Scrum is fragile, not Agile
dennisweyland.net
dennisweyland.net
I used to think that was a contradiction. Now I think it's almost inevitable. I wrote more about it elsewhere [1], but the basic deal is that most companies have other priorities than being effective, so the processes that dominate will reflect those priorities.
On average, the first priority of managers and execs is maintaining the power structures that make them a big deal. But true Agile processes are about empowering teams to self-organize around serving users. Ergo, Sturgeon's Law [2] applies here too: 90% of company software process is crap.
I think reason why Scrum in specific won out is that it combined a process framework unthreatening to executives with the multilevel marketing scheme of Scrum certification. Executives got to keep doing what they were doing, the workers got a "didn't stop breathing" certificate, and coaches got paid. Win, win, win!
[1] http://williampietri.com/writing/2011/agiles-second-chasm-an...
The most common failure I see is when project leadership agrees to a fixed scope and timeline then tries to execute in an agile way. Agile is the contract. But it's also a tough sell. It requires a high degree of trust to say "we'll pay you $XM in exchange for X sprints of whatever we prioritize with no fixed end state" but that's what you need to be agile. If you layer sprints on top of a fixed timeline, you're just adding needless complexity since your end state is predetermined and how you get there is irrelevant.
I agree that fixed-scope contacts cause a lot of problems for Agile approaches. However, they also cause a lot of problems for non-Agile approaches. If I have to deal with supposedly fixed-scope situation, I'm going with and Agile approach.
There are two basic cases. One is that scope is truly fixed (which is rare). In that case, having a new releasable version every week with highest-priority features first is excellent risk management. When the date comes, you'll have something to ship. You also get to continually validate and improve internal processes, so you're more likely to be using the time available effectively.
The other is that scope was fixed in the contract but is in practice variable. So every week you deliver something to the customer. Every week you build trust. And every week you encourage them to do tests, deploy early, anything so they start getting value. At that point they get user and business feedback, and come to you with changes. That's the point where you start shifting from fixed-bid to time-and-materials. Maybe you keep to the same budget, but now they're getting more for their money. Maybe they look at the backlog and say, "Wow, there's still lots more value to deliver, let's keep going."
In many places, especially smaller ones, project managers are also developers (or lead developers etc), and they often drink the Scrum kool-aid.
Developers may advocate small-a agile, but I have never heard of an actual developer pushing for Scrum, SAFe, or any of the name-brand "agile" methodologies, that come with expensive consultants and certifications and "coaches" and conferences and a whole ecosystem around them. These are things for project managers for the benefit of project managers who see project management the main deliverable l in and of itself. The same with JIRA actually, it's not a software development tool, it's a project management tool, sold to the same people who previously bought MS Project. These things are all overheads in the software development process, not enablers, as another poster says, the goal of most organisations is not to be effective at their stated mission, but to maintain their internal power structures.
Developing new aircraft, building a new ship, building a custom-designed bridge (most of them are) are processes that often run out of time and / or budget.
If you want predictability, you want repeatability. But in software all reliably repeatable parts become automated away.
But I think good projects release early and often precisely so that they can learn as they go. At which point predictability goes out the window.
I have seen more success in cases where teams estimate for 90% percent confidence instead of 50% confidence. By that I mean "it should almost never take longer than that" vs. "it will probably take that long." Unfortunately, to do that, you need enlightened business management that appreciates the difference between an estimate and a promise, as opposed to business management that pays lip service to the distinction.
This sounds like a much better way to estimate. Do you have any links to content discussing this?
I haven't seen a good way to turn the concept of "it should almost never take longer than that" into a concrete process. I've always seen it as a gut check, often implemented as "take your initial estimate and double/triple it." What I actually see is that a lot of teams implement that but walk back the doubling when the business/PM delegation persistently asks for more in less time. Only in really engineer driven cultures have I seen engineering teams successfully push back.
The alternative is some sort of sub-iteration hybrid where you attempt to fix deadlines for much smaller units of work and constantly revise.
This is OK for some industries. But 2/3rds of a car doesn't quite sell...
That in turn means technical debt; which puts future deadlines and feature sets at risk.
Since in SCRUM as a team you try to not compromise on quality the only way out of this is to push back on deadlines or feature set (or both).
So instead of blindly executing orders; the development teams can push back on deadlines.
Most of the time these are actually negotiable. Even the "hard" deadlines....
It doesnt mean the dev teams always win however they are better equipped to inform "management" about the consequence of the deadline.
https://www.amazon.com/Doesnt-Have-Be-Crazy-Work/dp/00628747...
https://en.wikipedia.org/wiki/Project_management_triangle
If time is fixed, then scope and/or cost must change.
Even if so, that would be in itself a failure of Scrum. A process that works only sporadically and only with exceptionally well organised groups ain't really all that much.
> It requires a high degree of trust to say "we'll pay you $XM in exchange for X sprints of whatever we prioritize with no fixed end state" but that's what you need to be agile.
...no? It requires a well-defined end-state, and that includes not defining the irrelevant. It's not really "give me $XM and we'll deliver anything from accounting software to a really nice puppet."
Unfortunately Scrum is too often explained and implemented that first way, leading to the anti-Agile feedback we see on HN with some regularity. The second way is much more compatible with complementary tools such as Lean Startup and Kanban and I wonder if this (to me very welcome) non-exclusivity explains why it is less talked about.
[1] https://blog.agendashift.com/2018/07/04/righttoleft-works-fo...
From what I can tell, it has all the meaninglessness of an empty buzzword used primarily as a method of trying to engage people and give an opening on selling their services by stating "oh if you don't understand the difference sit down with us and let us show you how different and better it is."
It looks like if scrum doesn't work they call it "falling left to right scrum", and if it works it's their "brilliant right to left scrum".
Even by their own definition they both have a backlog that gets prioritized and selected each 2 weeks into a sprint backlog, and executed during the sprint.
The rest is just hand waving. "One is goal focused vs the other is backlog focused. Oh yeah but we do put our goals in the backlog." So they're both backlog focused then? The only thing they're really sayings is "When prioritizing tasks, make sure they accrue to something and aren't just random work." Which is both obvious, and of course too simple to write and sell a book about, so instead this whole other terminology is made for it.
Thanks for sharing the link though.
If I ever find a developer who can sell, selling is not a developer's USP in general, there are exceptions and I would hire those exceptions instantly!
aka, WaterAgileFall.
You aren't wrong, either. I found Scrum to be a series of waterfalls that weren't well thought out. "Iterative Design" essentially meant, "We aren't sure what the button should do exactly, but we know we need it there and it kinda has to do this and we'll figure the rest out for the next iteration."
That caused so many problems with tech debt.
Stories began to take longer due to the increasingly large and discombobulated code base, but the expectation was that we continue to deliver the same number of story points each sprint.
Then we got blasted for losing pace. "Why are you under-performing? We are just adding little features. You've already implemented this button on another form before? Why does it take longer the second time? You already know how to do it!"
It became absurd and no one could hear or understand what was happening was easily predictable and in fact -- was predicted by several members of the team many months prior.
When the project management was informed of these predictions now having come true, the development team was accused of intentionally causing the delays.
The most upsetting experience in my career thus far was being belittled by an incompetent project manager.
"How could it possibly take that long? How do you not know how long it will take?"
The happiest times of my career have been when we're lacking a PM -- coincidentally the most successful!
Well, if they really said that, you can just dismiss the comment as incompetent. A witty retort may be in order if a non-technical manager is in earshot, e.g.:
"How long would it take for you to earn a green belt in karate?"
"Hmm, what does a green belt entail?"
"Exactly, you don't even know what you don't know, yet."
"We don't know because we've never done this before, and we can only compare to the most similar work we've done, and any existing data."
Before I could say more, they I interrupted: "You have to know, it's your job, or you don't know what your doing," or something along those lines. It was the most heated I've felt. I know what I'd reply with today, but I was younger and more naive then.
> "or you don't know what your doing,"
That though was unprofessional, and should be addressed as such.
I don't think this will score you a lot of points.
Also, someone asking how could something take so long is not a belittling question. I don't see any problem with either question.
