm and cmmi are identical on this issue. They just used
                different words in cmmi to explain the same thing. My guess is that 'size'
                was read as 'Lines of code' by many readers, so 'work product attributes' is
                a more generic term. But they are the same. So WBD works for any estimation
                practice in any model

                Neil

                >
                > Regards
                > Rajendra
                >
                > On 8/31/06, Adail Retamal <adail.retamal@ gmail.com> wrote:
                >>
                >> Rajendra,
                >>
                >> Some models use a (somewhat) polemic method known as "Backfiring" , i.e.,
                >> to map a size based on some effort amount.
                >>
                >> This, as with all models, should be used with care and only after a good
                >> calibration of the model with your own data.
                >>
                >> There is a product called Estimate Pro, now part of the Borland CaliberRM
                >> product, that does all this and provides helpful estimations for effort,
                >> duration, cost, staff, size, defects, etc. It uses COCOMO II, Monte Carlo
                >> simulation and Putnam (SLIM).
                >>
                >> Also, due its integration with CaliberRM, you can inform your effort
                >> directly in the requirements, and then have EstimatePro import those data
                >> and present the results. Very fast and flexible!
                >>
                >> Hope this helps!
                >>
                >> Adail
                >>
                >>
                >> On 8/31/06, Rajendra Gupta <gupta.rajendra@ gmail.com> wrote:
                >>>
                >>> Hi,
                >>>
                >>> Based on the exchange of emails, I have understood that wide band delphi
                >>> (WBD) method satisfies CMMI L2 requirements if handled well. I support this
                >>> view but have some queries based on some of the literature that I have gone
                >>> through on this methodology:
                >>>
                >>> (a) While I have understood that it helps arriving at reasonably
                >>> accurate effort estimates, I think WBD does not provide size estimation, as
                >>> FP, LOC or COCOMO provides. Am I correct in my understanding? Here each
                >>> task in WBS is given effort estimate.
                >>>
                >>> (b) Since it does not establish a size estimate, some metrics such as
                >>> Productivity, Defect Density for the product cannot be arrived at.
                >>>
                >>> (c) Since WBD requires iterative sessions, as every iteration proceeds
                >>> there is greater likelihood that effort estimate will continue to increase
                >>> rather than decrease. The risk is loosing customer due to higher estimate
                >>> as compared to the customer. Need to get practical experience from other
                >>> users of WBD.
                >>>
                >>> Thanks in advance on your inputs.
                >>>
                >>> Regards
                >>> Rajendra
                >>>
                >>>
                >>>
                >>>
                >>>
                >>>
                >>
                >>
                >
                >

                Please feel free to email with questions.

                Thanks and regards, Neil Potter

                The Process Group
                Email: neil@processgroup. com
                Tel:972-418- 9541
                Fax:972-618- 6283
                http://www.processg roup.com

              • Ross Timmerman
                Does anyone out there have any feedback on COTS estimation tools that actually work, and that can help a L1 organiztion move to L2 and beyond? Assume that at
                Message 7 of 19 , Sep 6, 2006
                • 0 Attachment
                  Does anyone out there have any feedback on COTS estimation tools that
                  actually work, and that can help a L1 organiztion move to L2 and beyond?

                  Assume that at least some historical data is available to support the
                  tool. One option I am looking at is QSM SLIM Estimate. Anyone use
                  that?

                  Thanks,

                  Ross
                • Ross Timmerman
                  Patrick, Can I infer from this that it is acceptable to use WBD to establish effort estimates for a project directly without relating them to some type of
                  Message 8 of 19 , Sep 6, 2006
                  • 0 Attachment
                    Patrick,

                    Can I infer from this that it is acceptable to use WBD to establish
                    effort estimates for a project directly without relating them to some
                    type of "size"? Based on this thread, which I started, I have been
                    working on trying to get a grip on estimation for my small software
                    organization, and am constantly getting tripped up by the issue of
                    size in the CMMi model. Our most successful estimation method was
                    not based on size, and was typically within 10% of the actual effort
                    for a project which was good enough for our business. For over a
                    year we have been trying to estabish an estimation model based on
                    size with absolutely no success. For CMMi L2, we had defined WBD to
                    estimate efforts for WBS tasks, but this was not acceptable, both
                    because the appraiser did not like WBD and because we weren't basing
                    our estimates on the size of something. If estimates based on the
                    size of something are an absolute must, I see no way that we will
                    ever make it to L2 - it's not worth it to us at the moment. SPICE
                    (ISO-15504) might be a better, more realistic option. I can't
                    imagine that the SEI wants to exclude small organizations, but they
                    are certainly doing a good job of it.

