Q: I would be interested in anything you might have about rolling out Agile on a team that depends on components created by non-Agile teams. In particular, how that is affected by different approaches to Quality between teams in the same company (and no, we don’t have a common standard, only on paper). I was reading about the GM/Toyota experiment (first TPS plant in the US) and how GM had issues trying to roll it to other plants, one of the biggest one being the fact that unlike in Japan, they didn’t have the power to push their process down to their producers. They quickly found out that they could not build a Quality car without Quality components; and I am afraid we will find the same here.

A: The reason quality is generally higher with the output of Agile processes is related to the nature of feedback loops built into Agile. We feedback on the product/design much more rapidly. Practices like pair programming or lightweight peer review, automated testing, short iterations, automated build/continuous integration, and close collaboration with the customer/proxy, all tend to give us more feedback on the product/design… which tends to lead to higher quality. My recommendation would be to try to drive as much of those feedback loops up stream as possible. You don’t control their process but you may be able to influence it at the boundary between them and you.

Close collaboration. The lowest hanging fruit is probably the close collaboration with the customer one. In this instance, the Agile team is the customer. The non-Agile teams are the vendors. I’m thinking of setting up regular demo/review meetings (probably on the cadence of the Agile team – short iterations). You may also be able to visit (virtually or physically) on a near daily basis.

Automated testing. You might also try setting up automated testing at the interface level for the components delivered by the upstream teams. You’ll have to avoid the trap of using this as “contract negotiation over collaboration” but that is in how you handle it. The key here is that you want them to think of the tests as a tool to help them do their job as opposed to a way to enforce something. This means that they will need ability to run the tests before delivering to you. It would be better still if they owned the tests and you reviewed them. No matter who owns them, the tests become the specification for the API, which is a good Agile smell.

Peer review. At this point, you are collaborating/reviewing the test code. This might then lead to a situation where you might be able to do peer review of their production code. I’d prefer a peer review approach that helped them improve their code (and learn how to write better code in the future) over one that just allowed you to fix their code after the fact.

Automated build. If you were to give them access to your build process, they would also be able to test the compile-time agreement between their code and yours. This comes with two immediate benefits: (1) it serves as an additional automated test of of the interface, and (2) this (combined with the other automated tests) gives them more confidence to refactor their code and make improvements. The assumption here is that most teams know their code has warts but they are afraid to modify it to improve it because they are afraid of breaking code that depends upon it. Running your build script lowers the fear.

There is a third (and potentially more powerful) benefit to a shared build process. It provides you with a place to plug in other quality improving tests and analysis. The automated testing that I proposed above are tests that run against their upstream code. With an automated build, you could include tests that run against your downstream (but higher level) code. This means that they could see if changes that they make break your higher level functionality. You’d have to use a stable version of your source so they could be sure the problem was theirs but a distributed source control tool or careful branch management could overcome that obstacle. The build is also a common place to run automated bug finders like FindBugs or even custom analysis like a tool to highlight any changes in the calling signature.

Please let me know if any of this helped. Maybe I can refactor and improve my answer (upstream product) based upon your feedback (from downstream).

Recently, I have had the chance to work closely with a number of projects that were not pure software. They all had some software or firmware component but they also included an electronics or even mechanical design aspect. Below are the top ten questions I recorded when working with these teams and the answers on how the teams effectively answered them.

1.    Are Agile practices and processes effective when conducting non-software projects (firmware, electronic, mechanical, etc.)?

Absolutely. Some of the XP engineering practices are not directly applicable or need to be modified based upon industry or your particular situation. However, surprisingly minor adjustments are all that is necessary for the Scrum process framework to be highly effective… even compared to processes that evolved specifically for hardware.

2.    What adjustments need to be made to make to the Scrum process framework work well for these projects?

Surprisingly few. The primary adjustments center around expectations in two general areas: (1) minimal marketable feature/emergent design/thin vertical slices, and (2) user stories.

3.    What adjustments need to be made to our expectations around minimal marketable feature, emergent design, and thin vertical slices?

Focus on feedback. True, the break even point between thinking versus building encompasses more thinking for non-software projects. Even so, push to build something sooner and even when you are working on early design and infrastructure, get feedback on some “product” each iteration… and get that feedback from a source as close to the user as possible. In software, it’s almost always less expensive and more effective to build something rapidly, get feedback on it from an actual user, and change it, than it is to think longer about how to build it expecting to avoid later rework. This is the primary reason why Agile practices encourage you to break up the work into increments of minimal marketable features (MMF). Agile software projects try to build as little infrastructure as necessary to implement the current MMF and let the design emerge from those increments rather than nail down all the requirements and design up front. You are encouraged to build the system in thin vertical slices where all levels of the product experience changes with each increment.

