Five capabilities of a successful DevOps team

So you have gathered developers and operations and put them in a team, voila DevOps! Or not? What kind of things does a DevOps team need to know to be successful? Are they developing safe software? Let's see.

Metrics

A DevOps team needs to measure everything, they are responsible for everything, so they need to know about performance. Performance of the application they are responsible for. Performance of the team to be able to improve. And performance of their delivery of value to the customer. Here are some basic metrics I think are important to measure.

• Lead time of a change: how long did the change live before it was put into production?
• Mean time to recover: how long do most outages take to be resolved and availability is restored
• How predictable is the team: prognosed work / delivered work (the closer the numbers, the better).

Security

A DevOps team owns the security aspects of the application, no one else is responsible. Things like static code analysis, OWASP Top 10, basic thread modeling and service accounts should be understood.

Testing

Testing is not the job of one person in a DevOps team. All disciplines must have a fair understanding and even an implementation of testing. Think about the testing pyramid and accompanying strategy. Think about unit testing on at least all new functionality. And least but not last; knowledge of disaster recovery or use of chaos engineering.

Transparency

Being transparent as a team means to show what you do and what you didn't do. It is about communication with all the stakeholders, including the DevOps team itself and management. A visible backlog and open review sessions are key to transparency.

Learning

In order for a DevOps team to be successful and stay successful they need to learn. While the world keeps moving, a team cannot stand still. Any lack of learning will result in a form of debt. New technologies appear, team members can change, the organisation will change and dependencies change, just to name a few. DevOps teams actively need to learn and develop their skills as they do their applications.

This is just the tip of the iceberg about what a DevOps team needs to be able to do, there are many more. But on a basic level you need to understand and own the above. With an increases autonomy that comes with being a DevOps team also comes more responsibility of doing the right things right.

What are the most important capabilities for your team?

Kano vs WSJF

How do we decide what to put on the top of the backlog? How can we prioritise one story over another? I have seen many different ways, here are a couple:

1. low hanging fruit: pick the easiest items that can be solved fast. These are not always the most important ones and can lead to technical debt or dissatisfaction of the customer.
2. The loudest yeller: in a group, there is one stakeholder that yells the loudest about their item, silencing the other stakeholders. This leads to a dishonest balloting of items and doesn't put the most important items first. 
3. The managers priorities: in traditional or hybrid organisations that are adopting scrum, managers are about people and projects. This can lead to managers making priorities instead of the stakeholders or customers and thus not delivering the most important items first.

So how can we do a better job? I want to mention two: the kano-model and WSFJ.

Kano

The Kano model is an approach that prioritises stories based on customer satisfaction plotted against functionality.

• In the left top customers is the functionality that brings excitement to our customers. If you invest in those, your customers will be delighted. Mind that not implementing these doesn't lead to customer dissatisfaction. A minimal viable product will not include these functionalities.
• On the right bottom are the basics for the service to work. These are basic functions which are taken for granted. Customers won't be satisfied by these functions, but will become very dissatisfied if they are not implemented.
• And on the top right are the performance functions. The more you invest in these functionalities the more the customers like it. The better the performance the happier the customer.

You can read more about the Kano model on the website of ProductPlan: https://www.productplan.com/glossary/kano-model/

WSJF

WSJF is an abbreviation for Weighted Shortest Job First. I came upon it during a SAFe course. It is based on the idea that we can weigh stories on importance and work and get the most economic benefits. For WSJF we need a bit of simple math. First we need to find out what our CoD (Cost of Delay) is. So how much money will we loose when we don't implement this functionality? We plot that against the job size with the formula: WSJF = CoD / Job size.

It is easiest to make a table with all the functionalities or stories or stuff you want to make, add a column for size, CoD and weight (which is CoD / Job size). After plotting this in the table, the item with the most weight is the most important to implement.

You can read more about WSJF on the SAFe website: https://www.scaledagileframework.com/wsjf/

How do you do your estimations? Which model do you prefer and why?

Five signs you know you are working in an agile team

Working agile is more than just a daily standup, but you knew that already. What are more distinct features of working in an agile team? Here are a few of them.