                    Ross



                    --- In cmmi_process_improvement@yahoogroups.com, "Patrick OToole"
                    <PACT.otoole@...> wrote:
                    >
                    >
                    > Rajendra/Neil,
                    >
                    > Stepping back from this a bit, the way that I typically see
                    Wideband Delphi used is when the project is estimating effort.
                    However, if done properly, with each iteration, the team will be
                    documenting their "estimation assumptions" - which REALLY turn out to
                    be the project attributes that serve as the basis of the effort
                    estimates. So, while trying to get their arms around the project's
                    effort estimate, they wind up discussing "size-like" parameters.
                    >
                    > That, to me, is one of the primary strengths of the Wideband Delphi
                    method - it pools the collective knowledge of the estimators,
                    discusses a much broader array of work and work product attributes,
                    and (hopefully) results in these being documented.
                    >
                    > If the organization is on their toes, they will develop a good
                    understanding of which attributes lead to more/less effort being
                    required, and will use that information to get better estimates on
                    future projects.
                    >
                    > I have always considered the Wideband Delphi method perfectly
                    appropriate for ML2 provided that the "basis of estimate" assumptions
                    are documented.
                    >
                    > Regards,
                    >
                    > Pat
                    >
                    >
                    >
                    >
                    > ----- Original Message -----
                    > From: Neil Potter
                    > To: cmmi_process_improvement@yahoogroups.com
                    > Sent: Monday, September 04, 2006 8:28 PM
                    > Subject: Re: [CMMi Process Improvement] Size based estimation
                    >
                    >
                    > Hi Rajendra
                    >
                    > > Thanks Neil and Adail.
                    > >
                    > > Refering to Neil's response, WBD (Wideband Delphi) can be used
                    to derive
                    > > either effort or size. If I use WBD to derive size, then I will
                    be able to
                    > > work out metrics such as Productivity. I have follow-up
                    questions:
                    > >
                    > > (a) If I need to know Productivity, can I still use WBD even
                    when I am using
                    > > it to derive effort and not size.
                    >
                    > You can use WBD just to estimate effort and not size.
                    > >
                    > > (b) While CMMI talks about 'attributes...', the older version
                    CMM clearly
                    > > mentioned size estimation. Does it mean that WBD (when used to
                    derive
                    > > effort and not size) will not work in CMM even though it works
                    fine in CMMI.
                    >
                    > Assume that cmm and cmmi are identical on this issue. They just
                    used
                    > different words in cmmi to explain the same thing. My guess is
                    that 'size'
                    > was read as 'Lines of code' by many readers, so 'work product
                    attributes' is
                    > a more generic term. But they are the same. So WBD works for any
                    estimation
                    > practice in any model
                    >
                    > Neil
                    >
                    > >
                    > > Regards
                    > > Rajendra
                    > >
                    > > On 8/31/06, Adail Retamal <adail.retamal@...> wrote:
                    > >>
                    > >> Rajendra,
                    > >>
                    > >> Some models use a (somewhat) polemic method known
                    as "Backfiring", i.e.,
                    > >> to map a size based on some effort amount.
                    > >>
                    > >> This, as with all models, should be used with care and only
                    after a good
                    > >> calibration of the model with your own data.
                    > >>
                    > >> There is a product called Estimate Pro, now part of the
                    Borland CaliberRM
                    > >> product, that does all this and provides helpful estimations
                    for effort,
                    > >> duration, cost, staff, size, defects, etc. It uses COCOMO II,
                    Monte Carlo
                    > >> simulation and Putnam (SLIM).
                    > >>
                    > >> Also, due its integration with CaliberRM, you can inform your
                    effort
                    > >> directly in the requirements, and then have EstimatePro import
                    those data
                    > >> and present the results. Very fast and flexible!
                    > >>
                    > >> Hope this helps!
                    > >>
                    > >> Adail
                    > >>
                    > >>
                    > >> On 8/31/06, Rajendra Gupta <gupta.rajendra@...> wrote:
                    > >>>
                    > >>> Hi,
                    > >>>
                    > >>> Based on the exchange of emails, I have understood that wide
                    band delphi
                    > >>> (WBD) method satisfies CMMI L2 requirements if handled well.
                    I support this
                    > >>> view but have some queries based on some of the literature
                    that I have gone
                    > >>> through on this methodology:
                    > >>>
                    > >>> (a) While I have understood that it helps arriving at
                    reasonably
                    > >>> accurate effort estimates, I think WBD does not provide size
                    estimation, as
                    > >>> FP, LOC or COCOMO provides. Am I correct in my understanding?
                    Here each
                    > >>> task in WBS is given effort estimate.
                    > >>>
                    > >>> (b) Since it does not establish a size estimate, some metrics
                    such as
                    > >>> Productivity, Defect Density for the product cannot be
                    arrived at.
                    > >>>
                    > >>> (c) Since WBD requires iterative sessions, as every iteration
                    proceeds
                    > >>> there is greater likelihood that effort estimate will
                    continue to increase
                    > >>> rather than decrease. The risk is loosing customer due to
                    higher estimate
                    > >>> as compared to the customer. Need to get practical experience
                    from other
                    > >>> users of WBD.
                    > >>>
                    > >>> Thanks in advance on your inputs.
                    > >>>
                    > >>> Regards
                    > >>> Rajendra
                    > >>>
                    > >>>
                    > >>>
                    > >>>
                    > >>>
                    > >>>
                    > >>
                    > >>
                    > >
                    > >
                    >
                    > Please feel free to email with questions.
                    >
                    > Thanks and regards, Neil Potter
                    >
                    > The Process Group
                    > Email: neil@...
                    > Tel:972-418-9541
                    > Fax:972-618-6283
                    > http://www.processgroup.com
                    >
                  • Neil Potter
                    Hi Ross I ll let pat reply. But my take is: WBD can be used for any metric estimation (effort / size / ...) For L2, size does not have to correlate to effort
                    Message 9 of 19 , Sep 6, 2006
                    • 0 Attachment
                      Hi Ross

