Agile Business Analysis: Interview with Scott Ambler

23 Likes To complement those competencies and skills, what tools and techniques do you see the analyst using on an agile project?

Scott Ambler: The techniques are all there, they’ve been there for a long time. A lot of the techniques in agile modeling are actually reworking those user centered design concepts from the late 80s, early 90s.
So, doing a little bit of modeling at the beginning, initial requirement positioning… just spend a couple days, maybe a week or so of gathering initial requirements to do basic scoping so somebody can ask you for some sort of rough estimate or rough schedule. Some sort of description of what you’re going to build.

That doesn’t mean that you need to create detailed documentation with it. It means you need to do a little modeling. Modeling on a daily basis. Do model storming. Doing iteration modeling at the beginning of each iteration…

As far as tooling goes, the most common tools are white boards and paper, particularly for working with stakeholders; this is a uniquely specific tool technique to allow them to be actively involved in the modeling effort.

The people are also using drawing tools. The more sophisticated ones are using CASE tools. At Rational, we got a very interesting product called Rational Requirements Composer. It’s a more simplified modeling tool that’s geared for the business analyst that enables them to do analyst level modeling to get the value out of modeling, which helps to communicate but not take the hit of unnecessary documentation.

If there’s a need for detailed documentation then RequisitePro or DOORS; those are better options for you. But for the vast majority of projects that aren’t under a regulatory environment, something like Requirements Composer would probably be better off for them. So, on an agile project, there isn’t much need for the more traditional requirement management tools such as RequisitePro or DOORS?

Scott Ambler: Yeah. All things being equal, it’s not a paradigm issue, it’s an environment issue. If there’s a regulatory need for strict requirements traceability or strict requirement specification, then you’re going to have that need if you’re agile or traditional. Those are the type of factors that are going to drive the decision for do you need a more formal requirement management tool; at least in my mind. Are you suggesting that, for the vast majority of the projects, agile can be the methodology of choice?

Scott Ambler: I would think so, yeah. When you look at the facts, all the numbers point toward agile… and away from traditional. But the problem is that the religion that we see around process and the IT community doesn’t see them; so ends up not motivating people to question their beliefs.

There’re a lot of people that really still believe in the traditional way of working and don’t really understand agile, don’t want to understand the implications of traditional and yet they still cling to their old beliefs. It’s really unfortunate.

So yeah, all things being equal, the vast majority of projects should in fact be agile. And we’re seeing a movement towards this… But it requires greater discipline. It does require people have a wider range of skills, to be more open minded, to be flexible. So, some of the things that we rewarded people for in the traditional world sort of hamper them in the agile world.

Some people are not going to make the jump to agile and not everybody needs to. There will still be traditional work out there but not so much over time. What advice would you have for business analysts who are pondering moving on to agile project or have landed a job on an agile project and they’re wondering, “Now what?”

Scott Ambler: I think the important thing is to be flexible; to view documentation as almost always as a risk. You need to do some documentation but it should be your option of last resort. You should always be asking, “How can I achieve these goals in a more effective manner?”

The real goal of analysis should be: let’s identify the requirements, let’s explore and understand what people actually need, and let’s communicate and make sure that the developers actually do that… 

  • You’re going to have to be actively involved in the team because there’s not going to be eight hours a day of analysis work to do. Splitting yourself across multiple teams is highly frowned upon in the agile world. That should be highly frowned upon in the traditional world as well. We need to be able to go and talk to somebody with business knowledge right away. 
  • Becoming focused on capturing requirements specifications in the form of tests is critical. You’re still going to be doing some diagramming and high level stuff; you’re never going to get away from that. But the detailed requirement specifications, if you’re not capturing that as test, that could be a problem. This is a very clear direction that we’re seeing in the agile community. So, I highly suggest doing that. 
  • Being prepared to work with others; actually trying to pick up skills from other people, trying to help people pick up your analysis skills. Wanting to collaborate is vital. Is there a case or a place for so called “live system documentation”, documentation that evolves over time and is kept up to date for various reasons?

Scott Ambler: Yeah. That’s why in the agile community, there’s this focus on executable specifications, because then a specification actually adds true value to the developers, which in this case just validates their work and to do so in an test manner, they’ll keep it up to date, right? Because, if I change my code and I break a test, I’ve either got to fix my code to make sure it still performs that test or I’ve got to update that test. And if that means to renegotiate the requirement, then so be it.