There are so many places where this can sneak in and become an issue, too:
- The development team can groupthink, underestimate and/or fail to account for technical debt payback in planning poker. They may estimate the cost for the new feature, but not include any time for refactoring to keep the codebase healthy
- The Scrum Master may not shield the development team from pressure to deliver coming from the product owner, which may cause the team to cut the above corners
- Codebase health may be split out from vertical stories into its own tasks that get deprioritised by the product owner
- Even if the development team identifies the need for codebase health, this may not be communicated back to the product owner in a way that makes business sense ("but we can still get this feature out the door quickly and fix the mess later, right?" - yes, once or twice, but the downside is a slowing of velocity)
- Even if the team started well with ground rules the Scrum Master is supposed to enforce surrounding technical quality, the Scrum Master may not be effective in enforcing the process - either because they lack authority or because they lack the ability or knowledge to do so
Among several others.
In your situation, the big red flag is that the team predicted the issue many months prior. The process has failed to extract that information from the team members and deliver a consensus between them and the product owner.
Though ultimately, as with any project, if the product owner has more organisational authority than the development team and chooses to exert this authority, no process will be able to save you.
The tricky bit is having a solution and implementing it.
Having something well thought out is hard work. Is hard work rewarded? Is initiative to even try rewarded? I'd say the answer is no.
The answer becomes yes in small teams that are given time, freedom and resources, or in times of crisis (same thing as small team really, minus the time bit), as far as I can tell. Otherwise, most people's natural tendency (including me) is to phone it in. We'd be wise to let people in software development and many other professions that are not manual labor, to switch careers after 30-35 if they haven't done anything worthwhile by then.
Working with people who have to phone it in for the next 30 years and know it is modern day hell for anyone with an iota of ability.
It's permanent compromise and insincerity, or being an outcast that's temporarily tolerated for doing 5-10x the work of the person next to you.
The problem is always : features and customer request over bug fixing or design fixes.
The system, project type does not matter. That's why some PO recommend 10 percent allocation to devops / ops, 10 percent to improving code base . And "no bug survives the Sprint" philosophy. And what is left,is you next feature or customer request.
In other words, Scrum methodology in itself is not mature as a methodology and depends on individual fiat, just like any project that doesn't have any process at all.
The only tangible benefit to Scrum might be paying lip service to development departments while firmly retaining the status quo.
The rest of your post is just warping your initial conjecture to arrive at conclusions you'd already decided upon. I neither agree with those conclusions nor the chain of twisted logic you used to arrive that.
I'm not going to stand on a soapbox and sing for the glory of scrums. People - like yourself it seems - can get very tribal when talking about such things. Which is as weird to read as it is pointless for you to argue. In my experience leading different teams using different methodologies, the thing that really makes the big noticeable impact is people and not methodologies. If you work in a blame culture or have colleagues in your team who don't follow process - then whatever process you put in place will be undermined at every opportunity regardless of it's methodology. However if you work in an environment where people respect one another and want to collaborate in getting work done, then you pick a methodology that works best for the day to day work (eg project work or support operations) and for the way people like their work organised. All the rest of the arguments are superfluous.
Yep. People do not realize that the whole company has to adopt agile/scrum for it to work. It's a shift that many companies can not or will not make. They are tied to fixed deadline, fixed scope for various reasons.
I’ve noticed comments for and against scrum are often so focused on the technical aspects of the methodology that they overlook the human aspect. Which matters more in my opinion.
I think it works well enough, for a few years. I don't think it's sustainable - the blinkers of "sprints" encourage growth of tech debt because nobody has an eye on the future and Product won't prioritize refactorings, and sprints are too short for devs to sneak refactoring into the schedule.
No, it isn't. Iterated waterfall is at least as old as the first paper discussing waterfall, but while scrum mandates interations, it doesn't mandate much about how work is done in the iterations, and specifically does not mandate the process steps associated with waterfall; further, it emphatically rejects the role separations and handoffs associated with waterfall during the iterations.
> munching through backlog items fed to them by product managers.
That's...not actually Scrum, as it implies that either the role of Product Owner is taken by a PM outside of the Scrum Team or that that the Scrum Team is not self-organizing, either of which is a significant (even if common) deviation from Scrum.
> the blinkers of "sprints" encourage growth of tech debt because nobody has an eye on the future
Tech debt should manifest in reduced velocity which should be noticed, taken as a signal of a process defect, and addressed in the Scrub Team’s various inspection and process adjustment points.
OTOH, if the Scrum Team is properly cross-functional and self-organizing instead of having a non-team-member imposed as Product Owner, then including appropriate restarting as components of completion of relevant backlog items shouldn't be a problem.
About the same percentage as that of “Agile” software development shops that put people and interactions above processes and tools.
OTOH, at any place that is considering implementing either, there are decision makers who can influence (or in the case of Scrum more than Agile, authoritatively direct) whether or not that's the case, so for them, at least, it's worth distinguishing between problems with Scrum as prescribed and problems which often occur because decision-makers decided to ignore key parts of Scrum-as-prescribed.
This process adjustment usually means "sorry about your vacation" and/or "you aren't doing enough overtime".
(Of course, if the wrong actor is monitoring velocity and controlling process, then the team has an incentive to mask the effect of tech debt but continuously adjusting estimates to maintain the illusion of constant velocity, a d avoid the first bad response, or, if that opportunity is missed, to avoid the subsequent external interventions.
"Fortunately," hardly anyone in Silicon Valley plans to be working on the same codebase in a few years. There's a good chance the problem space won't be relevant anymore by then, and on the off chance it's still funded, the new team will rewrite it whether or not it's good.
>sprints are too short for devs to sneak refactoring into the schedule.
As I've gotten more senior, I've just gotten more brazen about doing this less sneakily.
A lot of 6 month old code gets thrown away either because its been rewritten or because it didn't achieve its stated objective. Meanwhile a bunch of companies are relying on systems that were first created in the 90s because that software achieves its objectives and the cost justification for a rewrite isn't there.
The "spiral" model is closer to a true iterated waterfall. I've seen it used successfully in more mature companies.
The effective strategy is to build an oblique, fatally limited prototype, which can not possibly be mistaken for a shippable product. The bad prototype can only be used to test the hardest parts of an idea and must have major holes in it with no way of filling them. It should also have a largely fixed timeline to ensure it is put to bed before it gets "hamstered" into the shipping product.
Of course the business didn't know all the scenarios at the beginning of the feature development and didn't care because: iterative development means we'll figure it out later.
A lot of devs trap themselves by insisting they can write the tests after. Once managers know the code exists they want to use it, or move on to the next thing people are breathing down their necks for. And code written without tests is difficult to test so becomes a self fulfilling prophecy.
In fact it’s such a reliable mechanism for self sabotage that I look at carefully at people who bring this on themselves and try to figure out if it’s naïveté, learned helplessness, or malice.
We need to unify around these principles, possibly revoke memberships for unprofessional behavior. One dev getting himself fired changes nothing..
We tend towards 4 or 5 sprints in which nothing is completed, only to suddenly have a flood of completed things the next.
It's perfectly possible to deliver small units of work weekly. One still gets frameworks while doing that. It's just that the work of making the framework happens over time, not up front. Which is good, because good designs are based on good information, and the beginning of the project is when you have the least information.
The terminology is so incredibly developer hostile.
Could you elaborate?
This creates a confusing code-base with lots of IF statements, copy/pasted code, data modifications from several areas of the system, etc.
You'll fix a problem with data getting into the database incorrectly and realize a couple sprints later, fixing that bug caused other bugs elsewhere.
Automated testing doesn't catch it because automated testing only handled certain scenarios.
It's a mess.
Well there's your problem. How is that inherent to scrum?
There is something odd with this, if you did the estimation. Shouldn't the story points/sizing/estimates preclude this?
Doesn't the team pick the items for the sprint that they think they can complete?
I often wonder if they picked these words because they knew nothing at all about sports, of if there is a coded warning in them...
