My latest news article on InfoQ -> http://www.infoq.com/news/2011/11/project-managers-problem
Business analysts, requirements managers, and project managers will find the greatest interest in this article.
Software Product Planning
It occurred to me last week during one of our weekly iteration planning sessions that one of the most esoteric methods around product planning is deciding which requirements to turn into software features. The more rigorous approaches look at the cost of the feature, the potential impact to ROI ( assuming there is one ), and the demand.
What’s wrong with this approach is that it considers the features and resulting requirements in isolation from one another. By not considering how each new feature affects the existing product as a whole teams can and do end up with products in which the original feature set, that made the software successful, become diluted. Those of you who’ve worked with me know my favorite example is CA’s Remedy product, but I think one could find other examples: the Microsoft Office suite of products may be in this camp.
Feature Dilution: A Formula
So how would one go about constructing a measurement for feature dilution? First – some assumptions:
- You know or can retrieve the cost associated with the original marketable software release.
- You know or can calculate the benefit ( ROI ) for the original marketable software release.
- You have an estimated cost and benefit associated with any potential new software features.
Ok, so knowing these let’s construct a model for software feature dilution. We’ll adapt a formula from the world of finance.
V – Value of sofware after Feature dilution =
((O x OP) +(N x (∑ IP1, IP2….IPn))) / (O + N)
O = original number of features
OP = Current NPV of product ( could use ROI too )
N = number of new features to be added
IP1, IP2, IPn = NPV of each new feature.
If you run this formula through some examples in time what you’ll find is that as a product matures new features need to continually generate greater returns to justify value to the original product and ultimately diluting the existing feature set.
This is exactly what should happen if we want to avoid the fate of an overly complex and unmanageable software product. Just like stock market share dilution the product management team needs to justify that further feature dilution will grow the value of the product in terms of existing functionality…..not just that it will add to revenue.
Simplicity in software design has always been something great software architects knew yielded great products. With this formula I hope I have provided at least a start to measuring simplicity in software.
Who is this article for?
All software development professionals will find interest in this article, but managers, CIOs and software architects will find the greatest interest. The topic may be controversial to many, but I offer this article as insight into what seems to be a growing problem in the Agile movement.
“Why are you here? Agile doesn’t need managers.”
Ever hear this one before? Imagine how shocking it is to hear that the developers think your position shouldn’t exist……as if you as the manager had some contribution in creating that position. It’s most commonly directed at project managers as they first meet the development team they’ll be working with. To be sure the original Agile Manifesto makes absolutely no mention of project management and subsequent agile theorists go further and suggest adjusting the project manager role to be more of a coach or support role.
However, this view ignores reality.
Small non-integration dependent development projects, to be sure, probably require very little supervision of any kind as long as you have a competent, experienced and capable team. However, the larger the project, the more integration dependent the project, and the less development centered the project…….the more a project manager is needed to coordinate, communicate and lead the overall effort. A project in which the development portion is only 10% of the overall budget can allow a scrum master to lead the development while reporting to the project manager.
Furthermore the development team is almost never aware or good at managing a budget. The amount of time required to develop software requires that little time be spent on anything else. This creates a bit of a blind spot for some developers as they begin to believe that everything they are doing *IS* the project and that anyone else is just a peripheral annoyance.
The bad attitude here is the inability to recognize other roles and professions as having value and strictly adhering to a philosophical interpretation without recognizing the need for flexibility given the winds of reality. If taken too far the attitude can come across as almost unionist or neo-communist in its presentation by extending the view to all management in all situations. Surely the individual who adopts such an all-encompassing wholesale reduction of the corporate culture and organizational structure into one flat level is a radical. His views are on the periphery, but if he’s the right person ( a leader ) his views can gain traction and worsen the relationship between the development team and management; turning the goal of project completion into a class warfare between management and workers.
“The team runs the project, not the managers…….we’ll decide what gets done.”
This view is often an outgrowth of the notion that management roles are no longer needed. It flies in the face of the truth; that many decisions require collaboration among many elements of the company; not just the development team……..including the software design and architecture.
In other instances developers positing this notion are just unaware that there are other aspects to a project. Or even worse a developer has been burned terribly by a bad experience and feels the need to “take control” of the project before some perceived breakdown occurs.
Regardless agile becomes the pre-text, a foundation, for an attitude which suggests that most, if not all, the management structure above the development team has no contribution and should be summarily removed from contribution to the effort. Letting this attitude take hold, in my experience, usually results in endless re-architecture sessions, severe budget-overruns, no real end date, and a fractured emotional team that becomes disillusioned with its own mission.
“There are no due dates or schedules in Agile.”