When confronted with this idea, even software-only teams push back. Software teams want an architecturally sound base upon which to build their features. However, I have found that if the team thinks on it for a bit, they can often find a way to build usable features and start to get feedback much sooner than the team originally imagined. Those projects are able to deliver a marketable feature even in the first few iterations and they move rapidly to a mode where very little of each iteration is spent on infrastructure. The primary difference for hardware projects is that you have a different cost structure for manufacturing so the break even point for thinking versus building encompasses a bit more thinking. This means that it may take longer to get into the mode of designing several MMFs each iteration.

Remember, the primary benefit of this approach is to get the most valuable possible feedback as often as possible. So even when it is hard to use thin vertical slices to accomplish that, you should still seek out opportunities to enhance the richness and frequency of the feedback you receive by producing something to get feedback upon during EVERY iteration. The next step down from demonstrating an MMF is producing a prototype but even that can be hard to do every iteration of a hardware project. So when the thing you produce is only a document, a design, or an experiment, make an effort to maximize the value by choosing to get your feedback from a source as close to the customer as possible.

4.    What adjustments to our expectations need to be made around user stories?

Understand that the “user” in user stories only hints at one of four good reasons to manage requirements with user stories… and it’s not the most important reason, “conversations.” Maybe they should be called “conversational stories.” The big hang-up of hardware teams managing requirements with user stories focuses around the word “users”. That’s unfortunate because I don’t think that is the most important benefit that Agile teams (even software teams) get from the practices surrounding user stories.

The benefits come from four aspect: (1) WHO, (2) WHAT, (3) WHY, and (4) conversations. This last one, conversations, is the most valuable so I’ll talk about it first.

Even when drafts of requirements documents are shared with the development team for early feedback, the development team doesn’t internalize them until they start working with them. By having the team size, and often write user stories, you force them to start this internalization much sooner. The early conversation between the development team and stake holders around the requirements allows for implementation costs to be factored into requirements tradeoff decisions. The ongoing conversation throughout the project lifecycle, provides a high-fidelity communication channel and continuous vision alignment.

Now, let’s address the other three beneficial aspects one at a time. The traditional user story format is, “As a WHO, I want to WHAT, so that I can WHY.” All three of these, WHO, WHAT and WHY, provide benefits. The tension of trying to always make the WHO be an end user is not unique to hardware. Even in the most Agile of software-only teams, there are times when the end user is only indirectly the beneficiary of a particular backlog item. For instance, the most direct beneficiary of a research activity, a mockup, or a prototype (collectively referred to as a “spike” in the agile world) is the development team and not the end-user. In those cases, specifying the WHO does not encourage you to think about the product from the end user’s perspective. If every user story were this way, then we wouldn’t call them “user stories”. The “user” is in the phrase to remind us to get the user perspective involved as often and as soon as possible but just like MMFs, this practice is harder to do as often especially early in a hardware project.

I will not dwell on the WHAT because this element is present in all approaches to requirements management, except to mention that it is important for the what to not drift over into the HOW so the development team has flexibility in how they meet the identified need. Note: The Rally tool’s entity for “user stories” is really more of a generic backlog item. There is nothing in the tool that enforces or even makes awkward, the use of this entity in a traditional work breakdown mode.

On the other hand, the WHY is somewhat unique to the practice of user stories and can be very valuable. Understanding why someone wants something empowers the development team to be creative about satisfying a need… sometimes even by explicitly not satisfying the WHAT of the user story.  If a team is told that it must implement a data layer with sub-millisecond response, they may blindly go about accomplishing that… at great cost. My first response to a user story written like that is that it crossed the border from WHAT and into the realm of HOW. Nevertheless, even if you give a team a user story like that but you also tell them that the reason for this “requirement” is the responsiveness of the user interface, they may take steps to provide low latency to user input even when the data does not make it all the way to the data layer for a second or more… saving cost AND improving the product.

5.    What about prioritizing user stories strictly by value to the end user?

Prioritize by overall business value, not end user value. Even in software-only projects, user stories should be prioritized by overall value to the business, not the end user. Often that is the same thing and certainly, the end user’s needs are the biggest factor in prioritizing any user story with an end-user as the WHO. However, a feature that is desirable to the end user but not saleable might not be valuable to your business. Similarly, valuable features that are too costly (either to produce or as tradeoff for against other desirable features) might not be good decisions.