It’s the same reason why demonstrably value-additive, cost-effective prediction markets are rejected by managers despite evidence in their favor: it removes their ability to create ambiguous Dutch books out of different project deliverables and resort to exploiting subjective psychological biases of superiors to win money & status increases. Anything that holds them accountable to the actual measured result status of a deliverable can’t be tolerated.
Scrum is just the by-product of corporate status-seeking evolutionary war games against in-house scientific thinking.
Does anyone know if there is an anti-movement to that over-communication trend? Something along the lines: "The only way to go fast is to go well", * Plan * Honor the flow * Adjust
I feel we kind of lost something where creating code/designs using strong feedback loops was a "first class citizen with royal honors" of the software development process. Scrum kind of makes it a sideline issue, a peon of middle management
My dad started developing in the late 1960s. He did a lot of in-house software, mainly manufacturing and insurance systems. His approach was to go out and see something in use, talk to the people using it, go back to his office, write something, release it, and go talk to the people again. He'd do that loop in as little as a couple of days.
Today we'd call that obviously agile. But both back then and now, higher-ups wanted more "control", by which they meant feelings and appearance of control. Which caused longer feedback loops. Which made things more out of control, which required more process and even longer feedback loops.
I think we have the same struggle today, and I have no idea how to get out of it.
IMHO, the idea is that over time the organization will adopt new Agile practices and not depend on Scrum anymore. In practice, however, many organizations think they are Agile after they introduced Scrum, which is a simple fallacy. After all, becoming Agile is about a culture change and not about the introduction of a new process framework.
To drive the point home I could easily transform "[o]n average, the first priority of managers and execs is maintaining the power structures that make them a big deal" into a derogatory comment about engineers, fad chasing, and resume padding by a simple word substitution. It would be neither accurate nor fair to many (perhaps most) engineers: just as your comment isn't really accurate or fair to managers or organizations.
That's not constructive in the sense of providing ways to solve the problem. But that wasn't my goal. My goal was to support and confirm the article's thesis. I have also written plenty of constructive things on software development, but I think it's absurd to suggest that every HN comment must contain that kind of constructive feedback. (And if that were a reasonable standard, you aren't living up to it.)
But that doesn't explain the success of Scrum, because Scrum does away with the power structure of managers, and should empower the teams. "We're doing Scrum now" is the big stick I see scrum masters use to beat back Business managers to keep them from interfering with and micromanaging the team.
At least, that's what they should be doing when management interferes too much. And then address this higher up with whoever decided that we're doing scrum now, and explain to them what that means, if necessary.
There's no excuse to tolerate micromanaging managers in a Scrum process. They don't belong there. Kick them out.
* Daily standup, nobody wants to be at. We have multiple teams arrive, with roughly 20 people in a small room. Some people stand, some people sit. Sometimes the front-end team goes, sometimes the back-end team goes. Its limited to 15 minutes, so nobody says much of importance, and just parrots what is already on the Jira board.
* Scrum master driving process above scrum. The scrum master's role is to make sure scrum rules are adhered to. But at this company, any deviation from scrum itself poses a threat to this person's job security, so it doesn't happen. More high-level processes are not optimized because the scrum master defines everything.
* Technical debt. I see this time and time again. User Stories are supposed to be forecasts, not commitments. But the business doesn't like stories carried over, so they become commitments. At the end of each sprint, everyone rushes to get their stuff done, and hacks are implemented to meet an arbitrary deadline. Many times I want to begin my work by refactoring something to what it needs to be first, then do the actual user story. But its risky because the refactoring might take more than the allocated story points, and you get dinged. So I do the story first, and if there is time do the refactoring but it almost never happens.
* Poor product owners. During maintenance phase, we could work on cleaning up technical debt, but this value is not appreciated by business so they keep us busy with bikeshedding. One gripe I have about scrum, is there is nobody representing engineering, as the product owner represents the business.
These things combined have dragged down the happiness of the people I work with, but we all feel imprisoned by it. I have a stack of scrum books here I plan on reading, I figure this process isn't going away and I need to up my game with how to play it - but I wish I could use something else, perhaps kanban.
The project mgmt response was, "Attendance at standups is mandatory."
Regardless, I didn't go to anymore standups and when I got flack for that, I stopped going to the office all together. When I got flack for that, I stopped working all together.
Then I got fired. That solved all my scrum problems.
That's it. (Ok, once a year they do reviews.)
[0] If they don't bug a dev to do it.
I got fired from any last job after being one of the only engineers willing to say something to management about this. I figured I had a popular opinion - many other engineers often DM'ed me on slack encouraging me to continue speak out, including my own manager, so I somehow figured I should be safe speaking out, politely, as a respected majority representative.
In fact I was dead wrong. By saying that scrum was a problem, I ended up making product managers / CTO who loved scrum and used it as an hour long opportunity to lecture the team every day feel threatened.
Getting fired was one of the best things that ever happened to my career, because in retrospect it was a dead end company that was run by fear. Management was afraid of ideas or challenge to power to the point where innovative ideas and feedback were never well received and a culture of fear and not stepping out of line arose, even though it was sugar coated with fake company values of "openness" that nobody really believed in but that managers loved as a way to elevate themselves.
I suspect many companies are a my least a little bit like this. Unfortunately it is human nature and there are so many stories in history, usually of narcissistic dictators, that mirror this.
This was a very extreme example but I'm sure many other companies struggle with this when it comes to challenging scrum. Managers tend to love scrum because it's a chance, daily, to "manage" and to get status updates so that they feel comforted. It makes it so that trust is no longer necessary, and bad managers are often bad at trust.
I guess another way to view scrum is to realize that it is often a reflection on management's subtle fears and insecurities being projected onto the processes of the company. It is an instance where employees unfortunately must manage up.
It is central to the idea of the daily standup is that it is one team.
> Its limited to 15 minutes, so nobody says much of importance, and just parrots what is already on the Jira board.
If you have another mechanism for sharing what each person has done and is doing, then the standup should just be for sharing information about and resolving barriers.
> Stories are supposed to be forecasts, not commitments. But the business doesn't like stories carried over, so they become commitments.
If you organizationally can't get away from commitments, then you need to cut back the amount you plan to do so that there is some slack. As long as you are also doing backlog grooming and have items groomed beyond the current sprint commitments, you can take additional items opportunistically, if there is excess capacity after doing the committed items properly.
The powers that be at my work decided that “sprint predictability” is the most important metric here. This means that if we pull in stuff near the end of the sprint but don’t finish, predictability goes down. Thus, we are encouraged to not take on extra work if we don’t expect to finish. What this means in reality is that we start working on it without pulling it into the sprint and then get a head start for next sprint.
What’s your take on that?
I also think speculative future make-believe planning junk should be kept out of the real getting-shit-done tool. Put it in there when it actually matters. There should be no "on ice" or "phase 3" (when in "phase 1") crap in the tool the devs use. It may or may not make sense to have them involved in planning stuff that early, but keep the output of that planning away from the real-work trackers until it's closer to time to do it.
I think that's part of why Jira (and similar, heavy, report-focused tools) is so god-awful. It not only tries to have features for teams to get stuff done and for five other roles to dig around in tasks and reports for whatever reason, it enables and encourages that, and I'm fairly sure, at this point, that it's fundamentally a bad idea.
This probably increases the accuracy of the work trackers but increases the B.S. level of management reporting in organizations where there are fundamental problems between management and the working level.(Which is the only reason anyone would want to separate these two things.)
While this probably seems like an improvement from the working level, it's probably bad for the organization. The solution is not dis-integrating communication tools so that management's view is not connected to the actual work tracking, but dealing with the fundamental trust issues. Which is hard, of course, but things that are important often are.
[EDIT] more to the point, I think if it worked we wouldn't still constantly see management surprised when things aren't delivered "on time", and yet, that still happens all the time. IMO the team needs someone "on their side" to report reality upwards, diplomatically, not their own work tools reporting up directly, or they'll lie to their tools, which leads to more surprises, not fewer.
When you do finish a bit early I've not normally found it a problem to find some small bits of tech debt, research, or admin work to fit in before the end of the sprint.
