Work habits and conditions vary dramatically between software development groups.
At one extreme of the spectrum are productive, tuned development teams delivering solutions on schedule and on budget, staffed with passionate experts happily building their skills and careers while providing valuable cutting-edge solutions. At the other extreme lie dysfunctional, under-utilized teams. The latter group endlessly catapults magnitudes past estimates of time and money, with a revolving door roster of contemptuously-treated employees.
Most shops fall somewhere in-between — imperfect, but continually working towards better, more efficient practices and process, all while trying to provide a rewarding, fulfilling, and career-building experience for team members.
This entry is targeted at that continually-optimizing audience, and is written largely for development group managers and team/group leaders, though it’s still informative for the general development community (lateral and upward management is critically important, and is the force that drives most process change).
These observations are gleaned from my experience with teams of varying sizes and types, acting in a variety of roles including management, team lead, developer, process consultant, mentor, mentee, and technology consultant. I have directly participated in, or observed, the development process in small, tightly-knit engineering groups stocked with professional electrical engineers, to mega corporations with walking-dead teams awaiting the next executive shuffle to mete out some doom on their orphaned projects.
I’ve learned from the successes, but I learned even more from the failures.
I’ve focused on best-practices that are optimal in virtually any team, and which are frequently evident problems, while avoiding those that are more likely to vary significantly between teams and organizations. Work environment, for example, is very subjective and context specific: Some developers and teams work optimally in private, quiet offices with comfy chairs and scheduled communications, while others thrive in hectic, noisy open offices, interfacing directly with the
“customer” throughout the day, eagerly setting up development kiosks on whatever can serve as a seat. Some organizations need armies of BAs and BSAs and QAs and UAs, with layers upon layers of heavy process and checks built atop a traditional waterfall development model, while other shops work best with a highly agile, frequent delivery model, completed by multi-role development resources. It isn’t possible to declare universal best-practices within those domains.
I also appreciate that most real-world teams have budgets, and that feel-good lists of .COM-style excess aren’t realistic or helpful in the current IT climate. Every developer doesn’t need every tool available, or every electronic toy on the market, to do their job effectively. I haven’t tried to pander to the development
community, blowing smoke about how entitled and empowered they should be, beyond the level that is actually beneficial to their career and the role.
On to the list!
Optimizing Software Development – Executive Summary
- Customize Your Development Process – Choose standards and processes that fit your team
- Use Source Control, And Use It Properly– Source control is a requisite of good development
- Formulate Written Standards– Don’t rely on by-example or word of mouth standards
- Mandate Regular Status Updates– Have daily status updates
- Organize Your Information– Standardize your information archive structure
- Allocate Appropriate Time For Research– Research appropriately, or enjoy wasted effort and redundancy
- Accurately Track Development Time– Build valid historical data, and time accountability
- Focus On Your Competencies– Now when to outsource tasks to stay focused on your competencies
- Focus On Results Instead of Effort and Sacrifice – Sacrifice and effort should only matter when they yield results
Select processes, practices, and standards based upon your particular type of organization, type of project, and workforce, and then customize them for your specific needs.
Of course this suggestion seems ridiculously obvious, yet this is an industry rife with groups foolishly mimicking the process of those who appear to be successful.
Whether it’s the previously agile vertical-solution shop trying to act like they’re a bank, overloaded with process and CMM levels far beyond necessity because that looks like what the “big boys” do, or the banking group trying to adopt XP practices and frequent deliveries because it’s the hip new trend that all the blogs are talking about, when their product only requires bi-yearly deliveries under a heavy layer of checks and balances: What works for one type of organization doesn’t necessarily work for another, and adopting those processes under the wrong environment will bring defeat rather than success.
Don’t act like you’re developing life-critical flight control software when you make a P2P app, just as you shouldn’t do the opposite. Don’t pretend that you’re an ISV, or assume that much of what’s right for a retail software vendor is transferrable, if you’re actually an in-house IT development shop.
This sort of cargo-cult mentality exists at a lower level as well. During the .COM boom, success apparently came from stocking up on Aeron chairs — Get the chairs, and you’ll do well. After the collapse the idiocy continued, with the same chairs now being demonized: Now they could only bring doom, and had
to be removed post haste. It’s a chair people. In the grand scheme of things, it really isn’t going to make that much of a difference, positive or negative.