Apple has been criticized for excluding multi-tasking from the iPhone. They realized that multi-tasking negatively impacted battery life and user interface responsiveness and explicitly left it out of product. They made a business decision that they could still sell the iPhone even without this high profile feature.

However, before they made this decision, they needed some information. How much did background tasks hurt battery life and responsiveness? How amenable will potential customers be to purchasing a product without it? Apple can easily justify investments in research to determine the extent of this impact on both the usability and marketability of the product. This information is of no direct benefit to the end-user but the work necessary to gather it, was of immense benefit to the business.

Development projects of all kinds benefit from good design and marketing decisions. Backlog items focused solely on these outcomes are of value to the business and should get appropriate prioritization. Similar to the above discussion on MMFs and the WHO in user stories, it just may be that non-software projects experience more of these tradeoff analysis backlog items early in the project and they keep seeing them longer into the development cycle.

6.    Should user stories be our only tool for requirements management?

Not usually for hardware/mixed projects. There are many reasons why you might want some other mechanism to compliment your user story practice. For instance, the concept of abuse cases is often part of a larger security review. Safety reviews often have a parallel mechanism. Protocols and other interfaces are best defined by other means. Hardware typically have requirements associated with the operating environment. Etc.

7.    But user stories are not even an official requirement of Scrum so why shouldn’t we just use our traditional requirements practices?

Consider alternatives but remember all four valuable aspects of user stories. It is true that the official definition of Scrum simply calls for there to exist a backlog of work. It only mentions user stories in a sidebar and even then, the sidebar also mentions other approaches like use cases. The essence of Agile is (1) self-organize, (2) do something, and (3) inspect and adapt. The definition of Scrum is just one step more detailed than this essential definition of Agile and is intentionally minimalistic so any iterative agile approach would fit.

User stories have emerged as a common and valuable practice because of the reasons mentioned above but it is not strictly required. Your team should feel empowered to consider alternatives.

However, if your team chooses another approach to doing requirements management, you should not deviate from the agile practice of allowing the development team to do the estimating. Also, I encourage you to think about the reasons the practices surrounding user stories are valuable (other than the emphasis on the user) as described above and enable as much of those reasons as possible, starting with the conversation aspect.

8.    What about when we need to send a board (or prototype part) out for manufacturing and it will not be done within an iteration?

Push for rapid prototyping but adapt to your capability. This is a very specific question that comes up often when folks are told that they need to produce something upon which to get feedback during each iteration. What if the time it takes to get prototype parts back from manufacturing is longer than an iteration?

My first response is to ask yourself, “Is there ANYTHING that we can do so that we CAN produce a prototype in a iteration?“ The world of prototyping has attempted to keep up with the ever-increasing pace of change. There now exist component suppliers that allow you to upload a part design in the morning so that they can produce and ship it overnight. Those services are expensive but so is the time of your team. Failing some solution like that, “Is there a different way to produce something to get the answers and feedback we need for  decisions within a single iteration?”

If you still cannot think of a way to produce it within one sprint, you can handle it by breaking the backlog item down. The first portion includes whatever work is necessary to place the order for the part. The later portion includes any evaluation activities. Collectively, they have value to the business.

9.    What about dependencies and critical path analysis?

Supplement when needed but ask if it is really needed. Dependencies are considered by the product owner and the development team when choosing stories for a particular iteration. However, the consideration of dependencies is informal and not explicit like in a Gantt chart format (think Microsoft Project). I have worked with teams where explicitly and continuously conducting this sort of critical path analysis is…. well…  critical to their success; but I have worked with many more teams where the use of a Gantt chart is merely the default and what they are used to. For those projects, the most important thing is for each team member to know what they should be working on right now and have a sense of urgency about getting it done-done! The mechanisms in the Scrum framework are highly effective at accomplishing this. If you do need to conduct critical path analysis at some point, I suggest that you do it only as needed.

Note: The Rally tool includes functionality for you to record dependencies so that they are readily available when you are making decisions about what to work on next.

10.    Maybe we don’t need continuous critical path analysis, but we still have specialists that are not permanently dedicated to the team. How do we deal with that?

Favor cross-training and using generalist team members but fall-back to explicit allocation and coordination when necessary. The Agile approach is to fully dedicate as many of these specialists to the team as possible. Even when you know it’s not a full-time job for a particular specialty, it still might be better to supplement those specialists’ workloads with team tasks that are outside of their specialty. We find that becoming Agile tends to encourage more generalists (or at least multi-specialists) to emerge. This cross-training is generally positive on its own merit but double so when you factor in the cost of task switching and the productivity befits you get once a team learns how best to work together (think Forming-Storming-Norming-Performing).

