Fabio Pereira

We have noticed on our current project that during retrospectives people usually raise good and bad things from the last days only. At the moment we have a 3 weeks iteration and it’s hard to remember what happened on the first days. I have already used from 1 to 4 weeks iteration and each project has its own reasons for choosing this length. To help the team improve around things that happened during the whole iteration, we have a box called the Future Retro Box, where people can all keep adding notes throughout the iteration.

The reason why the box is called Future is because it will be opened during the next retrospective. The items inside the box will bootstrap the retro wall when people can add other ones.

Make sure the box is in a visible point of the work environment.

Another option would be to have an on going retrospective wall, as suggested by my friend Lachlan. However, this might work better for a co-located team, we are distributed. We tried using a wiki for that… It didn’t work… No one adds comments to the wiki… Why? Perhaps the wiki is a Refrigerator as opposed to a Radiator. The box has been working fine so far. The notes are not visible, but the box itself is.

Technical Agile, IterationMgmt, Retro, ThoughtWorks

Although we should avoid as much as possible Technical Debt, most of the projects end up incurring some of it. However it is important to know “how much” we owe and how much interest we are paying.

Know your debt

Recently on our project we did a Technical Debt Retrospective. Basically we brainstormed on the wall all the Tech Debts we thought we had incurred since the beginning of the project. Then we decided to categorise them by Effort and Pain:

“The relative effort that the team would have to spend in order to pay this debt.“

“Pain is the direct impact on productivity that this debt causes. (interest)”

Keep paying down the principal, not only the interest

If we do not pay down the principal of our debt by refactoring our code we will always be paying just the interest, which means reducing the productivity. The question we asked ourselves was “Which ones should we pay first?”

Looking at the four quadrants of this wall we can identify which ones we can pay so that we can have the highest return.

One interesting thing that came out of our retro was that the High Return section was almost empty. We came to the conclusion that this was a good sign, it means that we had already fixed most of the LowEffort/HighPain debt during the normal development. Which is exactly how this type of debt should be paid.

After the retro we kept the Technical Debt Wall next to our Story Wall. From that day on it should become an ongoing wall that we will keep an eye on… Paying whenever possible…

Technical Agile, AgileWall, TechDebt

My colleague Jason Yip posted some very interesting thoughts on roles and hand-offs and I thought perhaps I could express them with some images… As I really like images better than words.
Jason mentioned that usually this is what happens:

Waterfall Project

Some Agile projects

They miss the point by assuming this is enough:

Ideally, this is what we want

“We’re not just re-ordering hand-offs, we actually want to remove them.” (Jason Yip)

Technical Agile, Collaboration, Communication, Photo, ThoughtWorks

Prioritization is crucial for a project to succeed. We need to make sure that we deliver what’s most important, what adds more value. The interesting and most challenging part of this is when we have to ask the business (client, product owner) to prioritize items. By calling them items I mean they are not necessarily stories, features or a product backlog. Sometimes we want to prioritize needs, purpose, goals and outcomes as well. The way we ask them is the trick…

I’ve seen many different ways of defining the priorities and usually there are ranges involved, sometimes with labels or numbers, like:

• Very High / High / Medium / Low
• Must Have / Should Have / Could Have
• Essential / Important / Not very important
• 1 - 5

There is nothing wrong with creating the labels, but when we ask the priority of one item only, that’s what happens sometimes:

How do we understand business priorities

The most effective prioritization exercise I’ve seen was to give the business a randomized list of items we want to be prioritized, then we ask them to give us the X most important items. And these will be the ones considered during the next timeframe. We should always let the list visible to the stakeholder who is prioritizing… It has to be clear for him that he is giving up on items in favor or others… at least for the next timeframe, which could be the next iteration of release.

It’s like a son’s birthday gift…

Good parents apply this rule with their children. Let’s say it’s close to my son’s birthday, I don’t have a son, this is totally hypothetical :), but my mother used this rule with me… So, I want to buy him a gift, ONLY ONE, of course, because I’m not the kind of dad that gives everything that my son wants. So I have a list of gifts that I know my son wants. Bike and video game are top of the list, but I can give him only one now, because it’s his birthday. We all know It’s close to Christmas (next iteration/release) and I will give him the other one on Christmas, but now it’s his birthday time, only one gift… I, then, make sure that my son understands the one gift per event rule and ask him which one does he want as his birthday gift… As a clever boy, he will think carefully which one really matters more to him because it’s clear that he has to give up on one in favor of the other. He will not give up on it forever though, it’s just until Christmas (next iteration/release), however, he knows that one of them will be bought and received first, and the other will come later…

As I said, there is no problem in defining labels and priority buckets, the only thing we have to bear in mind is that whenever we ask the business to prioritize items, we do not ask them in which bucket he wants to put one single item. And that’s when the principle comes.