Now can everything get specified in the form of tests? No. But can the vast majority of information be captured that way? And this, I think, is one of the reasons why we see Agilists producing measurably higher quality and achieving measurably higher business stakeholder satisfaction. We’re doing more testing and more testing generally leads to higher quality.

And working iteratively, working closer with stakeholders increases the chance that you actually understand what their true needs are and then implement that and act accordingly. This is why working in priority order helps insure that we get things out the door faster because we are much more likely to code the critical functionality first.

And then showing visible progress in the form of working software, in each iteration, is absolutely critical. It gives the business stakeholders visibility into the project. This is one of the very interesting implications of agile is that because we are working in a more collaborative manner, because we’re working in a more open manner, our stakeholders know exactly what’s going on. They can see software being produced regularly. They know exactly what they’re getting for their money. They can actually steer the project and assure they get software that meets their needs and not just something that’s built to spec. Building something to spec is a phenomenally risky way of working. Anything else that you want to get out there to business analysts about the role of the business analyst on agile projects?

Scott Ambler: I think the important thing is be open-minded about agile… If you find yourself in this position where you think Agilists don’t do ‘X’. Please go on to Google or your search engine of choice and Google the term “Agile X” and you’ll find there’s a wealth of material out there. I constantly run into people that tell me Agilists don’t document. They don’t model. They don’t test. They don’t do this, they don’t do that. You’re not doing yourself a favor by having a gross misunderstanding.

And the other thing is to just recognize that agile is coming your way. If it’s not in here now, it soon will be. It’s very clearly growing. It’s just not a fad. We’re on the exact same sort of adoption crew that would sell all this technology in the 90s. So it’s here to stay. All the numbers show that this works. This works in practice…

Don’t look for excuses not to do agile or not to be agile. There’s always an opportunity to be more agile no matter what you’re doing. agile is a continuum. So, you might not be doing everything that we’re talking about but if you can adopt one or two agile ideas to improve the way that you work, then maybe on the next project adopt one or two more ideas.

Take time to invest in yourself, invest in your career. Read and talk to people. Find out what’s actually going on.

Article Pages

Page: 3 Of 4First  Previous  1  2  3  4  Next  Last  
Like this article:
  23 members liked this article


Tony Markos posted on Monday, September 8, 2008 12:58 PM
I am all for "making decisions right now". Heck, once I have an adequate understanding of the as-is, that is a relatively easy thing to do. But I have to take the time required to learn the as-is; I lack divine intiution.

I may be hearing it wrong, but to me agile is about jumping right into the to-be - largely ignoring the as-is. Am I hearing correctly?

Scott W. Ambler posted on Tuesday, September 9, 2008 5:37 AM
If you visit and spend some time reading about the best practices, you'll see that it's common practice for agilists to invest a bit of time up front in requirements and architecture envisioning.

- Scott
Tony Markos posted on Tuesday, September 9, 2008 10:48 AM

I have to admit, it has always been my experience that, especially on larger scale software projects, more than "a bit" of as-is modeling is required. The way I learned it and experieinced it is that coming up with an adequate as-is model is the major part of the required work (I was actually taught that the as-is model is ninty- eight percent (98%) of the required work ).

BA's often lead Business Process Re-engineering projects - very as-is-oriented efforts.


Scott W. Ambler posted on Tuesday, September 9, 2008 11:46 AM
In Agile Modeling,, we talk about doing just enough modeling for the situation at hand. So, just enough is situational. The larger, more complex the project the more modeling you're going to need to do. A good heuristic is that for every month of the release consider doing one day of up front modeling (i.e. for a 12 month project consider doing 12 days), with a cap of two weeks (10 days) for complex projects and one week for straighforward projects. Naturally logisitics and your organization's ability to focus on getting the job done and stay focused on high-value activities can stretch this out, but I would consider that a significant risk.

My experience is that the professional modelers among us tend to think that we need to do an order of magnitude more modeling than what is actually required. This is because they're often overly specialized on modeling hence overestimate its value, because they've been trained to think that way, and because they often haven't gotten their heads out of the theory of yesteryear. Having said that, the "extreme programmers" among us often think we need to do an order of magnitude less modeling than we actually need to do.