                      I'll let pat reply. But my take is:

                      WBD can be used for any metric estimation (effort / size / ...)

                      For L2, size does not have to correlate to effort (it should be somewhat
                      proportional to be useful). Effort does not have to be derived from size
                      either. Size can be taken as a 'how big is my project' metric. It can be
                      estimated, used as assumptions for the effort estimates (eg I have 1000hrs
                      for 100 changes), and more importantly tracked in PMC to detect early
                      changes in project bigness. Eg if 30 changes come in tomorrow, something
                      tells me that's a approx 30% increase in scope. It might be 30 small changes
                      or 30 large ones, but a large enough number to cause a reevaluation of the
                      effort estimate (and risk and schedule).

                      That's L2 for size.

                      Neil



                      > Patrick,
                      >
                      > Can I infer from this that it is acceptable to use WBD to establish
                      > effort estimates for a project directly without relating them to some
                      > type of "size"? Based on this thread, which I started, I have been
                      > working on trying to get a grip on estimation for my small software
                      > organization, and am constantly getting tripped up by the issue of
                      > size in the CMMi model. Our most successful estimation method was
                      > not based on size, and was typically within 10% of the actual effort
                      > for a project which was good enough for our business. For over a
                      > year we have been trying to estabish an estimation model based on
                      > size with absolutely no success. For CMMi L2, we had defined WBD to
                      > estimate efforts for WBS tasks, but this was not acceptable, both
                      > because the appraiser did not like WBD and because we weren't basing
                      > our estimates on the size of something. If estimates based on the
                      > size of something are an absolute must, I see no way that we will
                      > ever make it to L2 - it's not worth it to us at the moment. SPICE
                      > (ISO-15504) might be a better, more realistic option. I can't
                      > imagine that the SEI wants to exclude small organizations, but they
                      > are certainly doing a good job of it.
                      >
                      > Ross
                      >
                      >
                      >
                      > --- In cmmi_process_improvement@yahoogroups.com, "Patrick OToole"
                      > <PACT.otoole@...> wrote:
                      >>
                      >>
                      >> Rajendra/Neil,
                      >>
                      >> Stepping back from this a bit, the way that I typically see
                      > Wideband Delphi used is when the project is estimating effort.
                      > However, if done properly, with each iteration, the team will be
                      > documenting their "estimation assumptions" - which REALLY turn out to
                      > be the project attributes that serve as the basis of the effort
                      > estimates. So, while trying to get their arms around the project's
                      > effort estimate, they wind up discussing "size-like" parameters.
                      >>
                      >> That, to me, is one of the primary strengths of the Wideband Delphi
                      > method - it pools the collective knowledge of the estimators,
                      > discusses a much broader array of work and work product attributes,
                      > and (hopefully) results in these being documented.
                      >>
                      >> If the organization is on their toes, they will develop a good
                      > understanding of which attributes lead to more/less effort being
                      > required, and will use that information to get better estimates on
                      > future projects.
                      >>
                      >> I have always considered the Wideband Delphi method perfectly
                      > appropriate for ML2 provided that the "basis of estimate" assumptions
                      > are documented.
                      >>
                      >> Regards,
                      >>
                      >> Pat
                      >>
                      >>
                      >>
                      >>
                      >> ----- Original Message -----
                      >> From: Neil Potter
                      >> To: cmmi_process_improvement@yahoogroups.com
                      >> Sent: Monday, September 04, 2006 8:28 PM
                      >> Subject: Re: [CMMi Process Improvement] Size based estimation
                      >>
                      >>
                      >> Hi Rajendra
                      >>
                      >>> Thanks Neil and Adail.
                      >>>
                      >>> Refering to Neil's response, WBD (Wideband Delphi) can be used
                      > to derive
                      >>> either effort or size. If I use WBD to derive size, then I will
                      > be able to
                      >>> work out metrics such as Productivity. I have follow-up
                      > questions:
                      >>>
                      >>> (a) If I need to know Productivity, can I still use WBD even
                      > when I am using
                      >>> it to derive effort and not size.
                      >>
                      >> You can use WBD just to estimate effort and not size.
                      >>>
