Ajedrez - antoniomartel.com

Archivos de Categoría: English

How we used Scrum in our software projects

These days I have been preparing and updating a Prezi presentation about how we used Scrum (or Agile development) in our software development projects.

You won’t find there a Scrum theory implementation but the process and methods we used to try to become more Agile (whatever that means) and to deliver software regularly and frequently (working software).

Reach the presentation at Prezi: Practical Scrum – How we used it.
or have a look at it here:

Functional tests with Selenium WebDriver

(Sorry, this is a technical post about software testing, but don’t worry, you can find here some of my non-technical blog posts)

A few years ago we used Selenium IDE in a web application project just to check the functionalities developed at that sprint were working as expected but there were some problems with that. I try to sum up them here:

  • Selenium IDE works only with Mozilla Firefox web browser. That was a problem when you did need to test another browser or your web app worked only in MS Internet Explorer.
  • Tests and tests suites used to become red frequently. Firefox updates (on security mainly) made script commands deprecated easily so it was needed to search for a replacement, when available, or to find a workaround (after a few hours in Stackoverflow). To keep them working in order was costly (and time consuming).
  • There was some kind of a mess with script commands: type or typeKeys, click or clickAt, mouseAt or mouseDown. You never knew.
There were also some advantages, of course, like the way the user actions could be recorded by the IDE. You just had to click somewhere, move the mouse or type something that your movements were recorded by the IDE as script commands. It was needed to adapt or fix them after recording it but it used to made things easier. 
These days I have been playing around with Selenium WebDriver. It doesn’t need to inject javascrit code into the web browser and you can use your favourite programming language to write the tests.
I have written a couple of tests to check one of my web applications just to show you some of its main features and how Selenium WebDriver can work on Chrome. At the end of this post you can find a video showing how these tests work.

This first test is a kind of smoke test just to check that the application is working properly and shows the main page when it is requested:

Below you can find how to get the driver variable for Chrome. Using it you can have access to all WebDriver commands. It is written inside the jUnit @Before annotation so it is going to be executed before any other test run. At the @After method there will be a driver.quit(); sentence.

This last test actually checks that 10 questions are asked at each attempt (no more, no less), chooses randomly one of the multiple choice options and ensures that there is a results page at the end.

This is the Selenium/JUnit tests video recording:

You can find texts like this and many other about how to manage agile projects in my book Agile 101: Practical Project Management (available on Amazon).

References:

Agile 101: Scrum’s sprint 0 and the initial design

Sometimes we’re in projects in which we need to prepare a large number of things before starting out. Can we start building without stopping to plan or analyse what we’re about to do?
When we read about Scrum, we get the impression that all texts start directly with development, but…what about environment, team or office preparation tasks? What about design? Shouldn’t we stop to think about the whole design and architecture of what we’re about to build before we start programming?

About preparation or “before” programming. Some people call it “sprint #0”, the sprint where you prepare your tools, install the stuff you need, train the team, etc.. It can be a longer sprint (3, 4 weeks or so). Some detractors criticise this sprint because it is not agile, and because it hides some work that is taken care of in the same way as the rest.

It is true that it can be very comfortable, because we tell the client that we’ll be back in 3 weeks to have the preparation meeting for the first sprint, and by then we have organized it all. I have also used those zero sprints in my first projects, but now I prefer to solve those issues also in an Agile manner.

I try to make every one of those tasks (installing the development server, installing the demo server, creating a project and tasks in Trac or Redmine, installing the Integrated Development Environment or IDE, etc.) part of the first sprint and, if possible, have a set deliverable at the end of each demo.

In that demo, we could show, for example, the project-management tool already installed in the final URL and the task list with the tasks ready to be assigned, or we could launch the IDE in that demo meeting, and check whether the environment has been created correctly and whether Jetty executes a “Hello World” application correctly.

Another important step is the initial architecture design. If we make a huge design with all the application architecture (known as a big up-front design) before we start programming, we will be returning to a waterfall design. Firstly, we analyse everything, three months later we do all the design work during a couple of months, and then we program everything in one go until we finish it six months later. If we finish all work and show it to the clients and they say: “that is not what we discussed a year ago…” then we have a problem.

It is better to design one of the modules of the application, or one of the new features, then draft the specifications, and finally send them to development. While part of the team implements them, analysts can gather requirements for another module or another feature, and go on doing so sprint after sprint.

There will be sprints in which we won’t have new stuff to analyse and others in which the development goes a bit faster than the specifications, but normally there is always some task ready to be carried out while Product Owners sign off the latest requirements, or until those requirements are completely defined.

If we work like that, new modules or features could be regularly deployed to production or preproduction, so users will be able to take advantage of them without waiting until the very last feature of the application has been designed.

You can find texts like this and many other about how to manage agile projects in my book Agile 101: Practical Project Management (available on Amazon).

Translation by Begoña Martínez. You can also find her on her LinkedIn profile. Proofreading by David Nesbitt.

