What skills should business analysts acquire to add value on a project?

One way for a business analyst to really add value on a project for a customer is when he/she understands the customer, industry and competition well enough to be able to give valuable solutions and suggestions. Analysts, in most of the projects that I have worked on, do not specialize in a specific domain. They start on a project with very little knowledge of the customer’s business and slowly learn the domain as they go. They are mostly taking requirements from the customer and managing scope. What should the analysts learn in order to really make a difference to the customer (and also remain in business ☺) and how should they go about doing this?

Do we need always need business analysts?

If the customer is reasonably clear on what they want and the nature of the application is such that unless you have a lot of business domain expertise you cannot really shape the requirements, then isnt it best that the customer interacts directly with the developers and get the app built? An iteration manager to coordinate things might be sufficient... Looks like BAs will be out of business soon ;-)

Tall stories

Should we always follow the story approach? What if the nature of the beast is that there is a continuously evolving application where one cannot define even a functional test with data? If you define a story, no sooner than its 'complete', it will need to be changed, since new information comes along that makes your acceptance criteria obsolete.

Since I am currently experiencing such an approach in a particular area of the project I work in, we are veering round to the view that we need not track velocity or story completion of such work. For purposes of documentation, we will still have stories which detail what work needs to be done at a point of time, but we are not going to bother with tracking the acceptance criteria after that and the points delivered once this work is complete.

Comments?

Estimation - why and why not?

I work for a project which follows agile methodology for software development.

My team has Programmers, Business Analyst, Testers, Automation engineers, Project Managers, Subject Matter Experts and Customer.

When we need to develop a feature, we need to estimate and find out when that feature would be complete.

When we need to plan a feature, we ask estimates only from programmers. We don’t really ask estimates from other teams like business analyst team and testing team.

Are we doing a bad planning by not asking estimates from analysts and testers?

The estimation is one of the attributes requires for planning. We do estimate so that we can plan. So, when we ask estimates from analysis, testing or programming team, it should help us to do planning. If estimation does not help us in planning then we should not estimate.

We follow agile methodology, which is more or less similar to Just In Time (JIT) method of manufacturing. We like to keep our inventory level as low as possible. For us, inventory is a piece of requirement, which is termed as “story”. We need to ensure that we have minimal number of stories in pipeline. My pipeline starts when a business analyst starts breaking a “customer need” or "feature" into stories. The pipeline ends when story development is complete and it is ready to go live.

I need to plan so that my pipeline is thinner. I should pick up the inventory that i can finish. Thus, I need to know how much programming team can consume in a given time period. Thus we need programming estimates. I also need to know numbers of BAs and Testers needed to support the work that programming team can take. Hence, BA/Testing estimates.

Now, let’s take a different point of view. JIT didn't become as popular in USA as it was in Japan. JIT demanded a strict discipline during planning and implementation phase, which many companies in west were not able to adapt. Though, western countries could successfully able to adapt an ‘Avatar’ of JIT called “Theory of Constraints (TOC)”. The basic principal behind TOC was similar to JIT -keep minimum inventory in pipeline. But, it was rather easy to adapt compared to JIT.

TOC suggested that the team will move as faster as speed of your bottleneck resource. In a highway, if overtaking is not allowed, the traffic will move with speed of slowest truck and not with speed of a sports vehicle. Thus, you need to ‘estimate’ the throughput only for the bottleneck resource to find how fast the team can go.

Applying the same analogy to an agile software development project, we can chart out project plan based on speed of bottleneck resource. To do that, first, we need to find the bottleneck resource. Though, we don’t need estimates from any non-bottleneck resources, we like to closely plan capacity for other resources so that those resources don’t change in a bottleneck. If a non-bottleneck resource changes to a bottleneck, then throughput will be based on speed of new bottleneck. The old bottleneck may not remain that relevant in this scenario.

How do we know that a particular resource is becoming bottleneck? The stories will always queued up in-front of bottleneck resource. If your testing team is a bottleneck, you will see lots of stories lined up for testing.

If you start with an assumption that your programming team is going to be the bottleneck, then you should do only programming estimates. If you take programming estimates for planning, ensure that small amount of inventory (stories) always piled up before programming team. This is to ensure that bottleneck (or critical resource) is never having dearth of work.

The BAs should pile up stories for programming team. The pile in-front of programmers should never be empty. Thus, it is better to have some extra BA capacity so that BA team should always be able to satisfy the need for Dev team. The BA capacity should be increased or reduced based on pile of stories for developers.

The Testers should always be free (yes, free) to pick up stories for testing as soon as a programmer completes the stories. There should not be any pile in front of testing team. Again, the testing team capacity should also be closely monitored and adjusted as required.