The same could be said about foosball tables; catered lunches; big, corporate-style meetings; having big cozy offices for everyone, or tiny open plan offices for everyone. Someone seemingly did it to their success, so therefore it was perceived as an applicable practice for everyone.
Just because someone else benefited from it doesn’t mean that it would benefit your organization. It could very well hurt your organization.
Source control is the foundation of good coding practices.
Not only does source control keep your code centralized (facilitating organized access for developers, along with easy backups and automated integrity assurances), with the proper use of branching and labels it can greatly increase the agility of your team: Having the comfort of making major changes in a code branch, while retaining the ability to revert to a labeled release or trunk, is liberating, reducing developer paranoia and ameliorating the risk of change.
Whether you’re using Microsoft Visual Sourcesafe, Perforce, CVS, or one of the countless other available source control products, it is a critical component in your toolset. It’s critical even if you’re a single developer working alone in your one-man software
Source control is also invaluable for change management and auditing, historically tracking how much has changed, and where, over a given period of time, which is a benefit that often pays off handsomely years later. For example auditing the code history to determine when an errant algorithm was introduced to gauge how long it’s been screwing up the calculations.
Adequately trained quality assurance teams can even use source control to determine what has changed between releases to tightly focus their testing efforts without relying on the often incomplete word of the developers.
The technique to use source control properly depends upon the tool that you’re using, and the standards and needs of the group: Some groups use source control only to check-in production releases, while other groups check code in frequently, with every developer checking in multiple times per day, merging branches at convergence checkpoints. Ensure that you know the feature-set of your source control package inside and out, and that you’re following a rational, best-benefit standard of use for your scenario.
If your usage is non-optimal because of limitations of your SCM tool, consider a product change. It really is that important.
Tip: Frequently perform project builds on a clean machine with limited network access by following development environment configuration instructions, and then getting and building from source control. During this process many teams discover hard coded paths, missing dependencies, and solution defects. It’s better to detect it early than for development to grind to a halt the day before a big release when the corporate IT department decommissions that old server that supposedly wasn’t in use for the past year.
Standards can be extraordinarily detailed, or as simple as noting which industry standards a team subscribes to (e.g. “All .NET coding will be done in conformance with the Microsoft .NET Design Guidelines“). It could even be the formalization of the lack of specific guidelines in a certain realm (e.g. “Put the braces wherever you’d like, and either tabs or spaces are fine“). Documented standards should cover all aspects of development, including development environment, tools, accepted languages, best-practices, check-in behaviour, naming guidelines, and so on.
Of course standards should be living documents, with occasional reassessments leading to the removal and addition of new points.
If, on the other hand, your standards are word-of-mouth, or “by example” (usually with many conflicting examples throughout the organization), you don’t really ave standards — more likely it’s acting as a demotivating, inefficient confusion for newer developers to the team, and a constant political skirmish
between the veterans of the group: While good developers will happily adapt and conform to the idiosyncrasies and preferences of a group or project, it is frustrating and unproductive when they have to constantly go back and rename tables and rework classes because of unwritten “standards” that exist only in the minds of a few. When developers lose confidence in their code because of uncertainty over trivia such as spacing standards, something is seriously wrong, and people are
focusing their care and concern on something that should be thoughtless.
As an added benefit, the process of documenting standards is often a time for rational debate, when “that’s the way we’ve always done it” standards have to be legitimately rationalized and justified, or rightfully discarded. Many teams have improved their process, finally jettisoning unjustified legacy standards, during just such an exercise.
Having concise, documented standards (which can be the documented absence of standards in a particular realm) saves everyone from having doubts about whether they’re conforming, eases new developers into the fold, and increases the percentage of good standards while removing the excess baggage. It also facilitates external assistance more smoothly.
Every project, and every team, should have a daily or bi-weekly status updates by every member to their team leader or manager, or better still to every other member of the team (including from the manager to the team). This is regardless of your development methodology.
This can be a scrum, or a simple daily email stating what work the person is planning for the
day, and what they achieved during the past day/period. This practice not only keeps everyone focused on the task at hand, with better time accountability (just like everyone else, and possibly worse as a group, developers are procrastinators), but it also gives a transparency of development that let’s the manager, leads, team members, and stakeholders see where impediments exist, providing them an opportunity to help, or to recognize and respond to schedule slip early on, before it becomes a crisis. This transparency is one of the fundamental goals of the CMM process, and the benefits can be largely gained just by frequent, scheduled, mandatory and accountable communications.
