I have been using the Shift Left and even Shift Right terminology, as we say in Dublin, for donkeys years. I am not sure who I heard it from, I seem to vaguely recall a conversation with Gojko Adzic, or was it Dan North? I am getting more and more forgetful, whomever I pick I am very likely wrong, so forget about where i heard it from. and let’s move on.
I love it, it’s leaner!
I have to say I love Shift Left, it has helped me and my teams understand where we needed to take our attention to. In my mind it was never only about redistributing testing activities, but it brought great emphasis on prevention over detection, going towards a leaner testing, some people like to call #NoTesting.
I am all for prevention and I was always going to end up with the Shift Left party.
Now, reading what Janet says, I understand perfectly where she is coming from and there is no denying that the continuous infinity loops models describe the new reality much better than Shift + direction does.
Was it always like this?
Software development today is certainly non linear 100% agree.
Now correct me if I am wrong here, but waterfall is linear.
The water goes in one direction and there is no going back unless we are on some planet with different gravitational laws, which would be pretty cool to see now that i think about it 🙂
The key is “what” are you shifting from?
My take is that, maybe, in the early days, testers suffering with a bad waterfall hungover, would have understood Shift Left with more ease than if presented with the new continuous models (that in fact, didn’t exist yet).
We are improving
Think about being a waterfall tester in the mid late 200x. You are in a room and I talk to you about how to think and use Shift left. Somebody else maybe Janet, talks to you of either of the (excellent) continuous models we have today. What do you reckon you are going to do next?
If I had to guess, I’d say you’d Shift left. It is easier to understand if all you know is waterfall.
There is a left, a right and a centre in waterfall and people get it when you ask them to go left.
It was the first step to move from waterfall to agile, at least for me.
I see the continuous models as the reflection of the growing maturity of modern testers.
We are ready
We are now ready, bring the new continuous models on and let’s leave the shifting to rally drivers.
The models are good now, because testers are ready to understand, embrace and apply them.
A lot of people say yes, my personal experience says not.
My personal experience tells me that the most effective and efficient approach uses a test coach that slowly makes himself scarce. My experience focuses on creating competencies for completing an activity (testing) in a collaborative context.
One aspect I am finding difficult to explain is the impact on queues of a test coach approach, I will use a series of Scenarios and pictures to facilitate my reasoning.
If you feel like substituting activity A = development and activity B = testing feel free, but this approach is activity agnostic in my experience as I have applied it to analysis, and development obtaining similar results.
Scenario 1 – Test specialist
In the presence of one specialist on Activity B and many specialists on activity A. Problem: Long queues form in Ready for Activity B (Case1) or worse specialist multitasks on many activities at the same time slowing down flow of work as per Little’s law (Case2)
Scenario 2 – Test coach
Stage 1: In the presence of one coach on Activity B and many specialists on activity A Initially coach pairs on Activity B with people that do activity A. This way we obtain 2 benefits:
Queue in “Waiting for Activity B” is reduced as one person normally performing activity A is busy pairing with coach on one activity B task
By pairing on activity B feedback loops are shortened
Person with activity A acquires new skills to perform activity B from pairing with coach
Quality of activity B increases as it is a paired activity
Flow improves because of 1 and 2
Stage 2: When cross-pollination of skills required for activity B starts to pay off, we have 2 benefits
Normally some activity A person will show particular abilities with activity B, in this case this person can pair with another less skilled activity A person to perform activity B
Queue in “Waiting for Activity B” is reduced as more people with activity A skills are performing activity B
Flow of value improves lead time decreases
More activity A people get skills on activity B
Stage 3: All activity A people are able to perform activity B Activity B Coach can abandon the team to return only occasionally to check on progress. Benefits:
Activity A and activity B can be performed by every member of the team
the WIP limit can be changed to obtain maximum flow and eliminate the queue in Ready for Activity B.
The flow of value is maximised
The lead time is minimised
WARNING: I have applied this approach to help teams with testing for many years. It has worked in my context giving me massive improvements in throughput and reduction of lead time. This is not a recipe for every context, it might not work in your context, but before you say it won’t work, please run an experiment and see if it is possible.