Starting work on stuff from the next sprint is not ideal because it'll throw your estimates off for the next sprint, plus you don't actually know for sure what is going into the next sprint.
Then, two weeks later, being asked why I did not complete what I committed to.
"Development Teams do not commit to completing the work planned during a Sprint Planning Meeting. The Development Team creates a forecast of work it believes will be done, but that forecast will change as more becomes known throughout the Sprint."
https://scrumguides.org/revisions.html, "Changes between 2010 and 2011 Scrum Guides"
I say this as someone just moving into project planning and management. From that perspective you start to see that some level of estimation s critical. My preference is to ask devs for just the tightest bounds to timeframe they're actually comfortable with, then estimate based on that and pass up the chain. I consider it to be on my shoulders if things take longer than I communicated upwards.
Anecdotally this has worked out remarkably well throughout my career - from single dev to cto - even stuff that I _new_ was going to take for example 2 days, if done properly ended up in like 5.
Didn’t matter if I was doing the estimate or someone else. At some point I just gave up and started doing the “my gut says 3 hours, so it must be 7 hours of work then, and send that up the chain. That cheeky “+1” had saved my ass more times than I can count.
But without the feedback of actually having done the work, any estimate is bound to be wrong.
I made exactly this suggestion and was laughed out of the room.
Granted this was for large and small estimates, but I still don't think it's the worst idea.
Generally they don’t though.
That is definitively not be the feeling you should have. In "proper" scrum the Team is supposed to represent engineering. If you, for some reason, don't feel like you can represent these concerns then that is a huge issue. The PO shouldn't be your boss either. At least it isn't so in the company I work for. It sounds like the PO is the boss of that team, if he "keeps you busy". I am super grateful that our Scrum Master knew what he was doing and suggested creating a separate line on the Org chart for the POs so that they are explicitly not in charge of development. The team is in charge and POs are more advisory.
Sounds to me like there are some deeper problems in the company then just the scrum. There are so many red flags in your comment. I don't know if those books are going to help you "play" scrum better, because what you guys are doing does not sound like scrum at all.
Kanbans probably also not a solution. I really like Kanban, but it isn't going to remove the bad elements that seem to control that company.
In scrum the dev teams maintains a technical backlog and fills the sprintbacklog as THEY see fit- from the product backlog AND the technical backlog.
the PO has no say in this! he can complain that he thinks not enough features from his product backlog are planned, but should trust that the team knows what they are doing.
>These things combined have dragged down the happiness of the people I work with, but we all feel imprisoned by it. I have a stack of scrum books here I plan on reading, I figure this process isn't going away and I need to up my game with how to play it - but I wish I could use something else, perhaps kanban.
Spoiler, your Scrum books aren't going to help you figure things out, unless your org actually wants to change. You have your own unique process, but that you have a unique process isn't an issue. The issue is that your management seems to not respect the input of engineering. No process will work the way you want as long as that is true.
Some thoughts, * Most critical, your team is way to big. Should be half as big
* Scrum master should be a team member and rotated along the team. The process should be a result of constant tuning via the Retrospective. The scrum master here, should lose his job.
* Engnineering owns the implementation of the stories, product owners should not be a part of the conversation on how to do something, they only can say 'what'.
* If you can't decide your own points, then again, you have your own process. If you can, and you need to refactor, you need to add points for that.
If you can't do anything to change, then again reading scrum books isn't going to help. You can say, 'this isn't the way your supposed to do this' as much as you want. processes always reflect the organization. Unless the engineering team is an equal to the product team, it won't matter what process you have.
I don't agree. For us the scrum master is a fixed position who has as a responsibility to advise the teams on how to fix problems with their processes or to help build these processes in the first place. This works very well for us, because the scrum master is explicitly only an adviser.
Meaning I, as a team lead, do not have to do what the scrum master says, but I always have someone experienced with scrum who I can ask when problems arise or when the team wants to change the processes.
1) While I agree that 20 people seems like too large a team in general, I have a fundamental problem with the idea that scrum can determine what size team is right.
What seems to be happening, it appears to me, is that we have a fair idea of how large a meeting can be before it becomes unwieldy, and scrum has a daily meeting as it's fundamental component, so basically decides that for it's fundamental component to be effective there shouldn't be more than 8 people so therefore teams shouldn't be larger than 8 people.
That seems backwards to me.
2) Almost every scrum training usually has a separate scrum master. I do think the idea of a team member being the scrum master makes a lot of sense and would probably alleviate a lot of the concerns.
3) Scrum story points are supposed to reflect end user benefits. A refacorin's benefit will only show up after several sprints. In the meanwhile you hurt your velocity significantly because your refactoring is likely to be 0 points. Now, this actually makes sense to me, but the issue is that in most places management is carefully monitoring velocity and is gonna hold the drop against the team.
To be fair though, I think this is another issue where the real problem is management converting a measurement metric into a target, which basically negates any use it may have as a metric.
My biggest issues with scrum are: 1) A daily standup is ridiculous. It's highly disruptive and gives me the feeling of someone constantly peeking over my shoulder. A better strategy is a simple email when someone has an update, or a blocker to the rest of the team. Very few of my projects are done in a single day, and it's psychologically stressful to join every meeting saying I worked on story A, will continue working on story A for 3-4 days in a row, while the managerial types rattle off a list of meetings they attended, emails they sent, trainings they completed, etc. 2) Sprints are far too inflexible. Each story should basically define a new Sprint for the people involved in that story, which may or may not be longer than originally expected. I don't see the point of having a fixed 2 week period that applies to everyone. Why not just have a weekly or biweekly meeting where you go over everyone's stuff and see where you stand. It's basically a more flexible version of what sprints try to achieve, without the psychological issues created by having to fit things into a Sprint or alternatively having to carry it over into a whole new Sprint. That same meeting can track how many points were earned since the last meeting and you have your velocity being tracked as well.
Retrospectives are great. But they should be included in the same 2 week catch up meeting.
I find scrum tries to create a lot of artificial deadlines, possibly to encourage people to break things into smaller pieces and make regular progress, but I find that's not how people usually work. In my experience people tend to work more in spurts, delivering a ton of features and bug fixes over a couple of days, and then going relatively quiet for the next few days. Not because they aren't working, but because thats just how things usually pan out since software development is a decidedly non linear process.
That’s not my understanding at all. In our org, that’s the role of backlog priorities. Story points are the cost to get there, and are explicitly a tool to estimate relative amounts of effort involved.
If tech debt is to be incurred, it is expected that engineers negotiate that, and that other estimates will change as a result (and/or other tasks be created to track that debt). But then, nearly 100% of the managers in our org are, themselves, engineers or SMEs in the field they are managing, so they have realistic goals and understanding of the sausage-making process. What you’re describing really does sound more to me like a management problem than a process problem.
Story Points represent effort and uncertainty, not business/customer value - those are Business Value Points. Devs are only committing to X story points per sprint. If refactoring is needed, it's built into the story points, or added as it's own task. And as others have said if the Product Owners don't take engineering input on paying down tech debt/infrastructure/internal tooling, then you have a broken company no matter what process you're following.
Scrum.org states[0] A Story Point is a relative unit of measure, decided upon and used by individual Scrum teams, to provide relative estimates of effort for completing requirements.
[0]: https://www.scrum.org/resources/blog/why-do-we-use-story-poi...
According to Jeff Sutherland himself[1], story points are based on team effort and not end-user benefit: Estimates are estimates for the team to get a story done.
[1]: https://www.scruminc.com/story-points-why-are-they-better-th...
As an example, one could consider the various people working on an individual open source project as part of a team. And most successful O/S projects rarely require everyone to meet together at the same time.
I personally strongly prefer small teams (I find even 8 members too large), but arguably certain projects may need bigger teams, and I think they should still be able to practice Agile with the bigger team, which scrum doesn't allow for.
Effort makes a lot more sense, and I particularly like that it tries to incorporate uncertainty as well (I guess that's why they chose story points, to make it vague, since using hours leads to people treating estimates as deadlines instead).
I have never worked at a company where this actually goes right and could not be replaced by [insert ticketing system here]. We use Basecamp to do daily check-ins on a very high priority issue, if we have one, which we usually don't.