One: everybody is a developer

Everybody in the team has the same goal: getting stuff done. The distinction between your job and that of you peer is fading as you progress to developing stuff that works for your client. Everybody in the team shares the same responsibility. You do refinement together, work together in a peer setting most of the time. You share knowledge and you even take over from other disciplines in the team.

Two: you are the A-team

The team is more important than its individual members. You avoid local heroism with people that know it all. The best hero is the one that shares the way stuff is done. A hero is the one who shares most and educates the team on development principles. A hero helps all team members in getting a better performance so that all members become hero's. 

Three: reflection together

Reflection is done with the team. Together you reflect on what has been done in the review, feeling proud of what you, as a team, have achieved. You inspect what you failed to deliver and know what you need to change to do an even better job next time around. With the team you inspect how you are working and come up with things that increase your performance as a team and as an individual. You do peer reviews to inspect each other in an honest way so you learn and are able to improve. This implies no functional reviews by managers as well.

Four: dashboarding

As a team you use dashboards to make performance visible. Just like you use a dashboard in the car to read out speed, fuel and overall car performance. You have stopped sharing status in the form of reports or meetings. Reporting the old fashioned way has no added value to the customer. If anyone in the company wants to know how your team is performing they come to reviews and ask.

Five: working on products

The most important indicator of working in an agile fashion is perhaps that it is focused on products instead of projects. Your team works on a product with a distinct customer base. You and your team are solely responsible for the product and its success. A project is timebound while a product is customer bound.

What are the most important indicators for you? Please write them down in the comments.

Make it smaller

To learn faster, to deliver more value and to have a better general performance we need to make stuff smaller. Little stuff is easier to oversee than big stuff. Small problems are easier to fix than big problems. What can we make small in our work to make our work easier?

Shorter sprints

The shorter our sprints are, the more moments of inspection are available to us to learn and improve. A sprint is between two to four weeks according to the Scrum Guide. We don't need to adhere the framework here, if we can improve faster by shortening our sprint, we respect the framework as well. There are teams that run weekly sprints, and some even run shorter ones. A shorter sprint will provide more inspection moments and thus result in faster learning through more inspection and adaption moments. Because all events (except the standup) are relatively sized to the sprint, it will not consume more time.

Smaller PBIs

The smaller our product backlog items are, the faster they can be build and delivered to the customer. We get more revenue and more feedback. PBIs or stories are sometimes really big, they tend to have epic proportions. An epic PBI is one that can not be overseen or fixed in a certain amount of time. For most teams any story that takes more than three days to fix, the story should be considered an epic. So cut up the story into smaller chunks, make them independent of each other and smaller than three days. This might be hard at first, but practice makes perfect; think about the smallest possible change that might work. 

Less hierarchy

Hierarchy makes stuff complex. Hierarchy results in interdependence of PBIs, while we would like to have PBIs that we can resolve without any relation to another change (check the INVEST acronym). So an epic PBI, which is a PBI of epic proportion (and nothing more), should be made smaller. This will result in a couple of independent stories and the original epic story should end up in the trash. There is no added value in the original epic and it will only clutter the backlog with relations between stories that are useless to the outcome, because it's redundant.

There is another form of hierarchy, which we experience daily, that can decrease our performance as well. This is organisational hierarchy. The more autonomous a team is, the faster it can make decisions and learn from its mistakes. In a hierarchy people that don't do the work are made responsible for that work. Weird isn't it? Responsibility and autonomy go hand in hand. Make teams more responsible for their work in giving them the authority to make decisions about their work. Scrum solves this problem with two accountabilities: the Product Owner is made accountable for the work, the Scrum Master for the performance. 

Shorter meetings

I hear a lot of people complain about too much meetings, and too long meetings. I help them reorganise their agenda by skipping meetings that don't add value or performance. 

From a Scrum team perspective there are a couple of meetings: the daily; in which progress towards the sprint goal is shared. There  is refinement; in which stories are scrutinised with stakeholders until it is clear which work needs to be done. Each sprint there is a review, a demo and a planning session in which is talked about what is being or going to be made. And there is a retrospective in which the team introspects how it does the work. 

