 
Senior Member
Earned Value Analysis I was taught that EVA allowed you to calculate your actuals versus your plan on tasks, in terms of effort. That allows you to determine that given current work rate, when your end date would be.
Unfortunately, MS Project thinks EVA is BCWP (budgeted cost of work performed) which deals only with costs. Anyone know how to work out (without the Excel spreadsheet I currently use!) EVA in terms of effort?
 
Senior Member
Re: Earned Value Analysis It should measure the actual amount of work that was accomplished regardless of the effort that was actually expended of rht time that was elapsed, although you should figure out cost and schedule variance. Basically, you should have factors for Budgeted Cost of Work Scheduled (BCWS), Budget At Completion (BAC), Budgeted Cost Of Work Performed (BCWP), and Actual Cost Of Work Performed (ACWP).
I like to base it on Overall Proportion Complete (OPC) and Estimate at Completion (EAC), which sort of match the ideas above. As you know, OPC should be the labor rate and the proportion of the work products put through the various activities.
So what I do is have p<sub>i</sub> = (units through activity<sub>i</sub>) / (total number of units required).
Then I have:
EV = Size * (SUM p<sub>i</sub> C<sub>i</sub>) * OPC
Here note that you obviously sum over the activities and note that C<sub>i</sub> is the labor rate for given activity i. OPC is the Overall Proportion Complete. (It is the addition of the labor rate that helps figure out effort. You can also work out productivity measures, which you can take as the inverse of the labor rate.) Also note you could do EAC as:
EAC = CTD + ETC
(CTD is Cost To Date and ETC is Estimated Cost to Complete.) Then ETC is given as:
ETC = Size * (SUM p<sub>i</sub> C<sub>i</sub>) * (1  OPC)
The final ETC is given in the units of effort used in the calculations. Generally, all costs will be stated in terms of units of effort that are contextually relevant to the company.
Either way you go, then these ideas can kind of speak to those equations we talked about here.
 
Senior Member
Re: Earned Value Analysis Oh, if it were only that simple! Perhaps I should explain what I am trying to achieve here. I am trying to work out from estimates and actuals what the real completion date would be if everything stays the same in terms of current productivity. Admittedly, I am working from incomplete data in that I do not know the size, except in terms of effort estimate. This is because I am dealing with a 3rd party on a fixed priced contract (so I don’t care about cost), hence the only information I am getting is an initial effort estimate, effort spent and effort to complete. I had significant doubts about the estimates given at the start of the contract and those doubts have proven to be well founded. As time goes on, estimates do not seem to be getting any better, so I am trying to set realistic expectations within the company as to a completion date. So, based on the information I do have, I can work out what the 3rd party believes the productivity rate is and I can work out (somewhat laboriously) what I believe it to be, and, therefore, the estimated time to complete.
It gets more entangled in that there are Change Requests coming from both parties which are being estimated (I believe) at an inflated effort (I suspect in an attempt to recover the losses for underestimating the original effort), so these will have a different productivity rate – probably.
Now, maybe it isn’t EVA I need here (maybe it is a crystal ball!), but it seemed like the closest value I could use to achieve my aim. And this comes back to one of the original points, that MS Project does not seem to help me!
 
Senior Member
Re: Earned Value Analysis <BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by peternairn: I am trying to work out from estimates and actuals what the real completion date would be if everything stays the same in terms of current productivity.<HR></BLOCKQUOTE>
Tricky because nonstochastic productivity is often a simplifying assumption that rarely holds up in reality, but it also depends on how you are measuring "productivity". What is your measure for that? Understand that what I gave above has a basis in what you are talking about. Or, at least, I think it does if I am understanding you correctly.
The Estimate At Completion (EAC) is an estimate of the completed cost of the project (and that can be in monetary terms as well as in terms of effort). Initially EAC will be equal to the Initial Estimated Cost (IEC). Then, when the project has started, EAC will be equal to the Cost To Date (CTD) plus the Estimated Cost To Complete (ETC). That is why I said: EAC = CTD + ETC.
<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>... hence the only information I am getting is an initial effort estimate, effort spent and effort to complete.<HR></BLOCKQUOTE>
So what you are saying is that you want to work out a putative date of completion given the estimates you have and other information that you are treating as "actual". You have the IEC. You are getting the actual effort spent (which can be measured, presumably, by labor rate), and the effort to complete, by which I assume you mean the "estimated effort to complete", which can be ETC (because cost can be stated in terms of units of effort). It sounds like you are getting the right information but doubting some of what you are getting.
<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>So, based on the information I do have, I can work out what the 3rd party believes the productivity rate is and I can work out (somewhat laboriously) what I believe it to be, and, therefore, the estimated time to complete.<HR></BLOCKQUOTE>
And are you worried that what you are coming up is not going to match reality because of the unknowns you are dealing with?
I mean, you can always take the modeled approach and do something like Schedule Time = 3.0 * (Effort)<sup>1/3</sup>, where "Schedule Time" would of course be the maximum estimated schedule time and thus give you your putative completion date, but bear in mind this is based on your measure of effort. What you can also do, since you say that "estimates do not seem to be getting any better" is look at the deviation and variance that you are seeing. Model backwards to find out what the estimate "should have been" based on where you are and you might see where the actual estimating is going wrong.
Unless I am completely misunderstanding you, your situation is somewhat similar to what I have dealt with in the past and used some of these metrics to deal with. As far as your change requests, I treat those as service requests to the system (the project) and I usually try to model those (like defects) at a certain estimated arrival rate. That can then factor into your labor rate.
Speaking to your one issue, though, I tend to agree about Microsoft Project. Personally I find very few programs that will do the degree of computation I like to do.
 