Principle of Relative Priority:

One item by itself is usually important, otherwise it would not even be in the wish list… If I ask my son:

• Dad: “Do you want a bike?”
• Son: YES!
• Dad: “Do you want a video game?”
• Son: YES!

But when they have to CHOOSE… That’s when the most important items are picked… And during times of crisis, like now, with budgets constraints, we don’t even know if we will be able to afford a Christmas gift, so let’s make our kids happy during their birthday…

Technical Agile, Cool, GFC, IterationMgmt, Prioritization, Scrum

DRY is a concept applied quite broadly in code. In a web application it’s also important to extrapolate its use to the view layer (MVC) in order to increase clarity and consistency.

At the end of this post I’ll explain how a user story that had to be delivered became much easier to implement because we had a dry and refactored view.

Concrete example of a view pattern being refactored and dryed

As I usually try to explain things by example, I’ll show a JSP that can be refactored and dried. Please notice that the idea of refactoring and drying your view using conventions is technology agnostic and can be achieved with other languages and frameworks as well.

Let’s say we have to implement this form:

The example below shows one way of implementing the aforementioned form using Spring MVC JSP tags.

The convention used was that the default label will be label.${id}.

Tag Files and Conventions
If you want to know how to create your own JSP 2.0 tag file, like the one created above (form-ext:inputText), here are a few tutorials.
But always remember about the conventions over configurations.

Here is the file that was used to implement form-ext:inputText:

You can download it if you want: inputText.tag

If the user of the tag (inputText.tag) still wants to send the parameters label, size and maxlength, they would be able to do so. They are not required and have default values, though.

Basically if you apply these two concepts (Tag Files and Conventions) you can dry your JSP’s… But you have to identify the duplication patterns.

Extract Tag File - A view refactor similar to Extract Method

What was shown above is a step by step of a refactor that I call Extract Tag File. It’s quite similar to extracting a method, but instead of code, we have JSP.
I wish the IDEs had this refactor implemented … Who knows someday… Or maybe we could write a plug-in for Eclipse or IntelliJ.

How we applied this on a real project to deliver business value quickly

Recently, on my current project, we had to change the way the error messages were shown to the user. They had to be field context aware, which means that if there was an error related to the field “Given Name”, the error message had to be show next to the field.

This is what we had before:

This is what we had to implement:

Can you imagine how much easier it was just because we had our DRY VIEW? Our fields were wrapped and, therefore, it was just a matter of implementing this inside the tag file…

That’s what we aim… Deliver business value quickly!!! Happy client… Happy developers…

Technical Agile, Code, Java

We have adapted on our project James Crisp’s cool idea about changing the color of the command when the build fails or passes… Now it works for maven

mvncool.bat file

@echo off

color 07

call mvn %*

IF ERRORLEVEL 1 goto RedBuild
IF ERRORLEVEL 0 goto GreenBuild

:RedBuild
color 4F
goto TheEnd

:GreenBuild
color 2F

:TheEnd

It’s quite similar to the one that Crisp posted, so thanks Crisp for the interesting and reusable bat file.
Some small things are different for maven though… Check them out.

Step by step to have it working on your maven project:

• Download this file mvncool.bat (or copy it from the text box above)
• Move mvncool.bat to either:
     - the root of your project, where your root pom.xml is. You can also add this file to your version control so that everyone can use it.
     - or any folder which is in your PATH. It could be the same folder where mvn.bat is: MAVEN_HOME/bin/.

• From the root of your project, just from where you used to run your mvn you can from now on run mvncool instead. You can send the same parameters mvncool clean install for example and they will be sent through because of the %* on the bat file…

Before mvncool

mvn clean install
mvn package
mvn (anything)

Using mvncool

mvncool clean install
mvncool package
mvncool (anything)

Technical Agile, CI, Fun

During the Release Planning phase of a project we usually estimate and prioritize the MSL (Master Story List), also called Product Backlog in Scrum. The output of these exercises is a Release Plan which contains a vision of Time, Cost and Scope of the project. This phase usually happens for the first time during the inception, before the project starts. It needs, of course, to be constantly reviewed and recalibrated in a more fine grained level of details once the project is underway.

One of the main things we need to keep in mind at this time is that these are ballpark numbers.

Definition of Ballpark: A good numerical guess; an estimate.