No other meetings are needed!

Many people find themselves in a hybrid form of Scrum with archaic roles and responsibilities in which they are obliged to attend some meetings. There are three types of meetings: informational, decision making and cake. The informational are most of the time unimportant and could be done with a memo; when you are invited think about what you are going to get out of the meeting and decline if it is not enough. It is no shame to not waste your or anyone else's time. The decision making meetings are a bit more important but again, choose to go if you think you can gain something by attending. In the meeting, steer towards decision making and go when done. A one hour meeting can mostly be shortened to half an hour by this approach. The last meeting is the cake meeting, this is the fun meeting, this is the one you should always attend!

To summarise: make everything smaller. Small work items, short meetings, less hierarchy and short sprints. This will increase performance and added value. Sometimes it needs a bit of courage when you need to make choices. But it is worth it. 

Let me know in the comments what you do to make your work smaller.

Why do we want to achieve standardised work?

We standardise to sustain good practises, make audits less painfull and make our work sustainable. Standardisation is not the goal, sustainability is. But how do we get from our current way of working to a more sustainable way?

"To throw away what you no longer need is neither wasteful nor shameful." - Marie Kondo

We start with sorting out what we don't need. We clean our work process and get rid of clutter. Anything that does not contribute to the creation of value, we remove. Ask with every step of the work process if it is really needed; what would happen if the step is removed? When you create a work item, think about the minimal product you are happy with.

"A dramatic reorganization causes correspondingly dramatic changes in lifestyle and perspective. It is life transforming." - Marie Kondo

The next step is to reorganise. To make it easy to work, to reduce movement, lookups and extra meetings. We can improve on this by refining the work we do and collect or link all information on a work item to that work item. We stop going to meetings that are not helping us to add value to the product we create.

"Visible mess helps distract us from the true source of the disorder." - Marie Kondo

To keep stuff organised, we regularly clean it. This means looking at our processes and stuff we make, and clean out the clutter. Did we took a short cut somewhere? Fix it. Do we need to refactor to improve? Do it. Stuff that is not maintained will cause headaches in the future. And yes, it takes time to do this, but it will take up more time down the road, when it gets in our way.

Now we come to the part where we can standardise. We make rules about the previous steps. We define our way of working, our definition of done, our test strategies, our meetup schedules. We make rules about how we as a team work. And we publish the rules to make it transparent to everybody how we work.

"The space in which we live should be for the person we are becoming now, not for the person we were in the past." - Marie Kondo

Sustainability is the ability to keep a pace over a longer period of time. In scrum this long period is called a sprint; perhaps we should have called it something else, because we don't sprint like sprinter. We are not trying to achieve a new world record on the 100 meter, but we are trying to create stuff for others over a longer period of time. We are trying to set a pace and slowly improve on that pace over time. We do this with the previous mentioned steps. We aim for goals and get there with small steps.

What is the best team size?

Some say a team of nine people is more than enough. Some argue that a team of fifteen works fine as well. My argument is that the smaller the team, the better their performance. But who's right and why?

The bigger a team is, the more interconnected it is, the more time it needs to discuss and agree. Each new member adds extra interactions to each other member, growing interactions exponentially. Brook found this out in the 1970's and created a formula for it.

Cognitive load is the amount of stuff a team can handle. The Paas-scale can be used to determine how much the type of work effects the load. The lowest number on the scale is simple work that costs very, very low mental effort. The highest number on the scale cost very, very high mental effort. 

Depending on the work, it requires more or less mental effort. What would happen if we plot the numbers? Let's have a look. 

The green line is Brooks law. It starts at zero and curves up, like one would expect of an exponential formula. The line tells us that with a team of three, there are three interactions. It also tells us that a team of fifteen will have 105 interactions.

The red line represents low mental effort work. From the graph we can tell that the size of the team doesn't really matter in this case. It almost seems that interactions can take place while working.

The yellow line represents very high mental effort, it crosses the green line at three. So interactions will very much effect this type of work, suggesting a team size of a maximum of three.