                      >>> (b) While CMMI talks about 'attributes...', the older version
                      > CMM clearly
                      >>> mentioned size estimation. Does it mean that WBD (when used to
                      > derive
                      >>> effort and not size) will not work in CMM even though it works
                      > fine in CMMI.
                      >>
                      >> Assume that cmm and cmmi are identical on this issue. They just
                      > used
                      >> different words in cmmi to explain the same thing. My guess is
                      > that 'size'
                      >> was read as 'Lines of code' by many readers, so 'work product
                      > attributes' is
                      >> a more generic term. But they are the same. So WBD works for any
                      > estimation
                      >> practice in any model
                      >>
                      >> Neil
                      >>
                      >>>
                      >>> Regards
                      >>> Rajendra
                      >>>
                      >>> On 8/31/06, Adail Retamal <adail.retamal@...> wrote:
                      >>>>
                      >>>> Rajendra,
                      >>>>
                      >>>> Some models use a (somewhat) polemic method known
                      > as "Backfiring", i.e.,
                      >>>> to map a size based on some effort amount.
                      >>>>
                      >>>> This, as with all models, should be used with care and only
                      > after a good
                      >>>> calibration of the model with your own data.
                      >>>>
                      >>>> There is a product called Estimate Pro, now part of the
                      > Borland CaliberRM
                      >>>> product, that does all this and provides helpful estimations
                      > for effort,
                      >>>> duration, cost, staff, size, defects, etc. It uses COCOMO II,
                      > Monte Carlo
                      >>>> simulation and Putnam (SLIM).
                      >>>>
                      >>>> Also, due its integration with CaliberRM, you can inform your
                      > effort
                      >>>> directly in the requirements, and then have EstimatePro import
                      > those data
                      >>>> and present the results. Very fast and flexible!
                      >>>>
                      >>>> Hope this helps!
                      >>>>
                      >>>> Adail
                      >>>>
                      >>>>
                      >>>> On 8/31/06, Rajendra Gupta <gupta.rajendra@...> wrote:
                      >>>>>
                      >>>>> Hi,
                      >>>>>
                      >>>>> Based on the exchange of emails, I have understood that wide
                      > band delphi
                      >>>>> (WBD) method satisfies CMMI L2 requirements if handled well.
                      > I support this
                      >>>>> view but have some queries based on some of the literature
                      > that I have gone
                      >>>>> through on this methodology:
                      >>>>>
                      >>>>> (a) While I have understood that it helps arriving at
                      > reasonably
                      >>>>> accurate effort estimates, I think WBD does not provide size
                      > estimation, as
                      >>>>> FP, LOC or COCOMO provides. Am I correct in my understanding?
                      > Here each
                      >>>>> task in WBS is given effort estimate.
                      >>>>>
                      >>>>> (b) Since it does not establish a size estimate, some metrics
                      > such as
                      >>>>> Productivity, Defect Density for the product cannot be
                      > arrived at.
                      >>>>>
                      >>>>> (c) Since WBD requires iterative sessions, as every iteration
                      > proceeds
                      >>>>> there is greater likelihood that effort estimate will
                      > continue to increase
                      >>>>> rather than decrease. The risk is loosing customer due to
                      > higher estimate
                      >>>>> as compared to the customer. Need to get practical experience
                      > from other
                      >>>>> users of WBD.
                      >>>>>
                      >>>>> Thanks in advance on your inputs.
                      >>>>>
                      >>>>> Regards
                      >>>>> Rajendra
                      >>>>>
                      >>>>>
                      >>>>>
                      >>>>>
                      >>>>>
                      >>>>>
                      >>>>
                      >>>>
                      >>>
                      >>>
                      >>
                      >> Please feel free to email with questions.
                      >>
                      >> Thanks and regards, Neil Potter
                      >>
                      >> The Process Group
                      >> Email: neil@...
                      >> Tel:972-418-9541
                      >> Fax:972-618-6283
                      >> http://www.processgroup.com
                      >>
                      >
                      >
                      >
                      >
                      >
                      >
                      >
                      >
                      >
                      >
                      >
                      > Yahoo! Groups Links
                      >
                      >
                      >
                      >
                      >
                      >
                      >
                      >

                      Please feel free to email with questions.