This is not the only activity that a good test coach can help a team on, there are many shift left and shift right activities that will also reduce the dependency on activity B.
I have been told a million times “it will never work”, I never believed the people who told me and tried anyway, that’s why it worked.
Try for yourself, if it doesn’t work, you will have learned something anyway.
Previously on A Knowledge Worker’s Tale: A large bottleneck appeared in the exploratory test column of team X’s kanban board, but only Rick and Jane, the testers, seemed to give a damn about it working hard to unblock it. Gus saw the problem asked the team to take off their headphones, go to a room and get instructions from the testers on how to unblock the flow of work.
And the story continues…
Team Xers reappeared in their open space with renewed energy. Rick and Jane had explained what they needed to get their work unblocked and soon people started working in pairs. Rick and Peter were debugging together an issue that seemed difficult to track with the assistance of Fritz the business analyst that seemed positive he remembered something important that could help them. Jane was showing to Mike and Jimmy how, what she thought was the correct customer journey, was not what Zach the customer thought it was.
It was great to watch, a group of skilled professionals working in groups, using all their skills and knowledge to troubleshoot a complex problem.
As expected, the pairing and collaboration gave its fruits, defects were getting fixed and retested and the bottleneck was reducing fast. Flow was being reestablished thanks to the team swarming and focusing on the biggest problem in the system, it was great to observe.
Once things got back to normal Mike called the team to the board and said “Folks, it has taken only 3 hours to fix that bottleneck that had basically stopped our flow for two days, I think we have learned a very big lesson today and the lesson is that when we see a bottleneck, we all have to chip in and help unblock it!”. The other guys nodded with satisfied faces finally seeing their work getting to completion again. “Well done all!” Rick the tester added “without your help, we would still be where we were 3 hours ago, collaborating has been fantastic” and the lads gave themselves a small round of applause.
“Let’s go back to work!” said Mike.
“Hang on, hang on” intervened Gus, that had been lurking from distance. “It is great that you guys found a way to reestablish flow, I was delighted to see how you collaborated and forgot about your roles for the good of the customer, I say well done for that”. He continued “The fact is, if we go back to work without addressing the problem that created the bottleneck in the first place, it will be only a matter of time before it happens again, don’t you agree? Now that we have it fresh in mind, I suggest we try to understand the real problem that created the bottleneck, let’s do a quick retro, who’s with me?” The team nodded and followed Gus in the retrospective room.
“So WHY do you guys think the bottleneck formed at exploratory testing?” asked Gus. Immediately Peter jumped in “Because we need more testers!” to which some of the other team members nodded in unison. He added, “Rick and Jane are always super busy, they never have a minute free and at the same time they are always behind with work, the guys need help, we need more testers!”
“Adding more testers is a potential solution, but we haven’t identified the problem yet” said Gus. “I am going to ask a tester now, Jane, WHY do you think the bottleneck formed at exploratory testing?” Jane, slightly caught by surprise turns to Rick, mutters something, then says “Because exploratory testing was taking too long”. “Ok this sounds like a reasonable problem, and WHY was it taking too long?” added Gus. “Well, running tests would have been fast, but we found a lot of defects that needed fixing and that created a never ending back and forth with the developers and Zach” added Jane.
“Very good Jane, thank you, we are getting somewhere now” said Gus, and continued “so it appears we had too many defects in the stories you were testing, so WHY were we having so many defects?”. A long silence ensued.
Peter broke the silence and said “For my user story, it turns out I didn’t understand clearly what Zach (the customer) wanted, so I had to change it four times”. Jimmy added “Yes, the same for me, this feature is complex and the work me and Mike were doing had to be changed many times, every time Jane or Rick would find something wrong because they’d ask Zach and he changed his mind, yet again”
Gus smiled and went “Oh Oh Oh, I think we might be getting somewhere here folks, what do you think Mike?” Mike was deep in thought, head down. He stood up and said loudly “I get it! It looks like a communication problem early in the process presented itself as a visible problem in exploratory testing. It’s got nothing to do with testing!” He was truly excited by the revelation, then he added “The root cause of the problem is the bad communication between the team and the customer and the symptom shows up in exploratory testing.”