In order for the decision makers to have an idea of these three variables (Time, Cost and Scope) and then make a decision whether to start the project or not, taking into account mainly ROI (Return on Investment). Most of the time they need these numbers for budget approval so that the project can start. The numbers are not precise at all and they will change for sure when the project starts. But as I mentioned before, they are an estimate, something close enough to the real numbers.
We should only spend enough time to get to a comfortable stage where a decision can be made to take the next step.

One of the clients I’ve been to had a very good way of deciding about funding approval according to the output of the release planning. They knew the numbers were not precise so they had a form with:

How long is the project going to take?

• 1 to 3 months
• 3 to 6 months
• More than 6 months

How much is the project going to cost?

• 10K to 100K dollars
• 100K to 500K dollars
• 1M to 5MK dollars
• More than 5M

The main point is that it’s useless to try and figure out whether the project will take 30 or 35 days at this time, of course we need funding and budget allocation, but nothing stops us from allocating X amount of money, try it and then continue if it works.

“Don’t aim too much before shooting, shoot and then adjust and then shoot again and adjust…”

Technical Agile, IterationMgmt, Scrum

Sometimes it is good to remind ourselves about the obvious…
“9 pregnant women CAN NOT ‘deliver’ one baby in 1 month.”

To understand the metaphor imagine:

Technical Agile, Fun, IterationMgmt

Retrospectives are usually conducted in order to keep the continuous improvement engine wheels working properly on a project. It is present on most of the agile processes I have worked with, such as Extreme Programming (XP) and Scrum.

The image below describes retrospective that most of us are used to:

On my last project at ThoughtWorks I had the opportunity to introduce to our client, together with Jason, a new type of retrospective, the Goal Driven Retrospective.

Goals and Actions

The Goal Driven Retrospective brings the idea that we should be much more focused on creating actions to achieve common goals shared and agreed within the team and also that will definitely add business value to the client.

Examples of goals:

• Zero bugs in UAT
• Productivity x 2: If a similar feature is required, the team should be able to implement it in at least half the time that they took at the first time
• Improve work environment

Examples of actions:

• Showcase and run selenium tests on IE to decrease number of browser dependency bugs in UAT
• Refactor JSP’s and create a DSL to specify exactly what varies from one story to the other
• Team lunch to improve work environment

It is essential that the person who is facilitating the retrospective, usually the Scrum Master, XP Coach, or Iteration Manager is aligned with the goals of the organization.

Values

In order to prevent conflict of interest, all the goals are defined based on values:

• Productivity
• Cost
• Quality
• Morale

By following these values the actions will make the team:

• Faster
• Cheaper
• Better
• Happier

All the actions must be aligned with all the values.

Some examples of actions that would not be acceptable because they go against values:

Where is morale?
What if someone suggests that in order for the team to be twice as productive they need to work 16 hours a day? If you just had that face and thought: “Come on, we know that this doesn’t work…” you are smart aligned with the values, but remember: this is not common sense.

Slow slow slow… what about productivity?
Or what if to achieve zero bugs in UAT we try to test and automate every single possible scenario in the whole application?

The picture below shows the output from the retrospective we had on the client:

I strongly encourage everyone to give it a go and try this new way of improving.

Technical Agile, AgileWall, IterationMgmt, Photo, Retro, Scrum, ThoughtWorks, XP

If you won the lottery and you saw 10 million dollars in your bank account tomorrow, what would you do? You would probably buy a lot of things… right? But actually you do not need everything you would buy, do you? Ok, you would have a lot of spare money, so there’s no problem.

But what if you had only a hundred dollars in your account? What would you buy? You would have to think about it more carefully and prioritize.

The product owner needs to define what is really important for him.
If you ask him what does he want for a 2 years project (10 million) he will say: “I want everything!”, but if you break down the project into small iterations (sprints) and ask him to prioritize what does he want for a small chunk of time, that’s when the important things come up.

A good way to show the product owner that he does not have 10 million available is making him buy features with a limited amount of money.

How does it work?

Before starting the poker chips session, we need a list of estimated user stories. We call it Master Story List (MSL), or Product Backlog. There are many ways of facilitating an estimation session, I like planning poker. But always remember: The team gives the estimates.

Basically you need to define how much a story point costs and give a certain amount of poker chips to the client according to the size of your iteration. And then ask him to buy what he wants for the next iteration.
It is interesting because from my experience as long as he has money, he buys everything, but then when he starts running out of chips, that’s when the real process takes place.

The output will be a list containing the most important user stories at that time.

Technical Agile, IterationMgmt, Photo, Prioritization, Scrum, ThoughtWorks