What about the recommended team size of "ten or fewer people in general" according to the 2020 Scrum guide? For average mental effort work a maximum of eleven people should be in the team. If the work is mentally harder, the maximum team size shrinks. This means that the team size by Scrum isn't very far of what the numbers say, but it is a maximum team size, not an average! 

It is fair to say that any member above eleven won't increase team performance significantly. Splitting a team with many members into two or even more teams of five to six will boost performance again. The cognitive load is halved per team, but interconnections have gone down by a factor resulting in an increase of performance.

Software development is an above average mental effort, making teams too big is very counterproductive. When you aim for a high performance team, go for team sizes of about five to six people.

Let me know what your thoughts are about team size in the comments below.

Time management with Scrum

One of the goals of Scrum was to get rid of meetings. With meetings I mean unnecessary ones. The only thing that counts is delivering value at the appointed moment so users will be happy. How does Scrum help with effective time management? Scrum helps time management with events.

Scrum events

There are five Scrum events: the sprint, the planning, the daily, the review and the retrospective. In the sprint everything happens: building, refining, testing, delivering. In the sprint all the actions take place to deliver value to the customers. During the planning event, all the work the team thinks they can do in one sprint, is planned: a workplan is thought off, a goal is defined and work items are selected to be completed in the sprint. During the daily standup developers (everybody is a developer in the team) discuss progress and impediments on the defined goal. In other words, they meet to see if they are going to achieve what they have promised all the stakeholders (that includes users). At the review everybody is invited to inspect what has been done during the sprint. Possibly a demo is given to show what has been made. The last event is the retrospective; a moment of reflection by the team to inspect and adapt on the way they do things.

Time spent

About 22.5% of all time has been claimed by Scrum events. We need another 10% per day to learn. That leaves 5.3 hours of actual work time. This doesn't sound much, but this is net time. So all the overhead is gone. We should be able to work for a bit more then five hours a day without interruption. 

Let us look at some reasons why we cannot spent this time and on some things that we can do to spent our time more effectively.

Pitfalls

One of the biggest pitfalls is that we are going to listen to other people outside of our time and outside of planned events. The social part is fine, we need that as humans. But the part where it as about work is wrong, there are events and a PO for that. When developers approach a stakeholder to verify something it is fine, then it is part of the developers job. But the other way around will cost time instead of saving it; therefor decreasing performance of the team as a whole.

The second most heard excuse of not working on items from the sprint is that a person is needed in a meeting. There is no meeting where a developer is needed on something that is not on the sprint log. And if there was one, the developer should be the initiator. I've seen developers that could not do their jobs because their expertise was needed. By pulling out a developer, the developers flow will be disrupted, causing a delay of a couple of hours to a days work. Some developers are so busy in attending meetings, that they cannot do their jobs anymore. Attending meetings will make the team loose focus and decrease team performance.

The last pitfall I will mention is to not use all the Scrum events. Some teams cut short on planning, review and retrospective. Not paying enough time for planning will result in extra planning during the sprint, in a change of calculated hours, in a deviation of the sprint goal and result in unpredictability to the stakeholders. Not paying enough time to inspection and adaption events like review and retrospective will stagnate learning of the team and will guarantee performance will never improve.

Good practices

A way to be more effective is to go pair programming. At first sight it seems that it will cost time. But ever since Kent Beck wrote about it, it has been proven over and over again that working in pairs improves quality of the work disproportionately causing less issues and less technical debt, opening up hours to add features.

Sharing knowledge in the form of a center of practice (run by developers) adds to quality of work and to speed of delivery. It will cost time in the form of a meeting. So run the meeting as an Open Space, free for everybody to join if they have anything to learn or to contribute. This is a very cost-efficient way of gaining lots of knowledge and improve quality of work, resulting in less rework and an improvement of performance.

The last tip on spending time more effective is to talk about it during the Daily Scrum. Talk to each other about the work one is doing and find a way to finish the work today. This will increase focus on getting the work done, it will put importance on it. Making your work more important than meetings makes it also easier to say no and stay focused.

Share

Let me know what is holding you back to increase performance or let me know what you did to get things done in the comments below!