Optimally this information will also exist in an archived, centralized, widely accessible location, such as an
audio recording of a scrum, or a centralized collection of update notes.
It is critically important not only that this information is conveyed, but that it’s actually consumed and retained as well — management and the rest of the team must pay it proper heed, and they shouldn’t wallow over during the day to ask what was previously stated. That undermines the entire process when it turns into a wasted effort that no-one pays attention to.
Having lots of specifications, standards, code libraries, discussions, solutions, and examples strewn across the topology of the network is of little use to anyone, and can be detrimental. Outdated documents will continually reappear, and misinformation takes on an unstoppable life of its own.
Centralize and standardize, ensuring that everyone knows exactly where to find all pertinent information on a moment’s notice. Everyone should know exactly where they can find, and thus where they can store, specific types of information.
As it is, many teams start off with the “dumping bin”-style repository when they begin a project, planning on sorting it out later, however soon enough there are GBs of files — many of which have little or nothing to do with the product or project — confusingly cluttering a massive mess of a directory structure. This puts a serious damper on productivity.
Documents can’t be found. People have to constantly interrupt each other to ask where to find something or where to store something. Eventually an expensive initiative will have to be undertaken to separate the wheat from the chaff, at a time when most people have forgotten why all of the various unrelated information is there in the first place. Undoubtedly good data will be removed, and bad data will be retained.
Before you create the first project or product file, decide exactly how you’re going to organize the information, planning for the long term. Who will have access to it, and what permissions will they have How they will access it What will be contained where?
Tip: Documents such as specifications yield the same benefits of source control.
The first step of any project shouldn’t be to hit the IDE, or even to start high level design work. The first step should be to research and understand all competing or similar products, and any and all libraries and tools that could be leveraged to empower your solution.
As I write this, countless teams are spinning their wheels reinventing the Microsoft Application Blocks, or they’re developing a data transformation service that could be integrated in Biztalk or SQL Server Integration Services less expensively (and much more flexibly). Perhaps they’re developing a web portal that’s just an incomplete, second-rate clone of SharePoint, Community Server or Zope.
There are even teams duplicating functionality already existing in the .NET Framework or standard template library because they never spent the time to look.
For virtually any need there are libraries and existing solutions that could either replace the need for the project altogether, or more likely could serve as a foundation of a more robust, more capable solution. If you’re in the solution business, the desire to propose and implement the best solution should trump any desire to code for the heck of it.
Developers can’t complete that sort of competitive research in an afternoon, or as a footnote, nor should you rely upon off-the-cuff commentary provided in meetings by individuals vaguely aware of the options.
While it varies by project and industry, such research usually requires a significant upfront investment of time before coding begins, allowing the stakeholders and development group to all feel entirely confident that the project is justified and will yield the best possible results. Document every finding and
decision to save endlessly having to explain why one decided to develop what seems to be a clone of Biztalk.
One of the primary reasons why many shops continue to grossly under-estimate the time required for projects is that they have limited, or fictional, historical real-world time tracking to base future projects upon — they might have half-a-dozen developers working full-time, but the best they can
account for at the end of the week is that a couple of small app issues got worked on.
This is sadly very typical.
Where there is some form of time tracking system in place, often it is of little value: Developers, as a general observation, will try to find ways to under-report the amount of time they spent on a particular development problem or project, as it’s a bit of a badge to claim to have “thrown together”
complex solutions in a “couple of hours“. Furthermore few shops properly partition times (e.g. by project stage and task), further blurring the value.
Implement a time tracking system, and coach your team to accurately and regularly post times (daily, if not more frequently. Any less frequent and accuracy suffers dramatically (as someone with the bad habit of filling out timesheets at the end of the month, anxiously looking at file timestamps to figure out what I worked on when, I say this from personal experience).
Ensure that there is enough granularity that the time reporting is usable. For instance that you can see that competitor research actually took two weeks, instead of the foolishly anticipated 2 hours.
Avoid tacitly encouraging false timesheet entries: Don’t act concerned when real world values start appearing, and most tasks are much more time-intensive than originally thought. Adapt and accept that reality differs greatly from the mirage that most people imagine that they see.
