Quantity != Quality
-- Danny Kellett [email protected] On Thu, Jun 5, 2014, at 09:31 AM, Theo Fondse wrote: ** Hi Charlie! Although I have grown over the past few years to fully agree with LJ on this subject, I can also understand the need for metrics to have something to go by to know if our performance is on-par or not. Sadly, measurements in the LOC style no longer give a true picture of actual performance especially in terms of Remedy development. The company I am currently working for has a 100% custom Remedy solution, and are measuring performance based on number of requests closed per day irrespective of workflow object count (but they include all types of incidents, problems and change requests in this figure). In my opinion, this is a better performance metric than pure LOC count, but is also flawed because some types of requests are quicker and easier to close than others. Shawn Pierson nailed it very well in his mail, if the purpose of the exercise is to determine the "quality" of a Remedy developer or to guide decisions around which questions your metrics should answer if you want your company to keep the Remedy developer that will truly be most beneficial to keep when the time comes to make those "hard decisions". Dave Shellman also pointed out the efficiency of code argument. I would like to add to what he said by pointing out that the better Remedy Developer will: 1) Add config data to the system to configure it to do something rather than write superfluous/duplicate code. 2) Pre-allocate field ID's and share Filters/Active Links between multiple forms. These effectively lowers your LOC count and therefore LOC count does not paint a true picture of quality or quantity of performance in such cases. Server and network infrastructure performance also plays a role in developer performance. If you are working on a server that takes 38 seconds just to open an active link, you cannot be expected to churn hundreds of active links a day. Anyone will be able to (intentionally or unintentionally) exploit LOC-based metrics to their advantage by bloating their code, simply by: 1) Adding a number of extra Filters or Active links in stead of making something data configurable or generic. 2) Copy-Pasting Filters or Active Links rather than pre-allocating fieldID's and sharing workflow between forms. 3) Writing bloat-up Filters/Active Links that only seem to be doing something relevant, but has no consequence to the actual functionality. But, to answer your original question: If the company insistence remains on measuring performance on an LOC basis, and if 1) You are guaranteed to always be presented with complete, clear, signed off and sufficient requirement specification documentation, 2) You are guaranteed to have access to properly performing infrastructure (active link opens up in <3s), 3) You are focussed on doing development and are not required to stop what you are doing and attend to support issues, 4) You do not attend more than 2 hours worth of meetings or workshops a week, 4) Complexity of the solution is low to medium, 5) Good quality code is required. (As opposed to only evaluating if it seems to be doing what was required), 6) There are no external integrations to other systems where you do not have full admin access and responsibility, then my opinion is that, on an 8-hour working day, a good Remedy Developer should be able to produce anywhere between 15 and 100 objects a day counting a total combination of Forms, Active Links, Filters, Escalations, Guides, Applications, Web Services, Menus, and Flashboards. This is based on an approximate average of between ~5 to ~30 minutes time spent on average per object. If a Remedy developer is creating more than an average of 100 objects a day, then that developer is running the risk of probably not ensuring good quality code, because he/she is: 1) Copy-pasting code and not testing it the way it should be tested. 2) Unwittingly creating a bloated monster of a system that is going to be a costly nightmare to maintain. In such cases, one could start looking at: 1) Synchronising field ID's across al forms. 2) Writing more generic code that can rather be shared and/or data-driven. HTH. Best Regards, Theo On Wed, Jun 4, 2014 at 4:41 PM, Charlie Lotridge <[1][email protected]> wrote: ** Hi all, Thanks for all your responses. And, while I didn't get quite what I was looking for, it's certainly my own fault for not starting with the more narrow question I eventually posed. And even that I should have qualified by stating "assuming perfectly efficient workflow". I fully agree with all of the positions that the quantity of workflow varies significantly with the quality of that workflow, the complexity of the requirements, and many other factors. I also agree that in isolation, "workflow object count" is a useless number. I *do* think that as part of a broader set of measurable characteristics it can be used to say something useful about the developer, hopefully to be used constructively. But this is a conversation that is diverging significantly from what I was looking for. LJ, it's unfortunate that the poker point data was so misunderstood and misused, but I can only imagine that it must have been quite satisfying to the team that drove that point home with the 1000x formula. I'll take you up on your offer to take this offline. It might take me a while to put something together that makes sense, but please expect something within a day or so. Thanks, Charlie On Wed, Jun 4, 2014 at 7:05 AM, LJ LongWing <[2][email protected]> wrote: ** Charlie, I have a long standing hatred of performance metrics, that I won't go into the background for here, but I'll attempt to answer the basis of your question. Where I work currently, we went through an 'Agile transformation' a few years back. We all went through training on how to develop in an agile methodology, we discovered scrum masters, sprints, and all of the 'wonderfulness' of the agile methodology. During our grooming sessions we played Agile Poker ([3]http://en.wikipedia.org/wiki/Planning_poker) to estimate the level of effort of a given change. The 'points' assigned to the modification gave an indication of how hard the change would be, and a 'velocity' was set that said...ok, during this sprint we can handle '50' points of effort, with a sprint typically lasting 2 weeks, it would be agreed by all parties involved that the team could develop and test those 50 points in that 2 week period...it is typically assumed that given a general scrum team that the velocity can increase x% each sprint as the team gets into the groove. This process worked well for awhile