I've never figured out why you couldn't just do it over slack or equivalent. It's massively inefficient to get everyone to meet up in the morning.
Think of it from the point of view that managers have mangers they report to. When shit goes wrong they have to cover their asses. "So it's late. But did you check on their progress? Wait just over slack, no actual meetings? Ok, that obviously needs improvement".
All this talk of standup meetings being primarily for managers is really weird. It's not for managers, and they have no reason to be present. It's for the team, to improve cooperation within the team. If the team has a better way to accomplish that, then they should use that.
From the scrum guide... "The Daily Scrum is a 15-minute time-boxed event for the Development Team."
It is the team's event, if there are other people there it should be at their request. The guide goes on to say that.. "If others are present, the Scrum Master ensures that they do not disrupt the meeting."
Stand-ups and all the process attendant appears to be universally imposed from on high as a micromanagement technique, or because they were sold the idea as some kind of silver bullet by a talk or a snake-oil consultant.
I feel like Scrum is for teams that suffer Stockholm syndrome from poor management and need to be taught how to human beings again.
The main value of Scrum as an official process, is that it gives you a stick with which to chase toxic management out of the room. They're not sufficiently Scrum if they crash your meetings and demand to be in charge. Though in my experience, most companies that do Scrum have a management that keeps their distance unless invited.
And to be clear, when I say "blocked" I'm not talking like "I can't get any work done until X", I'm more talking like "this particular avenue needs X to happen before I can continue it, so I'm shelving it for now and doing something else until we can get it resolved." If you find yourself in the former case, something has already gone horribly wrong that should probably have been resolved days ago.
Smoothly running teams don't need this, but many teams don't run that smoothly, and then it's better to address it at the stand up than not at all.
If your daily stand-ups are just repeating the JIRA board, then just give people back their time. If your features are coming out badly designed, then stick a designer in the engineering pods, or call out the design phase in project lifecycle.
If your projects are consistently late, then... slow down the estimates!
Of course teams that do it right tend to know this, but the bad implementations of scrum I've seen consistently lack attentive iteration on the process itself.
Otherwise, here's a couple of notes freely given with no claim to excellence except that I've worked with successful scrum teams and unsuccessful ones:
* The 'business' is everything. All code is liability. Code that can't do things the business wants to do and hinders those things is a particular cost to the business. Product Owners who don't understand that "refactoring code to make it extensible before extending it" is the cost of doing business just have to learn that lesson. If they won't, let them go. Bad ones are a dime a dozen so you can always hire another clueless one if you want. It's like having a loan that you're paying lots of interest on. Saying "We can't spend time refinancing that. We have to go earn money." doesn't make sense as some absolute truth.
* Engineering is not in contrast to the business side. There is no need to represent engineering because the product owner is supposed to be able to understand all inputs into the program. If engineering is slowed down by the lack of dealing with some technical debt that's not some engineering problem divorced from the business. It's the business. This requires reaching across from the engineering side, being clear when things aren't going well, not starting on ridiculously long "refactoring" projects that are just lateral changes, and being clear what the expected outcome of any tech debt relief is. Avoid "feels cleaner", "is more elegant", "is more extensible". Talk in terms of outcomes: faster response to outages, the ability to add things like X (requires you to understand the product some, which is something an engineer should do), less time spent debugging issues. If trivially measurable, measure and demonstrate improvement. Builds trust.
* I'm not too convinced "Scrum Master" needs to be a solo job for anyone. The incentives are skewed. If you have a team of ideal people and they all pick up on the principles and you correct the bad patterns they have, they won't need you. The job inherently obsoletes itself which means the median person taking the role will entrench themselves in process to keep themselves relevant.
* 20 people at daily standup makes no sense. It's 10 minutes or 15 or whatever so you have 30 s to describe something. It isn't so much the time, though, but the fact that there are too many people here together in the room. It makes no sense. Find smaller units of organization. Conway's law presents but the alternative is much worse.
Indeed the commitment is supposed to be with the sprint goals. I've seen this vital aspect of scrum getting overlooked too often.
In Scrum, it's retrospectives.
Continuous improvement is a theme that runs across agile, scrum, lean. Without giving feedback, how are you going to improve?
Worse, the moment anything of interest gets being discussed, SM stops it, saying "let's take this offline" (in German-speaking projects at least, where there's a tendency to use English words to really say nothing at all). Which brings us to
> any deviation from scrum itself poses a threat to this person's job security, so it doesn't happen
MBAs have identified and captured the role of a Scum master (sic!) to comfortably represent a project towards management, without the responsibilties of a project manager (I'm just moderating the project), yet with an instrument to turn time-based freelancing and employed work into a fixed-price deal through commitments/forecasts.
But I've also been in projects where standups drag on and on and on, and they're hard to get under control. Usually, those projects had much larger teams.
I suggest smaller teams, and if you have this kind of multi-team standup, discuss with your team to have your own independent standup instead and send one representative to the big one.
Change them so that people talk about what is blocked then.
> Scrum master
Tricky, like anyone who doesn't want to adapt.
> Technical debt
Create tickets for debt, put them on the backlog.
> Poor product owners
Tricky.
That will risk making a huge technical debt section in the backlog that will find it hard to get priority. It's better to condition new features/bugfixes on the cleanup that caused them. "I can add this feature in 1 day but will need 4 days to clean it up. So my estimate is a week." I constantly nag developers to do this when they find that they are pressed for time and want to make a hack.
-"I'm nearly done but I can't find a nice way of doing this because of an existing code smell so I'm considering adding another smell"
-"Did you ever find that this was NOT the case?"
-"No..."
-"So do you always double or triple your estimates so you know you can comfortably refactor"
-"No..."
I think this is a disease of Scrum. Too short estimates exist because developers pat themselves on the back for finishing things quickly.
Keep technical debt in separate tasks to make the tech debt visible and to keep management accountable for not making efforts to get rid of it.
Essentially the "process" was:
- Someone gets an idea to do something, e.g. PMs wants to add a feature.
- The manager (maybe with the help of a dev) figures out which teams need to be involved (e.g. dependencies).
- Get a very rough estimate from developer. Are we talking a few days, a few weeks or a few months?
- Manager, PM and other teams get together to figure out priorities, schedules and who's likely to be working on it.
- Based on some discussions, areas of expertise etc, some number of devs gets assigned to work on this feature.
From here the devs know what they're trying to accomplish, what the constraints are and who to go to for questions (PM, design etc). They also know who else from other teams they're collaborating with and they just figure out amongst themselves how to get things done. They'll keep their managers and the PM updated on progress and any blockers.
Every once in a while manager etc have to step in. For example if some big that was previous unknown came up, there is a big risk, priorities need to be adjusted etc. But for the most part things just worked really smoothly.
Sometimes we need to give estimates and really hit it (legal or security issues, big marketing launch etc), but for the most part we were trusted to be doing things as quickly and efficiently as possible. So none of the commits, sprints and burn down charts BS. It's not hard to gauge develop's productivity based on output anyways. If there was something slowing the team down, we communicated the need to the manager and PM, and we worked on fixing it.
So for me, if you have good competent people that communicate well, you don't really need much process.
As opposed to trying to create a couple of metrics because you don't really understand what people around you are doing.
We are better off building over-all engineers that know what they are doing, than trying to catch up with the latest shiny thing (Scrum, Agile, Waterfall, Kanban, ...Jesus Christ)
Look at Spotifys videos on their process. Every team is self-driving, because everyone knows how to do stuff. If you work at a company where there is no process, and not every is able to complete the whole task from A-Z, you end up with deadends and people who get stuck. If the culture doesn't encourage knowledgesharing, people end up making crappy solutions.
At the end was there ever an instance of the client not using what you produced because it wasnt what they wanted?
Like anything else there is no silver bullet. Sometimes you got the feature right, other times it misses the mark.
The Agile Manifesto describes a set of ideals but gives no true set of instructions to follow, to do so would not be the Agile way. More than anything, it prescribes an attitude around which you should generally approach things.