                      Thanks and regards, Neil Potter

                      The Process Group
                      Email: neil@...
                      Tel:972-418-9541
                      Fax:972-618-6283
                      http://www.processgroup.com
                    • Adail Retamal
                      Ross, Borland CaliberRM comes with EstimatePro, a very good estimation tool based on COCOMO II, Putnam (SLIM) and Monte Carlo simulation. The good thing about
                      Message 10 of 19 , Sep 6, 2006
                      • 0 Attachment
                        Ross,
                         
                        Borland CaliberRM comes with EstimatePro, a very good estimation tool based on COCOMO II, Putnam (SLIM) and Monte Carlo simulation. The good thing about it, in addition to the estimation, is that you can use your requirements efforts as input to the estimation engine.
                         
                        It's worth a look! :)
                         
                         
                        Cheers,
                         
                        Adail

                         
                        On 9/6/06, Ross Timmerman <sum1.at.jci@...> wrote:

                        Does anyone out there have any feedback on COTS estimation tools that
                        actually work, and that can help a L1 organiztion move to L2 and beyond?

                        Assume that at least some historical data is available to support the
                        tool. One option I am looking at is QSM SLIM Estimate. Anyone use
                        that?

                        Thanks,

                        Ross

                         

                         
                      • Patrick OToole
                        Ross, Take a look at the documented estimation assumptions that resulted from your Wideband Delphi effort estimations. More than likely you will find that
                        Message 11 of 19 , Sep 6, 2006
                        • 0 Attachment
                           
                          Ross,
                           
                          Take a look at the documented estimation assumptions that resulted from your Wideband Delphi effort estimations.  More than likely you will find that those assumptions contain a number of "work product-related" or "task related" attributes.
                           
                          Things like:
                           
                          We assume that the website will require 12-15 pages to be provided.  Of these, 5 must be developed from scratch and the others will be modified versions of existing pages.
                           
                          We assume that 4 new tables must be added to the database.  Each table will have 4-7 fields.
                           
                          We assume that no change to the billing system will be required.
                           
                          We assume that the User Manual will be 100-125 pages, the Installation Manual will be 20-25 pages, and that the Maintenance Manual will be 35-50 pages.  About 50% of the User Manual can be reused from an existing application.
                           
                          Of the 30 requirements that affect the payroll system, we assume that 10-12 will be complex changes, 7-10 will be of medium complexity; and the remainder will be low complexity changes.
                           
                          Etc.
                           
                           
                          If your Delphi effort estimation exercise results in assumptions of this ilk being documented, then you're golden.  If you want to be PLATINUM, then revisit these "basis of effort" estimation assumptions after you've completed design.  You'll know a heckuva lot more about the project and the solution at that point.  If you modify both the estimation assumptions based on this new information as well as the effort estimate itself, you should be find for both PP SP1.2 and the size portion of PMC SP1.1.
                           
                          If you DON'T document these attribute-related assumptions, then you'll have no direct artifact to show for PP SP1.2 (nor the size-related portion of PMC SP1.1), and you'll get an unfavorable result.
                           
                          Getting people to document (and modify) such assumptions sure seems a whole lot easier to me than trying to generate a size-based estimation model!
                           
                          Regards,
                           
                          Pat
                           
                           
                          Regards,
                           
                          Pat
                           
                           
                           
                           
                          ----- Original Message -----
                          Sent: Wednesday, September 06, 2006 1:17 PM
                          Subject: Re: [CMMi Process Improvement] Size based estimation

                          Patrick,

                          Can I infer from this that it is acceptable to use WBD to establish
                          effort estimates for a project directly without relating them to some
                          type of "size"? Based on this thread, which I started, I have been
                          working on trying to get a grip on estimation for my small software
                          organization, and am constantly getting tripped up by the issue of
                          size in the CMMi model. Our most successful estimation method was
                          not based on size, and was typically within 10% of the actual effort
                          for a project which was good enough for our business. For over a
                          year we have been trying to estabish an estimation model based on
                          size with absolutely no success. For CMMi L2, we had defined WBD to
                          estimate efforts for WBS tasks, but this was not acceptable, both
                          because the appraiser did not like WBD and because we weren't basing
                          our estimates on the size of something. If estimates based on the
                          size of something are an absolute must, I see no way that we will
                          ever make it to L2 - it's not worth it to us at the moment. SPICE
                          (ISO-15504) might be a better, more realistic option. I can't
                          imagine that the SEI wants to exclude small organizations, but they
                          are certainly doing a good job of it.