Those of us with deep insight into capital budgets and corporate finance know how silly this is. However, if you read Ken Schwaber’s Scrum book it does talk of abandoning the Gantt chart for a burn down chart. In truth the burn down chart is a neat and well thought out innovation, but there are those who take this to mean that there is no schedule for delivery……i.e. the money never runs out.
This was a painful experience for myself. I watched as a team led by a strong, charismatic technical leader ,whom we all reported to, abandoned any time based goals in favor of just producing a “working product” for the customer. Without any time boundaries the team careened every which way. Work ethics declined or were non-existent. Those who wanted the product to succeed lost any motivation and drive. The customers became bewildered as to why so much emphasis was being placed on various technical architectures while features and product change requests became lost. The burndown charts further confused them. All they really wanted to know was; when will the product be complete? The team would only respond with; “We’re not on a schedule. We keep developing until we’re done.”
When attempts were made to set realistic goals by anyone; they would immediately get knocked down as “anti-agile”. When the team was informed that their project was hopelessly over budget; their eyes looked clouded, and confused. The connection between what they were doing, time and, ultimately, money had become lost in the abstract design patterns etched on their whiteboards.
Realistically…….there is always a due date and a schedule for delivery; explicit or implicit. No one puts up money for a development project with the view that it will never complete. Even more realistically, I’ve found that Gantt charts are still very useful for coordinating deep integration or non development aspects between non-agile teams and agile teams.
The no schedule attitude arises mostly because agile techniques put forth the notion that the project should continue to add new features until the money runs out. This is ideal and ignores what happens when a development team hasn’t even completed the bare minimum of functionality within the budget; rendering the application useless. The bad part of this attitude is taking a new technique for tracking team progress and accountability and twisting it into a reason for not being accountable for delivery.
“Agile code is self documenting. There’s no need for requirements, architecture diagrams or technical specifications.”
If you are a software architect or technical manager this attitude is usually targeting you between the eyes. The thinly veiled attack is meant to question your role, experience, and the need to have anyone coordinating the overall technical design of that 28 million line software program that generates 78% of the company’s revenues.
Certainly it is often put forth by ignorance. Maybe the 2000 line web app that the developer built recently required very few artifacts beyond the source code, but scale matters. You know that, your management knows that, but this bad agile attitude chalks up your role to not staying current on development techniques like Scrum. Major software systems require that a few minds are overseeing the direction and coordination of the technical vision and the many hundreds of hands creating it.
In my own experience this attitude came from a developer who, ironically, wanted to be one of the architecture staff. He felt by critiquing and arguing with the technical leaders and introducing his knowledge of agile techniques they would respect him more and give him the coveted position he craved. Instead, they found him to be annoying and a troublemaker. Furthermore his lack of tact in introducing agile concepts left the senior technical leaders with a bad taste in their mouths for anything agile.
“Agile rapidly embraces change; all change.”
My experience with this attitude came from a manager instead of a developer. It turned out he read “rapidly embrace change” to mean all kinds of changes……not just business requirements as was intended by the original agile creators. So, fundamental architectural changes became commonplace and shifting between different open source technologies was seen as ‘good’ even though this meant taking the team completely away from their skill sets and setting project delivery back by months. Organizational experimentation and rapidly dropping people in and out of roles also became part of ’rapidly embracing change’. The end result was a mess.
Clearly accepting change presented by customers is important, but without a system for managing that change; you’re asking for trouble. One needs to keep track of all requirements and changes and their impact to project delivery so that this can be communicated to customers. This is necessary to make effective project decisions. If you don’t then the customers get the unrealistic notion that anything they ask for will be included……we know where this leads.
So the bad attitude here is accepting change without managing it. An unmitigated free for all will only lead to dashed hopes and unmet expectations. Change is good, but violent change is chaos.
“Agile uses generalists; we test our own software. There’s no need for a QA group.”
Again this view is accurate in philosophical interpretation but my experience with this on, especially large, software development projects is…….you need a 2nd set of eyes looking at what the developers created and how well it works. Pride of workmanship is great and should be fostered, but sometimes pride can turn into blind acceptance and defensiveness. It takes a strong and deeply honest person to recognize their limitations and find ways to mitigate them.
Using generalists puts emphasis on making sure you’re staffed with a nimble group of multi-skilled individuals. In reflection this recognizes software development as mostly craft and less production assembly. However, as software development leaders we can’t assume perfection in human resources and ignore the facts. It’s better to see the risks and plan for them and history has proven that developers don’t find all their own mistakes.