until the 'metric' folks got a hold of these numbers. The metric folks said ok...well, we will start measuring teams on performance based on these 'points'. They started saying that this team was doing more work than that team because they were handling more points during a sprint...so one team started taking 3 0's onto the end of all of their points, they were then doing 1000 times more than any other team, and it became abundantly clear to the metrics folks that a 'point' system didn't determine how efficient a team was. Even within my scrum team our point values varied....if I was doing the work, I would assign the effort a 2 or 3...but if I knew that I wasn't going to the be the one doing the work, but instead, a junior member of the team, I would assign it a 5 or an 8 because they would need to do more research into the system to figure out how to get it done than I would because of my time on the team and knowledge of the inner workings of the app. The fact that myself and the junior member of the team might generate the same code, and I would do it faster, doesn't indicate that I'm better than them, nor necessarily more productive...just have more background than another. So...this long story is to say that every time I have ever encountered a performance metric that someone is trying to use to evaluate 'who is better'...I find that any metric that says 'lines of code per hour' or 'objects per day', etc don't show enough of the picture to properly evaluate someone. I instead prefer a metric that works on the whole environment/person instead. I prefer to look at 'how does the developer interpret requirements, does the developer ask any questions for clarification, how efficient is the workflow that is developed, how many defects come back on the code that is developed, etc. As others have pointed out....400 objects that don't work well are worse than 20 objects that work well. Other factors that determine a good developer are ability to communicate with team mates, ability to communicate with management, and ability to communicate with the customer. Some people are so 'heads down' that they might be able to program anything you want, but if you can't articulate your 'needs' to them in a way that they understand, and them get you what you are looking for back out of that...then they aren't a good developer in certain situations. I would be happy to take this offline with you if you would like...maybe get a bit more into your reasons for looking for this metric. On Tue, Jun 3, 2014 at 5:03 PM, Charlie Lotridge <[4][email protected]> wrote: ** LJ says 'performance metrics suck and don't work the way they are intended'. So, do you feel strongly about this? Yikes! ;) Really, though, while I didn't participate or even see any of those prior conversations about this subject, a couple points occur to me... First, while you're of course entitled to your opinion, I hope your blanket dismissal of the subject doesn't discourage others from voicing theirs. If the topic annoys you - and it seems to - my apologies. Not my intention. Second, I'd agree that "no one metric can accurately" say anything about anyone. My "one metric" examples were just given to spur the conversation. And perhaps others have more nuanced answers that involve more than one metric and include qualifications. I'd be interested in hearing about those. As a software engineer (my background), one of the metrics that has been used to judge my work has been "lines of code". In and of itself it's not a useful metric, but combine with other factors it can help provide a broad picture of the performance of different developers. Third, having such data doesn't make it bad or "wrong" data, it depends on how the data is used just like any other data. If used constructively, such metrics could, for example, be used to help assess a developer's strengths and weaknesses with perhaps the goal of working/educating the developer to shore up those weaknesses. And while it's certainly true that information like this can be misused, it doesn't mean we shouldn't have the conversation. Fourth, there ARE clear differences in the performance of different developers. Sometimes there are very valid reasons to judge the relative performance of developers. Sometimes it's because hard choices have to be made like downsizing. Is it better in these situations for the manager to just pick the individual(s) they like the least? Or who they *think* are the least productive? I smell a lawsuit. Wouldn't hard metrics be useful in these cases? Finally, a disclaimer: I don't now or have any near future plans to use such metrics to evaluate anyone...I don't have anyone to evaluate. And while my interest in the topic is more than just idle curiosity, I won't be using it to fire anyone soon. For me, this information is more for research purposes. Thanks, Charlie On Tue, Jun 3, 2014 at 3:03 PM, LJ LongWing <[5][email protected]> wrote: ** My opinion is that 'performance metrics suck and don't work the way they are intended'. There has been healthy debate over the years regarding exactly that subject, and every time it's happened, either on the list or otherwise, it ends up being that no one 'metric' can accurately say that this developer is doing 'better' than another developer. On Tue, Jun 3, 2014 at 3:46 PM, Charlie Lotridge <[6][email protected]> wrote: ** Hi all, I'm curious...what are your opinions about what might be useful metrics to use to judge the performance of Remedy developers? To narrow the conversation a bit, let's just talk about during the creation of a new custom application, or custom module to an existing application. In other words for code generation. So for example, you might tell me that a good developer can create at least 50 logic objects (active links/filters/escalations) in a day. Or create & format one form/day. What are you opinions? Thanks, Charlie _ARSlist: "Where the Answers Are" and have been for 20 years_ _ARSlist: "Where the Answers Are" and have been for 20 years_ _ARSlist: "Where the Answers Are" and have been for 20 years_ _ARSlist: "Where the Answers Are" and have been for 20 years_ _ARSlist: "Where the Answers Are" and have been for 20 years_ _ARSlist: "Where the Answers Are" and have been for 20 years_ References 1. mailto:[email protected] 2. mailto:[email protected] 3. http://en.wikipedia.org/wiki/Planning_poker 4. mailto:[email protected] 5. mailto:[email protected] 6. mailto:[email protected] _______________________________________________________________________________ UNSUBSCRIBE or access ARSlist Archives at www.arslist.org "Where the Answers Are, and have been for 20 years"