Gus said “That’s an excellent observation Mike, do you guys think that if you clarified your doubts with the customer before you started working you would have saved the time you have spent reworking?” Everybody in unison went “Yeah!”. Well, now we know what the real problem is, I’ll chat to you tomorrow morning at the stand-up and we’ll find a solution together, it’s getting late now. Great work everybody and well done for getting to the root of the problem with only four why’s, I normally need five…”
Gus stood up to leave but stopped mid-track and asked “Guys, so do we still think we need more testers?” Peter was first to answer “No Gus, you’re right that was a knee-jerk reaction, thinking about it ,we need more clarity, not testers”
Asking why enough times to get to the root cause of a problem is a great technique, thought Gus while leaving the office and heading to the pub.
A defect is anything that threatens the value of the product.
Before we start, let’s agree that:
we don’t want defects that threaten the value of our product
we want to give our customers as much value as possible at all times.
If you don’t agree with 1 and 2 then, don’t waste your time and stop reading now.
Testers are normally associated with finding defects. Some testers get very protective with the defects they find and some developers can be very defensive about the defects they wrote. Customers don’t like defects, developers don’t like defects, product managers don’t like defects, let’s be honest, nobody likes defects besides some testers.
Why would that be? The reason is that the focus of a lot of testers is on detecting defects, and that’s what they get paid for in a lot of organisations. If you are a tester and love your defects, you might find this article disturbing, if you decide to proceed, do so at your own peril.
Defects are waste
Let’s be clear from the start: defects are waste. Waste of time in designing defective products, waste of time in coding defective routines, waste of time in detecting them, waste of time in fixing them, waste of time in re-checking them. Even writing this sentence took a good while, now think how much time it takes you to produce, detect, fix, recheck defects.
Our industry has developed a defect coping mechanism that we call defect management. It is based on a workflow of detecting => fixing => retesting. Throughout the years it has become best practice (sic) to have defect management tools and to log and track defects. Defect management approaches are generally cumbersome, slow, costly and tend to annoy people no matter whether you are a tester that gets his defect rejected, you are a developer that gets a by design feature flagged as defect, or a product manager that needs to spend time prioritising, charting and trending waste.
Another dangerous characteristic of defects is that they can be easily counted and you will always find a pointy haired manager that decides he is going to shed light on the health of his product and on the efficiency of his team by counting and drawing colorful waste charts.
But, if we agree that defects are waste, why are we logging and tracking waste, creating waste charts seems even more ridiculous, wouldn’t it be easier to try to prevent them?
Oh, if only we could write the right thing first and reduce the number of defects we produce! I say we can, be patient and read on.
Software development teams have found many ways of being creative playing with defects, see some examples below.
Example 1:Reward waste
Some years back I was working on a business critical project in one of 5 scrum teams . Let me clarify first, that our scrum implementation was at best poor, we didn’t release every sprint and our definition of done was questionable.
Close to an important release, we found ourselves in a situation where we needed to fix a lot of defects before going into production. We had 2 weeks and our teams had collectively around 100 defects to go through. Our CTO was very supportive of the defect killing initiative and he was eager to deliver with zero defects. He put in place a plan that included free food all day and night and some pampering for the developers that needed to focus 100% on defect resolution. Then he decided to give a prize to the team that would fix the highest amount of defects.
I remember feeling frightened of the possible future consequences of this reward. I spoke to the CTO and told him that I would have liked more a prize for the team that produced the lowest amount of defects rather than the one that fixed the most. Our CTO was a smart guy and understood the value proposition of my objection, he changed his approach and spoke to the teams on how not introducing defects in the first place is much more efficient than fixing them after they have been coded. Soon after the release, we started applying an approach that focussed on preventing defects rather than fixating on detection. We never had the problem of fixing 100 bugs in 2 weeks again.