                          Ross

                          --- In cmmi_process_ improvement@ yahoogroups. com, "Patrick OToole"
                          <PACT.otoole@ ...> wrote:
                          >
                          >
                          > Rajendra/Neil,
                          >
                          > Stepping back from this a bit, the way that I typically see
                          Wideband Delphi used is when the project is estimating effort.
                          However, if done properly, with each iteration, the team will be
                          documenting their "estimation assumptions" - which REALLY turn out to
                          be the project attributes that serve as the basis of the effort
                          estimates. So, while trying to get their arms around the project's
                          effort estimate, they wind up discussing "size-like" parameters.
                          >
                          > That, to me, is one of the primary strengths of the Wideband Delphi
                          method - it pools the collective knowledge of the estimators,
                          discusses a much broader array of work and work product attributes,
                          and (hopefully) results in these being documented.
                          >
                          > If the organization is on their toes, they will develop a good
                          understanding of which attributes lead to more/less effort being
                          required, and will use that information to get better estimates on
                          future projects.
                          >
                          > I have always considered the Wideband Delphi method perfectly
                          appropriate for ML2 provided that the "basis of estimate" assumptions
                          are documented.
                          >
                          > Regards,
                          >
                          > Pat
                          >
                          >
                          >
                          >
                          > ----- Original Message -----
                          > From: Neil Potter
                          > To: cmmi_process_ improvement@ yahoogroups. com
                          > Sent: Monday, September 04, 2006 8:28 PM
                          > Subject: Re: [CMMi Process Improvement] Size based estimation
                          >
                          >
                          > Hi Rajendra
                          >
                          > > Thanks Neil and Adail.
                          > >
                          > > Refering to Neil's response, WBD (Wideband Delphi) can be used
                          to derive
                          > > either effort or size. If I use WBD to derive size, then I will
                          be able to
                          > > work out metrics such as Productivity. I have follow-up
                          questions:
                          > >
                          > > (a) If I need to know Productivity, can I still use WBD even
                          when I am using
                          > > it to derive effort and not size.
                          >
                          > You can use WBD just to estimate effort and not size.
                          > >
                          > > (b) While CMMI talks about 'attributes. ..', the older version
                          CMM clearly
                          > > mentioned size estimation. Does it mean that WBD (when used to
                          derive
                          > > effort and not size) will not work in CMM even though it works
                          fine in CMMI.
                          >
                          > Assume that cmm and cmmi are identical on this issue. They just
                          used
                          > different words in cmmi to explain the same thing. My guess is
                          that 'size'
                          > was read as 'Lines of code' by many readers, so 'work product
                          attributes' is
                          > a more generic term. But they are the same. So WBD works for any
                          estimation
                          > practice in any model
                          >
                          > Neil
                          >
                          > >
                          > > Regards
                          > > Rajendra
                          > >
                          > > On 8/31/06, Adail Retamal <adail.retamal@ ...> wrote:
                          > >>
                          > >> Rajendra,
                          > >>
                          > >> Some models use a (somewhat) polemic method known
                          as "Backfiring" , i.e.,
                          > >> to map a size based on some effort amount.
                          > >>
                          > >> This, as with all models, should be used with care and only
                          after a good
                          > >> calibration of the model with your own data.
                          > >>
                          > >> There is a product called Estimate Pro, now part of the
                          Borland CaliberRM
                          > >> product, that does all this and provides helpful estimations
                          for effort,
                          > >> duration, cost, staff, size, defects, etc. It uses COCOMO II,
                          Monte Carlo
                          > >> simulation and Putnam (SLIM).
                          > >>
                          > >> Also, due its integration with CaliberRM, you can inform your
                          effort
                          > >> directly in the requirements, and then have EstimatePro import
                          those data
                          > >> and present the results. Very fast and flexible!
                          > >>
                          > >> Hope this helps!
                          > >>
                          > >> Adail
                          > >>
                          > >>
                          > >> On 8/31/06, Rajendra Gupta <gupta.rajendra@ ...> wrote:
                          > >>>
                          > >>> Hi,
                          > >>>
                          > >>> Based on the exchange of emails, I have understood that wide
                          band delphi
                          > >>> (WBD) method satisfies CMMI L2 requirements if handled well.
                          I support this
                          > >>> view but have some queries based on some of the literature
                          that I have gone
                          > >>> through on this methodology:
                          > >>>
                          > >>> (a) While I have understood that it helps arriving at
                          reasonably
                          > >>> accurate effort estimates, I think WBD does not provide size
                          estimation, as
                          > >>> FP, LOC or COCOMO provides. Am I correct in my understanding?
                          Here each
                          > >>> task in WBS is given effort estimate.
                          > >>>
                          > >>> (b) Since it does not establish a size estimate, some metrics
                          such as
                          > >>> Productivity, Defect Density for the product cannot be
                          arrived at.
                          > >>>
                          > >>> (c) Since WBD requires iterative sessions, as every iteration
                          proceeds
                          > >>> there is greater likelihood that effort estimate will
                          continue to increase
                          > >>> rather than decrease. The risk is loosing customer due to
                          higher estimate
                          > >>> as compared to the customer. Need to get practical experience
                          from other
                          > >>> users of WBD.
                          > >>>
                          > >>> Thanks in advance on your inputs.
                          > >>>
                          > >>> Regards
                          > >>> Rajendra
                          > >>>
                          > >>>
                          > >>>
                          > >>>
                          > >>>
                          > >>>
                          > >>
                          > >>
                          > >
                          > >
                          >
                          > Please feel free to email with questions.
                          >
                          > Thanks and regards, Neil Potter
                          >
                          > The Process Group
                          > Email: neil@...
                          > Tel:972-418- 9541
                          > Fax:972-618- 6283
                          > http://www.processg roup.com
                          >