Even so, there may still be some centralized functions that your teams will need to consult. It is often possible to handle these situations by leveraging the team’s approach to dealing with outside suppliers.

When you move the solid line from a functional manager to a team lead and make the functional manager the dotted line, it will bring up many issues like personnel reviews and career counseling. The coaches at Rally have experience with companies making these transitions and can help you with those tough issues but you will have to work through them. “Agile is easy. Implementing Agile is a bit more difficult.”

I started to write a long comment in response to a post on my friend George Fairbanks blog, but when it got above three paragraphs, I decided to move it over here. The question was whether or not data modeling should be considered part of software architecture.

The short answer is, “Yes” but my reasons why may surprise you.

I’ve sat in many a team meeting where early design decisions are being made. Nothing helps engage the team better, than settling on a data model. At levels above that, where you just say, here’s an employee, here’s a customer, the team will just nod and agree to whatever the architect/design lead/etc. is saying. It’s only when you start to ask, “How are we going to represent these objects in the system?” (and really only after you discuss which objects go in which tables), do you start to get folks to actually think about (and commit to) building it. I’m not sure you have to go down to the data modeling level all the time, but for many situations you have to do it to get the team engaged.

Developers won’t require that you decide in the meeting whether or not it’s a varchar or a string(20), but they may not engage unless the conversion includes things like, “Are we going to use the same table to store customers and employees.” That’s when you start to ask questions like, “Is an employee ever going to also be a customer? If so, is it OK for them to have two separate entries in the system? How much data is in common between the two? How much is different?” This does get a little close to George’s fear about discussing N-th normal form but there it is.

But that wasn’t the original question. Nobody is saying that data modeling isn’t useful. We’re just asking whether or not data modeling should be considered part of software architecture.

Never forget, we have to get PEOPLE to build this stuff. The single most important driver for the success of a software project is the commitment of the people doing the work. If the coders think that the plan was pulled out of the air or the design is wasteful, pipe dreaming, or otherwise deficient, they will passive aggressively kill your project. If you’ve ever been part of an effort where the folks on keyboards think, “the PowerPoint architects who designed the systems are idiots”, you understand what I mean. So the number one priority is to get the developers to say, “our design”. If they ever think, “his design”, you are much more likely to fail.

The architect’s job is to get the team to commit to a good design. It’s not to deliver a good design to the team.

And the only way to be certain, that the team has committed to it, is to get them to find it themselves. You can tell them until you are blue in the face that you have analyzed this nine ways to Sunday and you are convinced… blah, blah, blah. I repeat, the only way to convince them, the design is good is for them to come up with it on their own.

As a great architect, you should have already walked through the issues yourself. Certainly, you’ll use the higher level abstractions that you are writing about, George, to think things through. But you can’t stop there. If you throw your design over the wall at that point, who knows how it will turn out. And don’t tell me that you sent it out for comment. Nobody reads that stuff. You have to host a meeting or a series of meetings, preferably with a good facilitator/coach (like me or Noopur Davis, who does this better than anyone I know). You go into the meeting with an architecture in mind and probably even well documented, but don’t hand those documents out. Start with a blank whiteboard. Add a simple block diagram and proceed from there with the Socratic method. Someone will suggest something you’ve already rejected and you’ll say, “Sounds reasonable, but what about foo?” They will then either 1) reject it like you did; 2) suggest a creative way around foo; or 3) convince you why your assumption or priority is wrong. Every time, I’ve seen this done, the architecture is different coming out of the meeting than going in. Sometimes the improvements are minor and might have happened anyway during implementation. More often, the changes are major.

If the architecture is for a large system of systems, then you may only have sub-team leads/architects in the meeting and they will probably not want to talk at the data model level. But when architecting for a 2-pizza team you better be ready to go down to the data model level.

You should of course follow up the meeting by delivering back to the team, the architecture that they came up with in the meeting. If it’s close to what you had already documented before the meeting, so much the better. To do this effectively, you may need several meetings at different levels. Don’t worry about the cost/time. If you delivered a finished architecture, you’d have to spend time training them on it anyway. This is just a fun (and much more effective) way of accomplishing that goal of them understanding what they are going to build. Your biggest problem will be resisting the temptation to share your wonderful architecture with the world and get credit for it. Your best approach is to make sure the team gets all the credit.

If the book you are writing is for software architecture researchers, then you can ignore everything I’ve written. If on the other hand, you are writing something that can be used by practicing architects, then never forget, we have to get PEOPLE to build this stuff.