Bottom line is that if you have the skill to model something today, you will also have the skill to model it tomorrow when and if you actually need the details. The goal should be to get the value out of modeling which is to communicate and think things through without taking on the risks associated with too much detail too early.

In the November issue of Dr. Dobb's Journal I summarize the results of a recent survey that we did regarding the current state of modeling and documentation in IT. From the looks of it the agile community appears to be a bit more effective at modeling than the traditional community and a bit more likely to create deliverable documentation. I hope that the article will be an eye opener for a lot of people.

- Scott
Tony Markos posted on Wednesday, September 10, 2008 4:43 PM

Thanks for the informative reponses.

haj posted on Friday, September 12, 2008 12:36 AM
Regardless of the name you apply to your approach to analysis, a characteristic of effective business analysis is clear communication. To that end, can you please clarify your fourth bullet point ("We’re self organizing, so that the people that do the work are the ones that are best suited to organize and plan it. That’s a bit different") which in its current form can be taken to mean either you advocate choosing the workers from the organizers OR that you advocate picking the organizers from amongst the workers. Thanks in advance
Scott W. Ambler posted on Friday, September 12, 2008 7:39 AM
Like I said, it's a bit different. The workers are the organizers. On agile teams there's a blurring of roles. We're moving away from overly specialized people, such as people who are just programmers, or just testers, or just BAs, to people who are generalizing specialists (see ) who have one or more specialities and a general knowlege of the process -- the sweet spot between being specialists and generalists. Generalizing specialists are more effective than specialists as they don't need as much supporting bureaucracy, they have the ability to apply the right technique to the situation to the right degree, and they can interact more effectively with others because they have a better understanding of the issues that they're trying to address. When you're just a specialist you're effectively someone who just has a hammer, everything looks like a nail.

So, instead of having someone do the planning for the team (i.e. the manager), the team itself does the planning. The manager may facilitate the effort, but the team does the majority of the planning work. When done on a JIT basis, this proves to be far more effective than traditional planning approaches.

You're correct that an important part of business analysis analysis is effective communication, arguably it's the most important part. But, we need to step back and observe that the traditional approach to analysis leaves significant room for improvement. The agile community has adopted several improvements, for the most part centered on reducing the feedback cycle and choosing more effective ways for people to share information (documentation, by the way, is the least effective method for people to share info). For several years now I've suggested that the business analysis community needs to point their skills back at themselves. I suspect that if they did that honestly that they'd start questioning the traditional strategies that they seem to prefer.

- Scott
haj posted on Sunday, September 14, 2008 9:43 PM
Thank you for directing me to the 'generalizing specialists article. Excising the circumlocution, your approach relies on an effective team. You pick (presumably expensive) resources who are skilled at more than one discipline. Ideal, that. I opine that such a team is by definition effective regardless of any particular method. In fact, left to their own devices, they may be relied upon to find and apply their own most effective methods, which may or may not be the methods you advocate. It is to improve the likelihood of success when the project team is less than ideal and/or operating in a less than ideal set of cirumstances that we codify and apply methods. The higher the risk, the more strictly we should enforce the methods which are, after all, insurance against failure. It strikes me as specious to promote Agile as a superior method if it requires a ‘better’ team than the alternatives, which themselves have evolved to maximize success even where the ideal team cannot be assembled, that is to say, almost always.
Scott W. Ambler posted on Monday, September 15, 2008 8:57 AM
My experience is that it's realistic to build teams from generalizing specialists, but you actually need to choose to do so. Just as a specialist team would be made up of some novices, some mid-level people and some experienced people so would an agile team made up of generalizing specialists. The novice GSs might just be specialists, the mid-level ones might be people with one specialty and a general knowledge, and the experts would have more than one specialty.

The deciding factor is that you need to build a team with a learning and collaborative culture going in, instead of building a Tayloristic team of specialists.

So, not having a group of experts on hand isn't a valid excuse for becoming more agile. It is, unfortunately, a common excuse for why organizations remain at lower levels of productivity. My advice is to choose to succeed.

Also, I've been pretty clear in my blog, and in other writings of course, that I certainly don't expect conditions to be ideal. Nor have I ever actually found myself in an ideal situation. So, not being in an ideal situation isn't a valid excuse to not become more agile either.