Example 2:Defects metrics
In my previous waterfall life, I remember when management introduced a performance metric directly linked to defects. Testers were to be judged on the Defect Detection Index calculated as (Number of Defects detected during testing / Total number of Defects detected including production)*100. An index lower than 90 would mean nobody in the test team would get a bonus. Developers were individually judged on the number of defects found in their code by the testers and business analysts were individually judged on the number of defects found by the testers in their requirements.
Welcome to the battlefield!
The bug prioritisation meetings were battles where development managers argued any bug was a missed requirement, product managers argued that every bug was a coding error or a tester misunderstanding and the test lead (me) was simply shouted at and criticised for allowing his testers to go beyond the requirements and make use of their intellectual functions outside a scripted validation routine.
Going to that meeting was a nightmare, people completely forgot about our customers and simply wanted to get their metrics right. The amount of time we wasted arguing and defending our bonuses was astonishing. Our customers were normally unhappy because instead of focusing on value delivery we focussed on playing with defects, what a bunch of losers we were!
Our customers were very unhappy.
Example 3:Defects as non-conformance to requirements
In the same environment as Example 2, testers, in order to keep their Defect Detection Index high used to raise large amounts of minor or non-significant “defects” that were in reality non-conformance to requirements. Funnily enough such non-conformances were generally improvements.
Testers didn’t care if they were requirements, code defects or even improvements, to them they were money, so they opened them. Improvements were filed as defects as they were in non-conformance to requirements. In most of the cases, these were considered to be low severity and hence low priority defects to make the testers happy and had to be filed, reviewed, prioritised and used in trends, metrics and other useless calculations.
This activity could easily take 30% of the tester time. Such defects would not only take testers’s time, but would also affect developers, product managers, business analysts and eventually clutter the defect management tool.
Waste that creates waste, exponentially, how wonderful.
Example 4:Defect charts, trends and other utter nonsense
Every week I had to prepare defect charts for management. These were extracted from our monstrous defect management tool and presented in brightly coloured useless charts. My manager got so excited at the prospect of producing useless information that she started a pet project to create charts that were more colourful than the ones I presented. She used 2 developers for 6 weeks to create this thing that was meant to wow the senior executives.
In the process of defining the requirements for wowing the big guys, she introduced a few new even more useless charts and consolidated it into an aggregating dashboard. She called it the product quality health dashboard, I secretly called it the dump.
Nobody gave a damn about the dashboard, nobody used the numbers for any reason, nobody cared that they could configure it, but my boss was extremely proud of it. A legend says that she got a big raise because of it. If you play with rubbish, then you will start measuring rubbish and eventually you will end up doing data analysis and showing a consolidated view of the rubbish you store in your code.
How can we avoid this?
1. Focus on defect prevention
Many development teams focus on delivering features fast with little consideration for defect prevention. The theory is that testers (whose time is sometimes less expensive than developers) will find the defects that will be fixed later. This approach represents a false economy; rework disrupts developers activities and harms the flow of value being delivered. There are many approaches available to development teams to reduce the amount of rework needed.
Do you want to prevent defects? You can try any combination of the below:
With BDD/ATDD/Specification By Example or other test first approach, delivery teams test product owners assumptions through conversations and are more likely to produce the right feature the first time.
The ability to have fast feedback loops also allows for early removal of defects, automated unit and integration tests can help developers quickly identify potential issues and remove them before they get embedded into a feature.
Tight collaboration between business and delivery teams helps teams be aligned with their real business goal and reduce the amount of unnecessary features. This means less code and as a consequence less defects. Because, your best piece of code is the one you won’t have to write.
Reducing complexity is very powerful in preventing defects, if we are able to break down a complex problem in many simple problems we are likely to reduce the amount of defects we introduce. Simple problems have simple solutions and simple solutions have less defects than complex ones.