Scrum conversely gives rules for how things can be organized and executed in a functional organization. It gives little room for flexibility. Any problems you have with Scrum, the first reaction should be "how are we following the rules wrong?".
Similarly, Tao and Agile are ideals for the individuals (and small teams), Confucianism and Scrum are rules for getting things to work functionally within a society/company.
Morality is society's rules for individual survival.
Ethics is the individual's rules for society's survival.
- daily standups (often one per day per team so multiple per person)
- sprint based development cycles, often with retrospectives
- fondness for the "as an X I want to Y" story
- story points defined in terms of developer hours or days
- Kanban style "pick a thing to work on", even in sprints
- burndown charts
Seems nearly inevitable that when anyone here says "we need more process" they're looking to have more of one or more of the above. I essentially never hear anyone in project management say anything like "we need to build more prototypes" or "we need to streamline our deployment process" or any of the other concepts that would actually characterize a (lower case a) agile development process.
(edited for formatting)
Its popular because managers can implement it. Agile- the manifesto version- was intended as a way to re-introduce the notion of discipline that newer generations of programmers had lost (where previously they were engineers and scientists who wrote code, now people who code exclusively eithout other backgrounds).
It feels like we have come full circle to where the new generations of programmers have delegated (or lost) all discipline to managers' Scrum plannings.
Managers and "product owners" aren't engineers or programmers. They shouldn't be expected to "manage" a developer's day-to-day, and yet them presiding over sit-down-stand-up meetings and demanding points and determining deadlines and features is precisely what it has come to. We are right back where we were when the manifesto was drafted.
User stories? Oh, a neat way to keep requirements general and open-ended so we can properly address how we're actually going to solve a user's problem. Surely it'll prevent PMs from over-specifying requirements that lose sense of true objectives.
2 week Sprints? Cool, estimates are hard, and now I never have to estimate more than 10 business days worth of labor.
Retrospectives? Great idea, we can finally do proper post-mortems and knowledge-sharing!
Scrum Masters? Wonderful, there's someone whose dedicated to running the process and making sure we have everything we need!
What I wasn't anticipating:
User Stories? But what about critical requirements that need to be prioritized that don't fit into "As an X I need Y"
2 Week Sprints? I now have so much technical debt a repo man is confiscating my laptop.
Retrospectives? This is always going to be 100% about how we didn't estimate correctly and not about far more important matters like: how these features didn't help our users, technical knowledge-sharing, and ticking-out technical debt.
Scrum Masters? Oh, you mean Project Managers?
Probject managers and product owners were kept at arms length in the sense that they didn't dictate how, what or when we did what we did. They translated the business requirements and timelines into something we could understand and react to.
Somehow, all of that turned into two week sprints that felt sustainable... At least until the company bought out another company, and everything went downhill. That, however, is a story for another time.
Here's the thing: I have been lucky enough to work with engineer-founders as my bosses for most of my career. You know what the downsides are? None. It's just pure awesome.
It's getting to the point that they're the only people I want to work for.
People who actually write software understand that you have to optimize for throughput first. Not to worry: latency won't be forgotten! But a primary focus on throughput will result in a clean codebase, that will maximize throughput and minimize latency.
In practice latency is the goto target for optimizing actual processes. It's the most linked with all the risks. But software development is not an actual process.
Exactly. And it's quite wrong even without taking the growth of complexity into account — as every engineer knows, or should know. Getting every task done as quickly as possible requires a lot of context switching, which is murder on throughput. When you add in the effects of complexity growth (aka technical debt, though I think "complexity growth" is clearer) the disadvantages of optimizing for latency become that much more serious. And the worst part is, as the disease progresses and latency deteriorates, managers try to cure it by applying even larger doses of the poison.
This idea that managers optimize for latency, while I optimize for throughput, occurred to me only recently. But as I look back over the disagreements I've had with managers through the years (including disagreements over the usefulness of Scrum processes!), it's quite remarkable how many of them seem to come down to this.
These books are not about software development, but of product development in general. The first one actually predates agile, published in 1997.
Core idea is that minimizing the size of the tasks is the best way to improve productivity. Not getting it done as quickly as possible, but to decrease the size.
Remember that product development (new, innovative, uncertainties) vs product manufacting (repeatable), is not a new problem, and not unique to the software industry. There's a lot to learn from product development in other industries.
Another interesting read is "The Toyota Product Development System: Integrating People, Process And Technology" which talks about ways of making product development predictive, and less risky.
It's interesting to see how little software is used to improve the process of software development. Other industries use a lot of software (cad/cam, visual modelling, testing, impact analysis) to improve efficiency and quality of product development. Software for product development is a huge market.
The problem is that except on the bare minimum, latency is a bad proxy on development projects. So the idea is perfectly correct, yet it's useless.
But we spent one entire workday (7 hours) on each sprint follow-up meeting, and then another entire workday planning the next sprint. That is what it took to write the stories, break them down into one-point pieces, prioritize with the PO, pass the stories out to the devs, etc.
Most places just don't plan for enough time to do things right, and quality is second priority, so no matter what methodology you use, you lose, because no cared about doing it right to begin with.
To be clear: are you saying that the one team that did Scrum well did so because they spent more time on the process? Reading what you wrote, spending two full days every two weeks to plan sounds, well, terribly dragged out. Does it feel like the time was well spent, or was it a slog?
By the end of the planning meeting, we had a clear idea of what we were going to accomplish, and a reasonable amount of confidence that we considered all the tasks that went into our plan.
Because when a story was, e.g. "add addresses to the clients page", it was broken down, discussed, thought out, and agreed upon. The moments when I discovered, oh shit, this story will actually take 10 hours longer than I had allocated, and now I have to stay until 10PM two days this week to meet my committment. Because our story points were approximately an hour each, and almost every story was broken down until there were no tasks more than 3 points each, 1 or 2 preferred.
It was all thanks to our scrum-master/project manager, who had actually spent a lot of time learning about scrum/agile/kanban/etc, read many books on it, and most of all, was committed to doing it right.
I think our sprints were a bit longer than 2 weeks.
During the project, it felt like a bit of a waste. We were spending a full 10% of our time on project management. However, looking back, I see that a) it seems to match up with other's experience, and b) it's pretty much the same amount of overhead for project management that the project would have had using any methodology.
At least with the 1-2 days every 2 weeks everyone sees it and it's something you can get better at. I'll take that over magical GANTT charts any day.
Creating fine-grained, detailed user stories and making sure that everyone understand them and agrees on the prioritization is not time spent on "the process", it's requirements engineering.
I would say the top, most useful, make-or-break practice that I would say is most essential to Scrum succeeding is a combination of:
* 1 point ≈ 1 hour of work
and
* no task above 3 points, try to only have 1-point tasks
Coincidentally, this process is what took the longest in our planning meetings, because the coders sat down and planned out the tasks, kind of the way you would in an algorithms course.
The payoff is that our estimates, after the first couple of sprints, were dead on, and there was very little "discovered work" mid-sprint. No midnight oil. And no corner-cutting.
At the end of each sprint/cycle we shared estimate vs actual effort, to improve our ability to make a good estimate and set an overall velocity.
On my first Scrum/Agile project, there were complaints of overwork the first sprint. Then I pointed out that we were the ones setting our own estimates, setting the pace and causing our own problems.
After that, the estimates got reasonable. We stopped playing "Name that Tune" with our estimates and the project settled down.
Maybe it exists and I am just not reading the correct articles.
[1] https://pdfs.semanticscholar.org/85d0/00404206914501d26e0bb4...
[2] https://www.researchgate.net/publication/261047173_Agile_and...
People attribute failure to the process because it is demeaning to put the blame on people when established best practices can receive the finger pointing.
Stop the notion of being nice to the lack of skills of people. Hiding behind being nice does not resolve the issue but rather propagates it. Fix the issue by providing the grounds for people to learn and become productive.
> Stop the notion of being nice to the lack of skills of people.
Projects staffed with fully qualified people with hardcore skills fail too. And projects staffed with utterly under-qualified people sometimes do just fine.
The thing is projects succeed or fail for many different reasons, usually multiple reasons operating in concert.