Strictly enforcing methods hasn't worked very well in practice over the years. Sounds good in theory though. You need to choose approaches which reflect human behavior, and as we know knowledge workers don't react well to being told what to do or to being strictly monitored. You might want to read the IBM whitepaper that I co-wrote with Per Kroll on Lean Development Governance which goes into this topic in more detail.

- Scott
haj posted on Monday, September 15, 2008 7:11 PM
No, thank you, I have read enough of the style, and in any case I will be busy searching for an example where “enforcing methods hasn't worked very well in practice over the years”. This could take some time.
Scott W. Ambler posted on Tuesday, September 16, 2008 5:15 AM
You're right, it's likely going to take some time because it's very rare for organizations to write up their failures and share them with the world. What's more common is that the problem gets ignored and then swept under the rug. Worse yet, many organizations will apply the same failed strategy over and over in the hope that somehow they'll get it right this time, being unable to observe the failure patterns that they're experiencing.

To see patterns of failure you need to work at a series of firms, something that consultants get to do but non-consultants generally don't. When you start to notice the same patterns over and over again you start to question what's going on. When you only see a one or two organizations struggle with a strategy, often several years apart, it's easy to assume that it's just bad luck that they both got it wrong.

You might find the results of the 2006 Data Management survey, , to be interesting because it found that two thirds of development go around their corporate data groups. Although this is a complex issue, it's clearly a sign that data management efforts are struggling in many organizations. Yet, when was the last time you read a case study about a struggling data management group?
Acadia posted on Tuesday, September 16, 2008 10:17 AM
Hi Scott,

I have been reading your articles for a while now and on Agile as well. I think Agile came in to place just to create a balance in a world that was relying too much on process and somehow forgetting the end user. I think Agile was a way to cut down the process drone and get some dynamic feedback mechanism into the picture. This is also clear in your article when you say one can choke oneself with all the modeling and documentation that no one will read.

Its also great to know you conducted a survey and you found out that Agile teams indeed focus and deliver modelling and documentation. Something that even you were surprised to learn.

And here is where I find the issue. Since politics is so much the talk these days, I can't help but say that throughout history individuals/ideologies comes to power in the hope of setting right and making better what existed, but end up doing some abject things themselves.

In the same light, I think the Agile community has to clearly advocate what it wants to do and there must be a absolute lack of ambiguity. Many people misunderstand Agile to mean no documentation. I think somewhere the founders did feel that way too and your "just enough" is not really a good enough term. Not having documentation is going the other extreme. Forget surveys and polls, I have found in my own company that after the product was delivered and shipped and the applause died down, people came back to fix something after months just to be utterly dumb founded and not knowing how to go about it. Or users in non-tech groups fumbling and cursing under their breath because no matter how good the usability Help does help many a times.

And when you talk of roles not being super specialized and planning being done by the teams, you have to have a precondition that it works in places where there is a good rapport and trust between the team. I cannot imagine Agile surviving in an environment where there is hostility backstabbing and an aggressive client. And there is no documentation to save your soul. So though your survey speaks clearly, I wish Agile community also understands the preconditions where Extreme Programming and Agile will work well to save a project and where it simply cannot. And I also wish that the Agile community mentions that documentation is really v important.