In my own experience the individuals holding this view disliked anyone testing their code and were prickly to any constructive criticism. In one case in particular we found the underlying reason was because the developer really wasn’t that good at coding. He was given training and mentoring and after many months of struggle it became clear that he was on the wrong career path.
So using generalists is fine, but the attitude becomes stale if the hard truths of decades past are ignored in favor of philosophical purity.
In conclusion these problems could be found in the pre-agile world as well. But in my experience these bad attitudes are finding refuge and justification in a new technique that in isolation probably never intended to present such a soapbox. As software development leaders it’s critical that we address these viewpoints before they take hold of the agile methodology and potentially darken a good movement. Agile has a great message; simplify, engage the customer during the product development, take ownership, and stay connected. It would be a great disappointment to see this message lost. So what do you think? Are these attitudes in your shop? How do you address them? I’d like to hear from you.
About the Author
Christopher R. Goldsbury is a software development professional who has played the roles of developer, architect, scrum master, development manager, project manager and quality assurance manager throughout his career. Chris writes on his experiences and ideas at his blog: http://www.anagilestory.com.
Who is this article is for?
This article is written for those with management and budgetary responsibilities for a software development project or team. Others, including developers, quality assurance personnel, and CEOs/CIOs may find interest.
Why would we need to estimate story point cost?
Story points are used to estimate work. Investment in that work is expected to derive some benefit. If that benefit is expected to be financial then understanding the cost of that work is essential to deriving any meaningful ROI. Even if no ROI is expected and the intended benefit is regulatory compliance ( as an example ) then company leadership usually wants to understand what how much of their limited financial resources is going towards any specific feature, iteration, or release.
How do we do it?
The technique presented here is a historical parametric approach. It relies on past data from previous projects. So, one has to have some of this data saved up before a reliable figure can be derived.
RC = Total dollar cost for a historical releases in a product
RSP = Total story points that contributed to that release.
RSPC = Release Story Point Cost
RSPC = RC/RSP
Once you have this for one release you should calculate it for all historical releases. The next calculation is an average:
Average RSPC per product = ∑ RSPC¹, RSPC²……..RSPCⁿ / N
If you want the story point cost across all products then average it again. Although, for most planning purposes it’s useful to plan by product line and this higher level of abstraction of cost might be too watered down.
What questions does this help answer?
- How much will it cost to add this feature?
- How much will it cost to deliver release 2.1.0 ?
- What is the cost of an average iteration?
How often should it be updated?
The astute among you will notice that we’re using historical data. Historical data is only accurate as long as change doesn’t take place. To counteract the shift and change in time size, capability, and mix one needs to do these calculations at regular intervals. How often? This is a judgement call. I do it monthly as I’m in rapidly growing team with many new products popping up. I constantly need to reassess my cost driver.
A more stable team and product might require only 6 month intervals. The relevant point here is; keep it accurate.
Story point cost ties a rather abstract and developer centered concept to the real world of business. This is necessary. If we intend to use story points in a meaningful fashion in our development environments than they must have some corollary to the spreadsheets, and ledgers that the world’s businesses run on.
- commitment to complete the stories and tasks within the iteration time frame.
- commitment to *try* and complete the stories and tasks within the iteration time frame, but you’ll probably need to move some to the next iteration.
- Team velocity is 40 story points.
- Iteration 1 the team commits to 35 story points. An easy iteration for them.
- Iteration 2 the team commits to 45 story points. Not easy, not hard.
- Iteration 3 the team commits to 55 story points. This is a stretch. They might make it ; they might not.
- Iteration 4 the go back to committing to 35 and repeat the stair step again.
How often do you hear that a company attempting to adopt agile practices fails? This article attempts to examine and explain the often overlooked key organizational reasons that agile fails, why it isn’t obvious to most of us and some potential strategies for coping with organizational impediments. The article’s target audience is managers with budgetary responsibility although technical groups might also find interest.
Historical Perspective on Agile
Where did agile practices originate and why? The Agile Manifesto was originated by a group of software developers. Their main pupose in creating a new software development methodology was to address some of the core problems with traditional waterfall techniques, specifically: risk around changing requirements, late feedback from quality assurance, and accountability of the development staff. Their focus was not on how this methodology worked with the budgeting and financial aspects of a funded development effort. In the information technology world there are two types of funding models. One is a large company that sees the software product as its bread and butter or at least a key differentiator for its business model ( think Oracle, or Scotttrade ). We’ll call this company X. The other model is a company whose development effort it not critical for their overall business model and the resulting funding is a fixed bid. This is company Y.
Why is this important?