You have your team of six hyper-intelligent and eager software developers, overloaded creating the next great scrum-tracking application. Based upon some user feedback, you realize that a small percentage of your users may want to interact with the system in a minimal way using J2ME equipped cell-phones, but your team is entirely staffed with .NET experts. Do you a) add a J2ME developer to the team, and just try to expand the J2ME product line, b) let one of your existing developers try their hand at J2ME, c) outsource that non-core component. Most shops would pick b without hesitation, incorrectly justifying it under the guise that it’s “free” because it’s covered by the payroll.
If you decide to try your hand at it in-house — despite the considerable domain difference — the results will almost certainly be a second-rate solution, delivered more slowly and more expensively than just outsourcing the need. This is worsened in that a resource originally dedicated to the core product has been diverted to a non-core need, and will come back to the core product with a need to spend time reacclimating.
This isn’t to say that this is always a bad solution: Sometimes it entirely makes sense to build such skills in the core team (if this is a component that will need regular modification, or if it’s an expanding customer base), not to mention that the work diversion can prove entertaining and rewarding for a developer.
Undertake non-core competencies — be it building a one-off data bridge to a partner, putting up a corporate website, designing logos and icons, or countless other non-core needs — with eyes wide open, realistically assessing the cost to do it in-house. Estimation of the “knowns” is
already notoriously bad in software development, but the situation is far worse for non-core competencies, where developers are prone to extraordinary underestimations.
Realistically consider leveraging external solution providers (or even other teams within the organization) where it can keep your group focused — with appropriate communications, teams will understand that it enhances their value and effectiveness, rather than presenting a risk to their employment. [Note: My organization provides outsourced consulting and development, so I do have a conflict of interests. However we always go in with the
intention of empowering and furthering the existing team, and never to supplant them]
Getting the team working on nights and weekends means little if it saps their passion to produce, expensively inflates the bug count, and increases turnover. Yet sadly a lot of managers work under the delusion that effort and sacrifice are worthy substitutes for results. Indeed, many management techniques are built entirely around deriving maximal effort and demanding complete sacrifice, inventing artificial deadlines and manufactured crises to prod the troops.
The results are invariable destructive and counter-productive.
This is rooted in basic human psychology. Just look at a famous cough syrup with a heavy advertising presence: Surely it must work better because of the sacrifice required to ingest it Contrast that with the presumption of self-destruction that most attribute to activities that people find enjoyable.
If it’s fun or requires little sacrifice, then it can’t be beneficial. If it is unpleasant, then it must be doing good. These foolish and unfounded notions have sold a lot of snake oils over the years, and they’ve supported a lot of human folly.
Savvy Machiavellian developers know how to exploit this management myopia, under-delivering until crunch time, and then putting in some superficial extra effort to get the project done, yielding kudos all around for going “above and beyond”. Just by finally delivering long overdue results that they could have achieved in the normal workday. This is especially the case in shops with limited time reporting, and no regular status updates: Show up early in the morning (sending emails as validation), grunt and groan and gripe, shuffle around regularly, and then leave late (again sending emails to validate how committed one is), maybe VPNing in during the middle of the night to again demonstrate great sacrifice.
In some shops that alone is a substitute for any results at all.
On the flip-side, the “slacker” that comes in at 10am, leaves at 4pm, and takes extended lunches somehow keeps generating the bulk of the product design and code, but his apparent lack of effort and sacrifice will be dealt with at the next performance review.
Many heavy and onerous processes are undertaken, and then maintained, not because they’ve empirically proven themselves to be useful, but rather because the effort involved gives a mentally cheap illusion of achievement. From volumes of documents that no one ever references or validates, to arduous signature gathering exercises. Most of the time they’re the mindless completion of a task with no real benefit to the project or the process, but it goes unquestioned as a part of the flow.
Discard all processes that are mechanically completed with little actual benefit. Tweak those that require more effort than necessary, maximizing the results while minimizing the effort.
Effort alone doesn’t make your quality better, software better, or team smarter. Focus on results, and welcome and encourage results that come with minimal apparent effort or sacrifice.
Building a top-notch, effective software development team is a lot of hard work, coupled with a bit of luck, but hopefully some of these points have given a bit of food for thought that might encourage some development in the right direction. This list most certainly isn’t comprehensive, and I’ve stayed away from domain-specific practices, but it is a start.