Agile 101: Scrum for fixed-price projects

One of the most common questions we ask when we’re wondering whether we should adopt a methodology like Scrum is how to face the fixed-price contracts that are so common in public procurement and many other sectors. One of the Agile principles says: “we welcome changing requirements, even late in development. Agile processes harness change for the customer’s competitive advantage.” This means that using Scrum we will be flexible with features we must develop, and we will admit new ones that were not in the original contract listing. This sounds very good but… the final price of the project is not going to change!


I work in the public-procurement department in a software-development company. That means that in at least 90% of the projects I work on I have a public institution as a client. Contracts are normally defined by a document with specific administrative specifications and another one with technical features. These contain all the services that are being hired, the characteristics of the product to be developed and the maximum price of the tender. They include even the technical features of the software and the profile of the projects’ participants. It is all very narrowly defined. It is the price that the public administration must pay to avoid arbitrariness.

If our company has been lucky enough to win the tender, then both of us, provider and client, have to face the harsh day-to-day reality of projects. Circumstances change. New legislation comes up that affects the project. Things that were deemed correct a few months ago are no longer valid, because needs have changed. What should we do? If we adapt the project to every change that the client needs, then we providers have to bear the brunt of additional cost. If we refuse to change anything for the client, always waving the contract and its specifications in their face, we risk delivering something that might be perfect for last year’s decree, but—for that very reason (or for any other reason)—is no longer of any use.

The way to avoid this, or at least the way that has worked best for me so far, arises from the way of working in Scrum projects. At the beginning of the project I make a list with the features to be developed, and give them a score that the client knows right from the beginning.

I normally give the example of a large jar full of ping-pong balls. Each ball is a feature that must be developed. The jar is full. If the Product Owner wants to include two new features, I ask that person to tell me which other features—or ping-pong balls of a similar size—should we take out of the jar. Normally this vision solves the predicament and the Product Owner does not resist giving up other features, now recognizing that they aren’t important. If it all gets recorded in the meeting’s minutes and, most important of all, if the Product Owner understands what is being given up and what will be gained instead, it is normally a comfortable solution for both sides.

You can find texts like this and many other about how to manage agile projects in my book Agile 101: Practical Project Management (available on Amazon).

Translation by Begoña Martínez. You can also find her on her LinkedIn profile. Proofreading by David Nesbitt.

Agile 101: Scrum in a maintenance project

There is a common doubt among those considering using Agile methodologies to manage a project: would it be possible to apply Scrum to projects where, on top of the upcoming and planned tasks, there are frequent interruptions to solve maintenance problems, to correct errors or to resolve issues? Many project managers face these types of problems and use several approaches to tackle them. Let’s see some of those approaches: 

Short sprints

With this solution, we will keep the tasks that have already been programmed for that sprint. If our sprints are 1 or 2 weeks long, we will be able to make the unplanned tasks a priority on the to-do list for the next sprint. If the sprint is one week long, we will start urgent tasks after 3 days, on average.
In an ideal world, this would work but it would pose some disadvantages. Not every issue can wait for a few days to be solved. A whole service could depend on it.

 

Low load factor

If we know that as a general rule we will have unplanned tasks or issues that we must solve quickly, we can lower our load factor during the sprint so we have “room” to solve these problems.
If in each sprint the team has the capacity to solve 10 planned stories, we could promise to deliver just 7, so the team has time to solve urgent issues. That way we won’t fail when it’s time to deliver what we promised, sprint after sprint.

In my opinion this solution can be useful in some projects but it could create another problem. Let me explain. Normally the load factor is 75%. If we lower it so we’re able to devote 30% of the team’s time to urgent tasks, we should apply a load factor ranging from 40 to 50%. On this 40% we’re using Scrum but, how are we managing the rest of the team’s time? What do we know about that pile of tasks that we’re solving sprint after sprint?

 

A different team for each type of task

This solution consists of having a Scrum team for the identified and planned tasks, and a Kanban team for the urgent issues. With Kanban we can add new tasks to the TO-DO column as they arrive, and we will follow them up until they get shifted to the DONE column. With this the team will be even more agile, reducing the overhead in meetings and planning that we could suffer with Scrum.
We still have a dilemma with unplanned tasks, which are, well, unplanned. There will be times when the Kanban team providing support will be oversized, because there are very few issues. But then a week later the number of issues and their importance could be so high that we will need any help we can get.

To minimise those risks, we could rely on the previous solutions. We could have short sprints, so the Scrum development team can plan the tasks for the next iteration. We could also use a lower load factor so the team can have a bit of room in their planning to give a hand if needed. In the same way, the support team can help with planned sprint tasks when their TO-DO column starts to look empty. To make the most of this, it would be good if all members in all teams rotated, so that everyone knows every aspect of the job.

You can find texts like this and many other about how to manage agile projects in my book Agile 101: Practical Project Management (available on Amazon).

Translation by Begoña Martínez. You can also find her on her LinkedIn profile. Proofreading by David Nesbitt.

Suscríbete

Esta web utiliza cookies propias y de terceros para su correcto funcionamiento y para fines analíticos. Al hacer clic en el botón Aceptar, acepta el uso de estas tecnologías y el procesamiento de tus datos para estos propósitos. Más información
Privacidad