The answer lies in how the development effort is viewed by company X and Y financially. In company X the software development effort is viewed as an investment, indeed the primary investment, in the company’s future. In company Y the development effort is a small application and is viewed as an expense to be time bound and tracked. In company X the team is funded. In company Y the project is funded. Read those last two sentences again.
In company X agile will succeed. In company Y agile will fail.
In a fixed bid development effort the software development is intended to end at some point. Securing funding for the project requires that you define it up front, estimate it, resource it, develop it, test it, implement it, and then turn it over to support. This is company Y.
In a time and materials funding scenario the company determines it has need for a software development team as there are many projects that require development well into the future. An estimate of how big a development team they can afford is created for the budgetary time frame (1 year, 2 years), it is resourced, and then project scoping and scheduling are done. This is company X.
See the difference? In company X there will always be software development. There is no end and the team is funded with that intention. So putting that work in a backlog, prioritizing it, and estimating and reviewing it in time boxed iterations makes sense.
In company Y the effort can only afford to be funded for some subset of the budgetary period (say 3 months). After that there is no more money or the company is unwilling to allocate additional money. They don’t want a long term development team because they can’t afford it and besides there wouldn’t be enough development work to keep them busy anyway. So deep controls and strong project management is required to ensure that something is delivered in that 3 month period.
Viewed this way…..financially, agile is a luxury. It assumes that you’ll always have a software team and there will always be development work. It assumes you’re team is funded year after year and you, as a manager, don’t have to worry about funding individual projects. As an agile manager you’re primarily concerned with schedule, scope and capacity. Budgets are an annual or bi-annual thing. You flex up or down depending on the economic realities facing the company. The criteria for success are largely centered on functionality delivered over time.
In company Y you might have $50,000.00 that is set aside to complete you’re project in 3 months. Budgeting and expense tracking are critical and will determine whether the project is a success or not. A manager here gets funding on a project level at various times throughout the capital cycle. You may deliver all functionality on time and over budget, but that won’t be seen as a successful project. As a manager in this company you are concerned with all 3 legs of the iron triangle. Your team is likely temporary and staffed by contractor labor. Adoption of agile in this situation is a mistake…..even superficially. Why?
The key lies in estimation.
In an agile software team you don’t estimate your work till right before you begin. And you only estimate, in the case of scrum, the next iterations work. So how do you know how long it will take? You don’t. Furthermore, you really don’t care. You’ll continue to deliver functionality every iteration. As soon as product management and QA say you have a good enough product; it’ll be released as a production version. You might have a guess, but until the team estimates it….you really don’t know long it will take.
In a fixed bid situation….your estimation needs to take place up front. The company is asking you how much it will cost to build the application because they are unwilling to fund it forever. They want it to end…….preferably sooner rather than later. Funds are limited and your project, although perhaps necessary, is not viewed as critical to the company’s future. Its ROI may even be negative. Returning to the leadership of this company with the answer; “I’m not sure how long it will take….just fund the team for a year and we’ll see how it goes every iteration” would be a mistake that would likely result in your dismissal.
In the 2nd scenario, if I told the team, after securing funding and hiring them, “we’re using scrum”: they’d estimate the work the next iteration. They would assume their estimates would be taken seriously and you’d give them time to complete the work as it unfolds regardless of whether their estimates fit your original project funding or not. That’s only fair. Unfortunately, that puts you as the manager in the uncomfortable position of submitting budget/schedule variances and/or cutting scope when you’re original estimate is proven to be inaccurate. Hence, you’ve failed at managing the project and therefore…….”this agile thing doesn’t work”.
The mistake was to assume the company’s leadership understood and was organizationally committed to scrum and agile principles. The mere fact that they are asking you to estimate the funding you’ll need to complete the project tells us otherwise. If they had asked us, “How much does it cost to fund a software development team for the next 3 years?” Then we’d be wise to approach it from an agile perspective.
The Real Problem
So, what is the real problem with agile adoption in organizations? It can be boiled down to these points:
- Agile assumes that the company wants a long term software development effort not a short term project.
- Agile is often assumed by company leadership to be a development process with no impact on budgeting. This is not the case.
- Development teams assume leadership understands the implications of adopting agile at the budgetary level.
The complexity of these points can’t be underweighted. Developers and development teams often have zero visibility into budgeting so they are unaware of how their agile efforts are being accounted for in monetary terms. This is evident in so many agile articles on the web. Likewise, management is often ignorant of development and specifically agile development practices. Agile adoption requires education to ease the clash and misunderstanding of these two worlds.
So what are some of the consequences of attempting to adopt agile practices on a fixed bid project…essentially laying an agile façade over the waterfall project?