We will agree that the BA and testing team capacity cannot be adjusted on daily basis to achieve the same. The resources take time before they can join the team and be productive. Thus, for a new team, the BAs and testers should be added in a ratio to the programmers and then, should be adjusted to suit the need for the team for first few months.

After these adjustments to BA and testing teams, the BA and Testing estimates does not really required for the planning. The programming team sets the pace and BA/testing teams always able to meet that pace as they are extra in capacity.

Based on above analogy, I don’t see any reason why any team needs to take estimates from BAs and testers unless they are real bottleneck in the project!! If we think that BAs (or testers) are bottleneck (i.e. critical resources) then we should take estimates only from BAs (or testers) and stop getting estimates from Programming team, as BAs (or testers) will decide when project (or requirement) would be complete !!

How do I decide optimum release size?

So, my project sponsor wants to do a production release every month. He feels that we are not agile enough as we are doing releases only once in 3 months. Our team size is 100+.

His core reason is : He needs to give the functionality to his customer with shortest lead time possible. Also, he likes to keep giving functionality to his customer one by one rather than in one go.

If we make release duration too short, it will not allow us to complete some features scheduled in the release. This is an extra overhead for the release when a feature is not production ready, as dev team needs to switch off that functionality and QAs needs to ensure that all workflows works fine after switching off the half baked feature.

So, what are the factors that i should consider before deciding on a release size?

User Stories: Value and Size

ShareKhan brought up the issue of story value and size in his recent post here. My comment was getting too long so here's a post.

There are three different issues that ShareKhan is talking about:

1. Business value of a functional story
2. The reason for splitting requirements into stories AND
   
3. The practice of "storying up" technical / environmental tasks

Business value of a functional story
Maximizing work not done is the art of our trade. The business value of a story is looked into precisely for this reason. If you don't have a good reason (business value) behind a given piece of functionality (story), you are wasting time by working on it.

This means that technical necessities, like refactoring of code, also have to pass the business value test to be played. They are not stories by themselves at all. Refactoring happens as part of a functional story which requires the code to change in a certain way.

Splitting requirements into small stories
The basic reason why we split requirements into stories is to deliver working software to the client as soon as possible and get her feedback on it. Development convenience, Better tracking of the iteration progress / velocity, etc are just side effects of having small stories.

Another thing that we aim at doing is delivering value (read "usable software") with every build. This means that if the plug is pulled, the client still walks away with something she can use. Of course, if you pull the plug after the first day itself (like in ShareKhan's example) you won't get much but I guess that's true for anything in life.

What small stories do is they let you build a coherent piece of software that has some value. Let's say we have the following stories.

1. As a sales person, I should be able to enter the product code by typing it in
2. As a sales person, I should be able to enter the product code by scanning the barcode
3. As a sales person, I should be able to enter the product code by scanning the RFID tag
4. As a sales person, I should be able to make a sale for the product entered
5. As a customer, I'd like to get an itemized bill for the products that I bought

I'd have 1, 4 and 5 built first and keep 2 and 3 as niceties. If the client pulls the plug after 3 stories, she still ends up with a system that let's her sell products. Now if you don't do it this way, your client ends up with a piece of software that can take in a product code in multiple ways but can't make a sale. Making a sale is the crux of the system (taking the POS example from ShareKhan's post).

Technical / Environment / Infrastructure tasks and the practice of tracking them as stories
I think my main problem is with calling these "stories", unless you can actually clearly define the value that they bring. Most of the times the value of such tasks is as vague / intangible as your NFRs. You know these tasks are valuable but it's a waste of time trying to find out and put down in words exactly what value they bring in. I don't even think that it affects prioritization of these tasks in any way.

That leaves one problem, tracking and I think we should track them as convenient to the project. Track them as cards, count them in velocity, deduct time spent on such tasks from development capacity. Anything that suits you and the client.

When do you decide that a story has business value?

We were discussing an issue with a client today. It was about getting some hardware installed that involved a series of tasks like getting racks, loading them, having the machines setup with some config and so on. The person in charge of this area was asking for stories to be created for each of these tasks "so that the work could be tracked".

The manager we were talking to was questioning the value of this with the logic that these are just simple tasks that need to be done by one small team to get the machines installed. How will having stories to track this help?

The idea of having a story is to track how delivering upon a small discrete requirement can generate business value for the business. Whats the point of generating more overhead with creating, delivering to and tracking a story for such low level tasks?

Then I thought, how do you quantify this business value per story? If the complete goal is to build a retail POS system, then what business value can I have by just getting a story to work where it lets the user scan a barcode and have the value register? If the rest of the system (and maybe 200+ stories) don't get done or don't work, then the miniscule "business value" of this scanning story is meaningless.

So then at what level does story breakdown make sense? To me, it now seems more of a development nicety to break it down so it fits in an iteration, give closure, blah blah.