Senior Member
Re: Earned Value Analysis Thanks for all this, Jeff, it is really helping to get my mind round it.
<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by JeffNyman:
Tricky because nonstochastic productivity is often a simplifying assumption that rarely holds up in reality, but it also depends on how you are measuring "productivity". What is your measure for that? <HR></BLOCKQUOTE>
As you say, tricky. Given the information I have, I am using the rule of thumb that the first 50% of the effort achieves 33% of the activity (remember, I am not actually seeing anything as output to base things on). So, there is a Relative Productivity Factor depending on how far through the activity you are of 1.5 for the first 50% and .75 for the second 50%. (I know it is not really straight line, but it makes the calculations less complicated and is a rough approximation). Therefore if an activity has an IEC of 20 days, the CTD is 10 and the ETC is 15, I am performing the following calculation:
EAC = CTD + ETC = 25.
OPC = (CTD/EAC * 100)/1.5 = 27%
EPC (Estimated Proportion complete) = (CTD/IEC * 100)/1.5 = 33%
Productivity = OPC/EPC = 0.82
Which means that they are working at 82% of their predicted productivity.
My concern is that I think they are assuming that this 82% will increase to 100% for the remainder of the activity, which I don’t believe, so my calculation goes as follows:
ETC = ETC/Productivity = 18 days.
Therefore the EAC = 28.
So far, this calculation is proving to be more accurate.
<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by JeffNyman:
And are you worried that what you are coming up is not going to match reality because of the unknowns you are dealing with? <HR></BLOCKQUOTE>
Absolutely. The fact that my calculations are proving more accurate so far may just be pure luck (The 3rd party have no knowledge of my calculations).
<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by JeffNyman:
I mean, you can always take the modeled approach and do something like Schedule Time = 3.0 * (Effort)<sup>1/3</sup>, where "Schedule Time" would of course be the maximum estimated schedule time and thus give you your putative completion date, but bear in mind this is based on your measure of effort. What you can also do, since you say that "estimates do not seem to be getting any better" is look at the deviation and variance that you are seeing. Model backwards to find out what the estimate "should have been" based on where you are and you might see where the actual estimating is going wrong. <HR></BLOCKQUOTE>
That is, effectively, what I have been doing – modelling backwards – as a prediction of the future, it may not be good, however.
<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by JeffNyman:
Unless I am completely misunderstanding you, your situation is somewhat similar to what I have dealt with in the past and used some of these metrics to deal with. As far as your change requests, I treat those as service requests to the system (the project) and I usually try to model those (like defects) at a certain estimated arrival rate. That can then factor into your labor rate. <HR></BLOCKQUOTE>
Ah, here is a known! They will all arrive together in terms of work starting on them
<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by JeffNyman:
Speaking to your one issue, though, I tend to agree about Microsoft Project. Personally I find very few programs that will do the degree of computation I like to do. <HR></BLOCKQUOTE>
So, you have found at least one?
 
Senior Member
Re: Earned Value Analysis Yes, I could agree with the Productivity = OPC/EPC equation given some of the unknowns you are dealing with. Again, OPC is really going to be a summation of activies. So you have OPC = (C<sub>i</sub> P<sub>i</sub>) / (C<sub>i</sub>).
Remember: C<sub>i</sub> is the labor rate for activity i. P<sub>i</sub> is the proportion of software that has passed through each activity, as I described earlier. Obviously with "Productivity" you can take the limit of that as:
Productivity = Size / [(Effort / C)<sup>1/3</sup> * Time<sup>4/3</sup>]
(C is just a scaling factor.) Even if you do not know size, you can solve for its estimation:
Size = (Effort / C)<sup>1/3</sup> * Time<sup>4/3</sup>
(C, again, being a scaling factor.) I suppose you could just remove size from the initial equation as well and treat it as an unknown and base your estimate strictly on the effort and the time (i.e., matching persontime with calendar time) but then you run into the issue: relative to what?!? I also wonder why they feel their 82% will necessarily jump up to 100% or even why they assume it will necessarily increase? Granted, certain trends can indicate that (for example, work products slowing down in terms of need, defects slowing down in terms of find rate, change freeze put in place, etc).
<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by peternairn: Ah, here is a known! They will all arrive together in terms of work starting on them.<HR></BLOCKQUOTE>
They will?! So change requests all come in at the same time and all get worked on to the same degree (effort) immediately? (That is more why I actually asked about this because development of said changes plus the possible defects found plus the fix rate plus the retest rate could be the variables that skew the results.)
<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>So, you have found at least one?<HR></BLOCKQUOTE>
Actually, no  at least not in recent memory. I usually build spreadsheets based on the formulae that I like to use.

Posting Permissions
 You may not post new threads
 You may not post replies
 You may not post attachments
 You may not edit your posts

Forum Rules 