Story points are often used on agile teams to determine the complexity of the work being done. The number of points completed each iteration determines their velocity ( points per iteration ) and gives management an approximation on how much work can be completed in a given iteration.
If you come from a fixed bid shop like company Y your immediate question is, “How does this relate to hours? How do I project costs and ROI?” Truthfully, it doesn’t. It isn’t intended to. Again, the agilist is assuming you’re funding a software development team not a software development project.
In company X you could estimate things by hamburgers or cigarettes in each iteration. It doesn’t matter. You’re going to get the product done at some point ( +/- functionality requested ). The only real question is at what point to do we call it complete and release it to production. Funding for the team is not contingent on estimation of effort.
In company Y project funding is directly related to estimation of effort. It is critical that we tie this to time because our cost driver is often an hourly rate. Story points have no meaning here.
Scrum master vs PM
“Agile does away with the need for a project manager.” Ever heard this before? It’s scary for traditional PMs and unintentionally threatening. However, it is correct. If you assume that a team is funded year after year regardless of projected effort then the needs for organizing and managing the development effort are more centered on technical leadership, task and risk management. Timelines and budgets go out the door. A scrum master is sufficient, preferable for getting the job done.
However, if you’re in a non-agile situation, like company Y….then traditional PM practices are not only valid, but essential to making sure the effort is kept within budgetary and schedule tolerances. In this situation the leader of the development effort is being entrusted with precious company resources that can’t be wasted and needs to have the skills of a CEO-Lite.
A funded development team does change the project manager role. A fixed bid project does not.
Daily Stand Up Meetings
Agile uses daily stand up meetings for a variety of reasons: motivation, risk mitigation, status, accountability, team building, etc. It’s a good idea that is equally at home on a fixed bid waterfall project. There is no reason this practice can’t continue to be used, but the team on that fixed bid effort has to realize that you’re not really doing agile and there is a deadline looming. You also might want to weigh the time needs. The daily stand ups should be short, but 15 minutes a day adds up when you only have 3 months of funding.
With fully funded software teams that use agile the question of when something is done is answered incrementally. Functionality is reviewed at the end of each iteration ( say 30 days ) and evaluated for readiness for production release. Again, this is a good idea that could still be used in a fixed bid situation, but the business owners need to be coached to understand the iteration review as a risk mitigation and accountability technique and not a demonstration of the completed product.
Iteration planning really does assume you’re using agile in a team funded scenario. It necessitates your costs are known and fixed for the budgetary cycle and that any estimates the team comes up with won’t play havoc with your budget. Doing iteration planning in a fixed bid situation will almost certainly result in confusion, budget variances, and/or loss of functionality.
Burn Down Charts
Burn down charts show the progress of completing functionality in a given iteration. They are a measure of team performance over time. They do not illustrate how close the project is to completion. If we were to sum all of them up…..they might show this, but given that they are usually used strictly for the development effort of a project; this won’t always be the case.
In fixed bid scenarios the question is not usually one of how much functionality the team is doing per time period. It really doesn’t matter. They need to get all the functionality done within the time frame that the money allows.
So using burn down charts and iterations in a fixed bid project sends the wrong signal to the team and your customers.
Firstly I would suggest that trying to adopt agile in a fixed bid scenario/project funded situation is not recommended. Instead, as a manager, you should make the assessment of whether the company can support an agile practice/fully staffed development team through time. If the company can; then you should use agile…it works well. If the company can’t….then you’d be wise to stick with traditional project management practices.
If you determine that your company has the resources and the workload to support a software development effort but isn’t using agile then it comes down to convincing the leadership that agile makes sense for this effort. Put on your change management and sales hats….you’ll need them.
Secondly, project manager and scrum master are not static roles and titles. In one company a PM/SM may have budgetary responsibilities. In another they may not. In some they may have direct reports in a traditional organizational structure. In another the structure may be more matrixed and the PM/SM might have no direct reports. I mention these differences because agile articles, like all writing, is written from the viewpoint of the author. Too often what worked in one situation is attributed to being a superior process or technique…….when in actuality the organization and roles fit the process and technique. Change the roles and organization and it might not work as well. Context is often missing in the agile vs waterfall blogosphere.
Lastly, the agile debate is often likened to a philosophical war. But from my experience and vantage the confusion is largely an outgrowth of misunderstanding. Too often a technical manager hasn’t thought through the business implications of adopting agile. Likewise business folks frequently misunderstand agile as being ‘some development thing’ with no relevance to how they just funded the effort. I’ve been fortunate in my career to walk across the bridge of misunderstanding and see both sides. Doing this has given me insight into the background budgetary assumptions that so frequently go unrecognized as the cause for agile adoption failure.