I think the best approach is not to be dogmatic about process and to recognize that "pointing fingers" rarely resolves anything, regardless of whether one is pointing at the process or the people.
- the dictionary definition of "agile" - how many times the word "agile" appears in the Scrum Guide - how "lengthy" the Scrum Guide "seems" - that Scrum is a process, so does not "sound Agile".
Ugh. How did this reach #1 on HN?
Then ppl came along and said that they were "too complicated" we need something Agile! In actual fact, those ppl didn't understand the process frameworks and just created a specific instance of one and they wondered why it didn't work for all projects, organizations, technologies, etc. And then they said those processes are "too simple" - what can we do? "Maybe we should create a way of generating a process?", they then said. And so it went...
Agile is a mindset.
Understand the mindset, absorb it, make it yours, and build the process that's working for you based on that mindset.
Stop complaining about Scrum, it's just a tool, and tools deprecate.
Since everyone is doing, or claims to be doing, agile, the whole notion is completely meaningless. Unless they are actively advertising to be doing some form of waterfall, you can safely assume there is some notion of iterations involved.
What matters more these days is whether the team is mature enough to have continuous deployment without process bureaucracy. Deployment fear is a good sign the process sucks. If you have a need for human gatekeepers, something is wrong with the test automation. Deploying often and with confidence is a good sign you are dealing with a smoothly running team.
ACM recently had a series of webinars[0] by Ivar Jacobson on Essence[1]. Essence was kind of confusing at first, but it is essentially a language to describe practices from the different processes. The idea is that you can build up a library of the various practices in your company and allow teams to pick their own and evolve their process by swapping out practices that just don't work out. Essence seems like an interesting idea, especially if it allows teams to create a development process that fits them.
> find what works for your team
For the combination of team and project. Different software projects need to take very different tradeoffs in how they managed. Use scrum in aerospace for flight control, and people will die. Use scrum for system software, and you'll have hard time delivering quality and performance, system software often requires non-trivial amount of engineering. Implement NASA's best practices when working on videogame or web app, and you'll blow budget.
This article is from 2002 but still good: https://www.joelonsoftware.com/2002/05/06/five-worlds/
Every team at my workplace runs a different process. We’ve got a team developing new features who are running an iterative agile process, allowing them to react to changes ok requirements or unexpected issues with as little disruption as possible. We’ve also got a team doing B2B projects, essentially rebranding our mobile apps and platform for different companies, and that’s pretty traditional waterfall - I wouldn’t in a million years use waterfall for new features, but when you’re running through a set of well understood tasks that need delivering on schedule to coincide with work by third parties it gets the job done.
I hate artificial deadlines (i.e., sprints). Kanban let's you focus on the work without all the process distractions.
Is there anywhere this would be documented to explain - a. How an agile process in initiated? b. How an agile process is maintained?
I have been part of a team with mis-applied agile processes -
1. Requirements change frequently. And not due to users asking for them. More like the Project Managers asking for more configurablity.
2. Requirements are not thought through. Even the simplest cases.
3. Changes don't get vetted by users. Instead, we go through another round of requirements!
4. Every once in a while requirements (and teams) get re-organized, re-planned which puts all the developers off.
I have also been part of a team which focussed on deliveries. This team has been able to write tests and iterate over requirements fast. The satisfaction level of all the developers in the team as well as the project owners were high.
1. Requirements were phased.
2. Whatever was required to be done in each phase was thought through as far as possible. Anything that wasn't clear or was more complex to think out was sent back to the project owners / users for clarity.
3. Code and tests get written. Phase delivered.
Although this process was extremely successful, I don't know if I would call this Agile.
It's agile up to the point that it is broken, which happens often due to it's fragility. That it is easily broken is an unfortunate but unavoidable consequence of its complexity, and common misunderstandings/misuses.
Scrum is complex because large software projects (large as in number of people, budget and expected velocity) are never simple to manage. Scrum is complex because correct usage can not be read from a book nor gained from a qualification - it requires experienced judgement to pick the optimal usage and apply it to a team and project.
Nothing can beat the effectiveness and efficiency of a single strong developer working on a stream of well considered backlog items. That is engineer nirvana! However many projects simply require more velocity than one person can manage. From here, there is a whole spectrum of processes that should be chosen depending on resources and needs. From my experience, full scrum kicks in once you have four or five engineers on the same system.
Product management is a key point of failure in Scrum, true also with other processes. But it's felt so much more when an entire team grinds to a halt and an iteration fails.
If Scrum doesn't feel agile then you are doing something wrong. This is exactly the question the team should be asking itself at retrospectives (doesn't need to be three questions!). What's changing in your process sprint to sprint? Scrum masters are there to facilitate incremental process changes driven by team members, not a system that doesn't result in agile software delivery.
SCRUM is meant to be owned by the developers - not management. Shame on companies that hijack and impose such process.
SCRUM gives dev teams rhythm with sprints. Allows them to deliver mini releases of functional software in 2-week increments (could be 1 to 4 weeks too - lenght should be decided during planning). The goal of a sprint is to be decided by the developer and doesn't have to be about the software. Can be about learning, improving as a team or helping team members to grow. Here are a few examples I've seen: - Measure 5 metrics in Kibana - Marie-Pier leads sprint review demo - Try self-merging PR's and measure the results - No more loose ends Of course, in parallel the team builds and delivers an increment of functional software.
The planning, review and retro rituals are paramount. SCRUM can only work well if those are well planned and executed. The retrospective is particularly important to allow the team to improve themselves. Its an opportunity to change their own process (i.e. change the definition of done).
Contrary to wide belief, story points are not part of SCRUM. Its a mechanism that can be used to gauge the size of work - it should be up to the dev team to decide whether to use them or not. Again dev teams should own that process - should not be imposed by management.
These lessons I've learned in the past two years thanks to an awesome SCRUM evangelist we've hired. I'm proud to say its changed our lives for the better.
Meetings should only be between business / leads to get scope onto canvas.
If I'm blocked I'll seek out the person who can most help me. I can easily find 5-10 minutes to write out a summary, but on my terms when I'm not in coding mode.
Now, I'm freelancing, likely to be team lead for the company assuming they get financing - and I keep taking notes and reading ways to organize better. We have a ton of work to do, it's a financial mobile app that has 0 tests on the api (I'm single-handedly running the API development Laravel + vue (for some web views)). Money is tight so I'm the only dev on web side and we have 1 ios dev.
The hope is to be full-time at this by Christmas w/ a 6 figure salary (my first +6 figures).
From what I've seen, projects almost always start out in a requirements/analysis phase because management, stakeholders, etc., need the reassurance. I've yet to see any public-facing project of meaningful size start out releasing actual MVPs as opposed to "big bang" releases.
Design often happens first because engineers don't want to(and sometimes can't) begin coding away and then receive designs that are either counter to their existing work or are unworkable; management usually sides with design because, well, it looks cool.
Engineers eventually get coding, but non-automated testing doesn't really occur until later because it can be impractical to have people testing unfinished software. Then there's alpha testing, users break the software, engineers fix bugs, A/B testing of different versions of features, then beta testing, where users again break the software and the engineers fix more bugs.
The project continues to drag on because the first 80% of a project is always the easiest part, and management is always hesitant to release "unfinished" software.
Inevitably, there's a big bang release, by which point management has already gotten bored and has dreamt up other "big ideas" for the team to focus on, at which point the project is, for all intents and purposes, placed into maintenance limbo where bugs are fixed, junior developers place their awful code, and a new "feature" is added now and then to satisfy the marketing department.
Senior developers rationalize the waterfall-like nature of the project, so they opt for "continuous integration", but all that ends up translating to is maintenance limbo without versioning.
The actual next version of the software doesn't get built until after most if not all the original team left, and the original software became "legacy" enough that bugs keep popping up and the new engineers don't want to touch the old code. Management eventually gives in, for better or worse.
Am I wrong? This is what I've seen happen to the vast majority of software projects, all under companies that were either Scrum, Agile, or pretend "Agile". Maybe I've just had bad luck.