Good coding standards like for example limiting the length of a method to a low number of lines, setting limits on cyclomatic complexity, applying good naming conventions to help readability also have a positive impact on the number of defects produced
Code reviews and pair programming greatly help reduce defects
Refactoring at all times also reduces defects in the long run
Moral of the story: If you don’t write defects, you will not have to fix them.
2. Fix defects immediately and burn defect management tools
If like me years back, you are getting tired of filing, categorising, discussing, reporting, ordering defects I have a very quick solution. Fix the defects as soon as you find them.
It is normal for a developer to fix a defect he finds in the code he is writing as soon as he finds it without having to log it, but as soon as the defect is found by a different individual (a tester for example) then apparently we need to start a strict logging process. Why? No idea really. People sometimes say: “if you don’t do root cause analysis you don’t know what you are doing, hence you need to file the defects”, but in reality nobody stops you from doing root cause analysis when you find the defect if you really want.
What I am suggesting is that whoever finds a bug walks to a developer responsible for the code and has a conversation. The consequence of that conversation (that in some cases can involve also a product owner) should be let’s fix it now or let’s forget about it forever.
Fixing it now, means normally that the developer is fresh on the specific code that needs to be fixed, surely fresher than in 4 weeks, when he won’t even remember he ever wrote that code. Fixing it now means that the issue is gone and we don’t have to worry about it any longer, our customer will be thankful.
Forgetting about it forever means that it is not an issue worth fixing, probably it doesn’t threaten the value of the project and the customer won’t care if we don’t fix it. Forgetting about it forever also means that we won’t carry a stinky dead fish in a defect management tool. We won’t have to waste time re-discussing the same dead fish forever in the future and our customers are happy we are not wasting time but working on new features. If you decide to fix it, I’d also recommend you write an automatic test for it, this will make sure that if the issue happens again you’ll know straight away.
I have encountered huge scepticism when suggesting to burn defect management tools and fix just in time. Only very few seem to think this is possible. As a matter of fact all my teams were able to do this for the last 6 years and nobody ever said, “I miss Jira and the beautiful bug charts”.
Obviously this approach is better suited for co located development teams, I haven’t tried it yet with a geographically distributed team, I suggest you give it a try and let me know how it goes.
Playing with defects waste index:
Epidemic: 90% – The only places that don’t file and manage defects I have ever encountered are the places where I have worked and have changed the process. In the last couple of years, I have heard of two other places where they do something similar but that’s just about it. The world seems to have a great time in wasting money filing, categorising, reporting, trending waste.
Damaging: 100% – Using defects for people appraisal is one of the worst practices I have ever experienced in my long career, the damage can be immense. The customer becomes irrelevant and people focus on gaming the system to their benefit. Logging and managing defects is extremely wasteful as well, it requires time, energy and can among other things, endanger relationships between testers and developers. Trending and deducting release dates from defect density is plain idiotic, when with a little attention to defect prevention defects would be so rare that trends would not exist.
Resistant: 90% – I had to leave one company because I dared doubt the defect management gospel and like an heretic I was virtually burned at the stake. In the second company I tried to remove defect management tools I was successful after 2 years of trying, quite resistant. The third one is the one where people were happy to experiment and as soon as they saw how much waste we were removing it quickly became the new rule. I have had numerous discussions with people on the subject and the general position is that defect management must be done through a tool and following a rigid process.
A classic problem for testers in agile contexts is the fact that they feel they are not listened to by developers. Testers, often rightly, warn developers from doing stuff because the consequences could be very bad, but developers in many cases don’t listen to them.
This is very upsetting, testers find themselves lonely within an agile team because of this. They get frustrated and if they keep on asking and screaming their needs they risk being alienated by their team members becoming completely ineffective while growing dissatisfied with their job.
But, but, they are right in telling the developers what to do! WTF?
When working as a tester in an agile team you’ve got to develop a skill that before you really didn’t need that much.
It is called influencing. Before when you worked in your separate QA department/test team you didn’t need it because there was a test manager that was fighting the battles for you and deciding the test strategy to be applied.
Things have changed, you’ve got to become good at influencing.