                        • Larry Fellows
                          Ross, Although this post was a direct question to Pat, which I m sure he ll answer, I thought I might add some to your discussion and help you with your
                          Message 12 of 19 , Sep 6, 2006
                          • 0 Attachment

                            Ross,

                             

                            Although this post was a direct question to Pat, which I’m sure he’ll answer, I thought I might add some to your discussion and help you with your apparent dilemma.

                             

                            Having used Delphi estimating in the past, the effort estimates do not come out of thin air.  The basis of the effort estimate has to be based on some concept of size.  It is clear that you don’t use typical “size” parameters like LOC or FP.  The answer to your problem may be to expand your concept of “size”. 

                             

                            When your estimating team comes up with an effort estimate for a task, some key provided them an idea of how large or small the task was so they could estimate the effort.  You claim an estimation method that is typically within 10% of actual effort.  This is way beyond guessing and tells me whatever your “key” is the relationship to effort is pretty well established.  The “key” can come from information your customer provides or from work you have done in the past.  Some potential ideas are requirements, features, or inputs and outputs.  Another possibility is similarity to previous projects.  If your estimates are based on the actual effort of a previous similar project than the previous actual effort is your size “key”.  If your projects are similar enough, you don’t have to break the previous projects down to size to get a factor you can multiply to get the new project’s effort estimate.  By the way similarity is a historical based estimating key and if your estimation accuracy should deteriorate as your software becomes more sophisticated or the projects become more dissimilar, the previous actual effort can be broken down to its components to determine a new key that will bring your estimates back into a range you find acceptable.

                             

                            So rather than saying you don’t use “size” to prepare your estimates, determine the “key” that the estimators use as the basis for their effort estimate.  That key is your “size” because it establishes the relationship task size and the effort required.

                             

                            Larry


                            From: cmmi_process_improvement@yahoogroups.com [mailto: cmmi_process_improvement@yahoogroups.com ] On Behalf Of Ross Timmerman
                            Sent: Wednesday, September 06, 2006 11:18 AM
                            To: cmmi_process_improvement@yahoogroups.com
                            Subject: Re: [CMMi Process Improvement] Size based estimation

                             

                            Patrick,

                            Can I infer from this that it is acceptable to use WBD to establish
                            effort estimates for a project directly without relating them to some
                            type of "size"? Based on this thread, which I started, I have been
                            working on trying to get a grip on estimation for my small software
                            organization, and am constantly getting tripped up by the issue of
                            size in the CMMi model. Our most successful estimation method was
                            not based on size, and was typically within 10% of the actual effort
                            for a project which was good enough for our business. For over a
                            year we have been trying to estabish an estimation model based on
                            size with absolutely no success. For CMMi L2, we had defined WBD to
                            estimate efforts for WBS tasks, but this was not acceptable, both
                            because the appraiser did not like WBD and because we weren't basing
                            our estimates on the size of something. If estimates based on the
                            size of something are an absolute must, I see no way that we will
                            ever make it to L2 - it's not worth it to us at the moment. SPICE
                            (ISO-15504) might be a better, more realistic option. I can't
                            imagine that the SEI wants to exclude small organizations, but they
                            are certainly doing a good job of it.

                            Ross

                            --- In cmmi_process_ improvement@ yahoogroups. com, " Patrick OToole "
                            <PACT.otoole@ ...> wrote:

                            >
                            >
                            > Rajendra/Neil,
                            >
                            > Stepping back from this a bit, the way that I typically see
                            Wideband Delphi used is when the project is estimating effort.
                            However, if done properly, with each iteration, the team will be
                            documenting their "estimation assumptions" - which REALLY turn out to
                            be the project attributes that serve as the basis of the effort
                            estimates. So, while trying to get their arms around the project's
                            effort estimate, they wind up discussing "size-like" parameters.
                            >
                            > That, to me, is one of the primary strengths of the Wideband Delphi
                            method - it pools the collective knowledge of the estimators,
                            discusses a much broader array of work and work product attributes,
                            and (hopefully) results in these being documented.
                            >
                            > If the organization is on their toes, they will develop a good
                            understanding of which attributes lead to more/less effort being
                            required, and will use that information to get better estimates on
                            future projects.
                            >
                            > I have always considered the Wideband Delphi method perfectly
                            appropriate for ML2 provided that the "basis of estimate" assumptions
                            are documented.
                            >
                            > Regards,
                            >
                            > Pat
                            >
                            >
                            >
                            >
                            > ----- Original Message -----
                            > From: Neil Potter
                            > To: cmmi_process_ improvement@ yahoogroups. com
                            > Sent: Monday, September 04, 2006 8:28 PM
                            > Subject: Re: [CMMi Process Improvement] Size based estimation
                            >
                            >
                            > Hi Rajendra
                            >
                            > > Thanks Neil and Adail.
                            > >
                            > > Refering to Neil's response, WBD (Wideband Delphi) can be used
                            to derive
                            > > either effort or size. If I use WBD to derive size, then I will
                            be able to
                            > > work out metrics such as Productivity. I have follow-up
                            questions:
                            > >
                            > > (a) If I need to know Productivity, can I still use WBD even
                            when I am using
                            > > it to derive effort and not size.
                            >
                            > You can use WBD just to estimate effort and not size.
                            > >
                            > > (b) While CMMI talks about 'attributes. ..', the older version
                            CMM clearly
                            > > mentioned size estimation. Does it mean that WBD (when used to
                            derive
                            > > effort and not size) will not work in CMM even though it works
                            fine in CMMI.
                            >
                            > Assume that cmm and cmmi are identical on this issue. They just
                            used
                            > different words in cmmi to explain the same thing. My guess is
                            that 'size'
                            > was read as 'Lines of code' by many readers, so 'work product
                            attributes' is
                            > a more generic term. But they are the same. So WBD works for any
                            estimation
                            > practice in any model
                            >
                            > Neil
                            >
                            > >
                            > > Regards
                            > > Rajendra
                            > >
                            > > On 8/31/06, Adail Retamal <adail.retamal@ ...> wrote:
                            > >>
                            > >> Rajendra,
                            > >>
                            > >> Some models use a (somewhat) polemic method known
                            as "Backfiring" , i.e.,
                            > >> to map a size based on some effort amount.
                            > >>
                            > >> This, as with all models, should be used with care and only
                            after a good
                            > >> calibration of the model with your own data.
                            > >>
                            > >> There is a product called Estimate Pro, now part of the
                            Borland CaliberRM
                            > >> product, that does all this and provides helpful estimations
                            for effort,
                            > >> duration, cost, staff, size, defects, etc. It uses COCOMO II,
                            Monte Carlo
                            > >> simulation and Putnam (SLIM).
                            > >>
                            > >> Also, due its integration with CaliberRM, you can inform your
                            effort
                            > >> directly in the requirements, and then have EstimatePro import
                            those data
                            > >> and present the results. Very fast and flexible!
                            > >>
                            > >> Hope this helps!
                            > >>
                            > >> Adail
                            > >>
                            > >>
                            > >> On 8/31/06, Rajendra Gupta <gupta.rajendra@ ...> wrote:
                            > >>>
                            > >>> Hi,
                            > >>>
                            > >>> Based on the exchange of emails, I have understood that wide
                            band delphi
                            > >>> (WBD) method satisfies CMMI L2 requirements if handled well.
                            I support this
                            > >>> view but have some queries based on some of the literature
                            that I have gone
                            > >>> through on this methodology:
                            > >>>
                            > >>> (a) While I have understood that it helps arriving at
                            reasonably
                            > >>> accurate effort estimates, I think WBD does not provide size
                            estimation, as
                            > >>> FP, LOC or COCOMO provides. Am I correct in my understanding?
                            Here each
                            > >>> task in WBS is given effort estimate.
                            > >>>
                            > >>> (b) Since it does not establish a size estimate, some metrics
                            such as
                            > >>> Productivity, Defect Density for the product cannot be
                            arrived at.
                            > >>>
                            > >>> (c) Since WBD requires iterative sessions, as every iteration
                            proceeds
                            > >>> there is greater likelihood that effort estimate will
                            continue to increase
                            > >>> rather than decrease. The risk is loosing customer due to
                            higher estimate
                            > >>> as compared to the customer. Need to get practical experience
                            from other
                            > >>> users of WBD.
                            > >>>
                            > >>> Thanks in advance on your inputs.
                            > >>>
                            > >>> Regards
                            > >>> Rajendra
                            > >>>
                            > >>>
                            > >>>
                            > >>>
                            > >>>
                            > >>>
                            > >>
                            > >>
                            > >
                            > >
                            >
                            > Please feel free to email with questions.
                            >
                            > Thanks and regards, Neil Potter
                            >
                            > The Process Group
                            > Email: neil@...
                            > Tel:972-418- 9541
                            > Fax:972-618- 6283
                            > http://www.processg roup.com
                            >

                          Your message has been successfully submitted and would be delivered to recipients shortly.