I wish more teams could pick and choose what methodologies to use for their purposes, rather than buy into MLM garbage.
However the article is not wrong, it's too easy to make SCRUM all about the process, without changing anything about interactions, developer empowerment and feedback cycles.
I think the SCRUM doesn't do a good job of putting limits on what is still agile SCRUM and what is not. Maybe because it's written in such a friendly language.
Unfortunately there ain't no such thing as a free lunch, and when you emulate you necessarily incur overhead, which can be prohibitively large if you're emulating something powerful -- plus, if you have actual good programmers on your team, they will either quit or start emulating mediocre programmers to remain compatible with your mediocre-programmer process. Either way, you lose the benefits of having them.
TDD/yagni/refactoring allows you to keep your code base nimble. I can't imagine doing an iterative process by slamming features in one after another without refactoring the code as you go to handle what "you now know", or not having good test coverage to support the changes you need to make.
Refactoring also shouldn't be something the PO is aware of, it's not a story, or a 2 month break from feature development. It's part of the job as each story is implemented. Yes, sometimes you don't notice a good way of doing something until some time passes and that area of the code becomes a bigger refactor than usual, just have to deal with it as soon as possible.
It has good principles like backlog, sprint and daily meeting. If teams only pick one of those or anything else and that it help them, great! By the way holding solely Scrum accountable for failures is narrowing down the analysis to only one part of all things that can go wrong and have nothing to do with Scrum. That's a cognitive biases concentrate.
I wish there was a prescriptive way to successfully make and deliver software. I have seen very little evidence that there is.
In my experience this seems to be an issue unique to enormous corporations. When I used to work at small / medium-sized companies this never happened.
Agree with the Software development priority queue part being good thing.
Long term planning was in the waterfall process which does not work that great either since it tends to over focus on planning.
How do you find a good balance between Agile and Waterfall?
I think it can also be different what process works well if you are a startup and need to deliver a MVP minimum viable product as soon as possible. Compared a bit to if you are a long term enterprise company and need to maintain your code base over time. Ie dealing with technical debt from short term solutions vs long term stable.
I think one should have experienced a successful scrum project before being able to criticise it.
Scrum is a tool: you may be Agile without using Scrum or you may find easier to be Agile using Scrum. Doing Scrum does not make you Agile. Scrum is hard and implementing it takes different skills than those proven by individuals who can write code.
What's the better alternative? What have you been doing to make things better? The answers to these questions as much more interesting than endless complaining.
In the early 1990s, Ken Schwaber used what would become Scrum at his company, Advanced Development Methods; while Jeff Sutherland, John Scumniotales and Jeff McKenna developed a similar approach at Easel Corporation, referring to it using the single word Scrum.
This is a many times debunked myth about the Agile manifesto. Even the authors themselves explained that "Individuals and interactions over processes and tools" does not mean and shouldn't be taken as "no processes".
Also, agile is not about moving quickly but about being able to easily change direction any time, therefore dealing better with changing environment.
Scrum is a problem when it is mandated and imposed on functional organisations that do better and more fluid communication. This is often done as a way of flexing corporate power and does huge damage.
A key problem is that we do not have ways of auditing and measuring the performance of software teams no matter what methods they are using - until then this is all anecdote.
Of course we have ways of doing that - one is the code quality, another one is historical performance on agility.
1 - Some people said that this can be a failing of scrum and I 100% agree.
Some large orgs really work like that, where the approval of one ham-fisted iron-gripped executive is deemed more important than hundreds of thousands of dollars of engineering talent going to waste. I don't think this is very effective but alright then.
Everybody else should just be using Kanban boards with periodic review.
The TLDR of Scrum is simple: (product) management gets to set the priority of things every 2 or 3 weeks. After that, we see what got done and check to see if the priorities are still the same.
If the priorities are wrong, don't blame Scrum, blame management. If tech debt is increasing, don't blame Scrum, blame management.
There's no magic bullet to determine what is important, and certainly Scrum Master training won't help an incompetent manager a competent one.
Of all the people complaining about Scrum it doesn't sound like any of them have read the Scrum Guide. Symptoms of Scrum are not managers pressuring people at stand ups, squeezing 4 weeks of work into 2 and having a "team" of 20 people. These are symptoms of poor leadership.
The problem with Scrum is that it's usually (always?) implemented as serialized waterfall projects with completely arbitrary looking deadlines (i.e. the sprint's length). All too often things get sort of done but not done quite as well as everybody would be happy with. It's stressful for everyone involved and, insofar as I've seen it used, it makes accumulating technical debt even more likely than more traditional waterfall projects.
What a virtuous cycle
A.k.a. common sense :)
Usually it's hack together something and see if it sticks, which is awful.
But isn't that exactly what Scrum wants you to do? After every sprint, you spend some time to reflect, look at how you're working, look at the bigger picture, etc.
Scrum is by no means perfect; it's a tool, not the infallible silver bullet the author wants it to be. If you misuse the tool, you're still going to get wrong results, but that's the same with every other method.
The real question is: is there a better way to do it that can easily and reliably be implemented by large corporations? I think the popularity of Scrum is probably due to it being more successful than what was used before.
It might not be truly Agile, but for large corporations, being truly Agile may be a bit much to ask. They need reliability and reproduceability, and that means they're always going to need some focus on processes and procedures, and can't always rely on people who might leave or have something happen to them. Not that Scrum is always a good fit for those processes and procedures, but it does help make it appealing for large companies, and at least it puts a good part of the process in the hands of the people who will be using it.
tl;dr: Scrum is not perfect, but what is?
and, I never saw so much interpretation of what something should be as with Agile.
not everyone is entitled to say or mean something, that includes myself and this post.
on the contrary: you have a PO in charge of communication and priorization of features and a dev team in charge of developing- no manager on top...at least in theory...
Without that info, this just feels like buzzword bingo, a la "What should I write about to stir the pot today?"
Speaks a lot bout what agile truly is and how scrum is a very large mismatch.
If completing "the plan" is seen as success, then of course sticking to the plan is the best way to success. But if maximizing business/user value is the success, then "no plan survives first contact with the enemy." (I don't like "the enemy" concept, but this is a famous quote that gets across the concept).
And I think the reason why "agile" fails is because many actors in many organizations find they are effectively rewarded for "completing the plan" regardless of business/user value, indeed.
“Sticking to the plan” does not mean rigidly enforcing zero changes, but rather means keeping a commitment to the general scope and direction that was mapped out. Compromises should require extraordinary hard evidence before being accepted.
This is really why Agile as a general set of guidelines is so easy to subvert and ends up being a misused tool in most every situation where Agile is deployed (Scrum of otherwise).
Project management guidelines need to start out by specifying a way that quality is strictly disallowed from being subverted by competing interests that lobby for changing the plan.
If a set of software project management guidelines doesn’t start out with an unchallengeable quality-above-all-else mandate that creates policy barriers to the natural entropy of different interests trying to lobby for why their preferred change has to be made, then it’s doomed to just get politically subverted.
Doesn’t matter if it’s Agile, Waterfall, extreme programming, whatever.
“Sticking to the plan” in the sense of setting up preemptive, high-cost barriers to anti-quality modifications to what was agreed is _the_ thing.
Either way, harping on the idea of “but what if circumstances changed” is mostly just what corporate politicians do to subvert people demanding evidence for the requirement of changes.
It's not even software specific. Which is actually a strength.
Example, one common misconception is the daily standup is a status meeting. In fact you are supposed to replan in standup, it's a tiny planner. Not sure how you can call the ability to pivot every day not agile.
EDIT: Unpopular option, software quality is poor and it has nothing to do with scrum.
If you don't have that level of trust in your team members, sure, formal process makes sense. And I'm all for it if that's what's needed to deliver business results successfully. But you're then valuing processes and tools over individuals and interactions, following a plan of when and how you may respond to change over letting individuals figure it out, etc. Again, it's perfectly okay if it's not agile if it's what the business needs. But my assertion is it's not agile.
Agile is a subset of TPS (toyota production system)/Lean applied to software.
Kanban for example is another instance of the agile manifesto.