This is what Gus did when he moved to an agile team as a tester many years ago.
First I tried barking orders, screamed and shouted, but it din’t work at all, so I decided to adopt a different approach.
I started to really listen to what developers said instead of listening for finding gaps in their thought
I listened and listened and listened a bit more
Third I started asking questions showing real interest in what they were doing. Being mindful of their fears and feelings. I made sure they knew I was there to help and that we were all in the same boat
I started praising them when they did something good, for example thanking them for adding testability to the application. Things like “without Roberto’s design I would have spent weeks doing what I can do now in 10 minutes, thank you so much Roberto, you made my life better”
I started coaching them on how to test by testing with them. When they saw what testing really involved, they understood its importance and challenges and started asking interesting questions about it.
Who will developers listen to?
Now compare the developers’ reaction when faced with a suggestion raised by Gus to the one that Jack gets. Jack is a tester that uses the classic approach of “what the hell are you talking about? This is going to explode in production!”
Who do you think will be able to influence developers actions when something important for testing needs to be done?
Me? Often, I always got the developers at least to listen to me and a lot of the times we did it my way unless somebody in the team had a better idea.
So, do you want to be Jack and keep on moaning about developers that don’t understand anything about testing?
If I were you I’d take Gus’s approach and build your influence within your team, start now, start listening.
In my last stint as a tester from October 2012 to Jan 2014, I helped my organisation at that time moving from delivering once every month, to delivering multiple times a day.
Let me first clarify that we didn’t move to multiple deliveries per day just for the fun of it, but because we needed it.
Your organisation might not yet know it needs this level of agility but more than likely it will at some stage in the future.
How did this transform the role of the testers within the organisation?
When I joined I found scrum teams that delivered either once a month or once every 2 months. The teams had 3 different defects management databases full with old and new defects. Testers were doing the following activities:
exploratory testing (~50-70&)
The batches were big, the exploratory sessions were long and found a lot of defects. The automation was not effective, as it was slow and unpredictable, its value was negative.
When I left
When I left, we were using kanban, delivering multiple times a day, defects were more or less a myth of the past, no defect management tool existed. Testers were doing the following activities:
Three amigos BDD sessions with customers and developers
Exploratory testing (1~5%) – never longer than 10 minutes per card, more often than not reporting no defects
Pairing with developers
Coaching developers on testing
Writing automation (0%)
Talking to the customer and the team
Improving the system
Designing the product with the team and the customer
Helping define what to monitor in production
Any other valuable activity the team needed them to do
As you can see the activities that before occupied 100% of testers time, now occupy from 1 to 5% of testers time.
Were testers busy before? Yes, absolutely
Were testers busy after? Yes, absolutely
Were testers complaining because they weren’t doing automation or enough exploratory testing? No, believe me. Most testers I worked with saw the new activities in the role as a learning activity and an opportunity to broaden their skills and become more valuable to any company.
If a tester didn’t want to adapt to the new reality and embrace the change and new ways of doing things, he would have been busy for 10 minutes a day (~2%) and he would have not been useful to the team.
Did we get there with the touch of a magic wand? No, the end stage was the result of many experiments. It was, back then, a good recipe for that context at that time (it is continuously changing)
So, tester, what’s your strategy for working in a company that releases multiple times a day?
There is one thing that bugs me about mind maps as a testing tool, it really bugs me a lot.
Don’t get me wrong, I find mind maps a great collaboration tool to get ideas out of people’s heads and put them on paper. They are great for designing a test plan, a few testers can sit down, brainstorm and create a map that can be efficiently used as a complete plan.
So what bothers me so much?
This: “Why don’t we do the same exercise together with the developers before the code is written?”
Are we so obsessed with finding bugs that we cannot share our thought process and help PREVENT (yes prevent) the defects instead of catching developers out and wasting customer’s money and time?
I really hope that somebody out there will give me a valid reason why we shouldn’t shift left the mind map creation and prevent the defects instead of detecting them.
Otherwise we should start doing it now, no excuses.