I strongly resent the fact that in life if you went about changing your mind, you would be booted, but in the software world, a client who changes his mind endlessly is accommodated. I need to understand yet how Agile even tags cost with all the changes and scope creeps with no change management and with documentation coming at the fag end?
Scott W. Ambler posted on Wednesday, September 17, 2008 7:31 AM
Some thoughts:
1. Software development is a complex and situational effort. As a result there is ambiguity in the agile message and justifiably so. One size does not fit all. Perhaps your need for definition is an indication that you may struggle to become more agile?
2. If people misunderstand agile to mean no documentation that is surely not the fault of the agile community, but instead the fault of those people to not do a bit of research. Just google the term "Agile documentation" and you'll find a wealth of articles on the subject. My Agile Modeling book, published in 2002, covered the topic in detail and the Agile Documentation book in greater detail in 2005. The topic has of course been addressed in other books too.
3. Just good enough is in the eye of the beholder. Just good enough artifacts are the most effective you can possibly create. Once again, software development is situational and one size does not fit all. I've described JBGE in detail at
4. If your organization is having problems then you need to learn from your experiences and improve your process. If there is an actual need for more documentation then learn from that and invest effort in it on future projects. Agilists treat documentation just like any other requirement -- it's estimated, prioritized, and then implemented appropriately. Also, is the real problem lack of documentation or poor hand-off procedures? I've seen lots of orgs treat documentation like a band-aid over their release process. Instead of just handing a system off to a maintenance team, keep some of the developers around or bring some of the maintainers into the development effort. Also, were the teams writing high-quality code with a full regression test suite? If not, then I suspect that's your actual problem.
5. Help features, user manuals, ... should be treated like any other requirement (see above). If the stakeholders choose not to invest in these things then who is really at fault?
6. If you don't have good rapport and trust between the team you're in very serious trouble regardless of paradigm. With agile you'll likely fail quicker, which is a good thing. Papering over problems like this only extends the point at which you declare failure, increasing your actual costs as well as opportunity costs.
7. The agile community has been pretty clear about the preconditions for various agile methods. For example, the XP folks have been very blunt about this (once again, invest some time with Google) as have I with Agile Modeling (drop by and search the site).
8. People can and will change their minds. Get used to it, this is human nature. Furthermore, it's the customer's money, not yours, and if they choose to spend it in "questionable manners" that's their decision, not yours. If you resent things like this you either need to work through your issues or you need to find work in another profession more aligned with your value system.
9. The agile community is spectacularly clear about how to handle the cost of changed requirements, see for one of many write ups on this subject. We treat stakeholders like adults and ask that they be accountable for these sorts of decisions, it's their money after all. One of the advantages of agile, and one of the challenges I suppose, is that it provides greater control and visibility to stakeholders. The stakeholders can change the scope, control the budget, and control the schedule. When you deliver potentially shippable working software on a regular basis, that enables stakeholders to make concrete decisions on whether the system is ready to be shipped, whether they're getting what they want, and how much (if any) they should continue investing in this project. When you implement requirements in priority order you deliver the highest ROI possible, as defined by your stakeholders. When you deliver high quality working software, with a full regression test suite, on a regular basis you can accomodate change easily. Working in ths manner requires greater accountability on the part of stakeholders (clearly a challenge in many orgs) and greater discipline on the part of IT professionals (also a challenge in many orgs). We've upped the game in the agile world.
10. If you step back and observe what happens on traditional projects, most of the critical deliverable documentation is written towards the end of the lifecycle just like on agile projects. See
Acadia posted on Wednesday, September 17, 2008 9:36 AM
Scott, thanks for your response and your time. I agree with your comments. The problem is when there is a lot of half-baked knowledge around and time is not invested in reading and researching. And I think practitioners and experts like yourself are doing a fabulous job of putting so much knowledge out there for people. There really is a need for advocates and enlighteners - who know the stuff to educate the people.

My worry is the misunderstanding of Agile must not undermine the role of a business analyst in the project sphere. Reducing our work to mere facilitation may not stand in the way of companies throwing out our role altogether.
Amar posted on Thursday, September 18, 2008 1:51 PM
It is a side question but can I ask anyone of you to shade some light on the type of questions asked on a BA skill evaluation test?
Scott W. Ambler posted on Friday, September 19, 2008 11:24 AM
Agile is real, it works well, it's here to stay, and it's growing. So, the BA community, and other communities for that matter, needs to accept this and find ways to add value on such projects. This will require them to be flexible enough to change. If not, they're going to struggle. BAs clearly have value to add, IMHO, but they need to choose to actually do so. The implication is that they will need to change the way that they approach their work.
Jim posted on Wednesday, November 12, 2008 3:25 PM
I think you've gone a step beyond the Agile Manifesto to say that Agile is against documentation. I don't see that at all, but rather that people and solutions are more valuable than documentation. There are many times that documentation IS part of the deliverable, and I don't see any conflict in that. Bureaucracy or not, if the business requires documentation, we should be delivering it, and why not deliver it in an Agile context? Many organizations believe that they cannot implement Agile because it's anti-documentation. They are legally bound to provide documentation. I'd much rather see the documentation as part of a sprint than a tired old traditional method.
Only registered users may post comments.



Copyright 2006-2024 by Modern Analyst Media LLC