[i_have_an_idea]

Agile/Scrum is one of those things that are impossible to criticize. You can come up with a well-reasoned critique of Agile and there's always some Agile evangelist that pops up to tell you that you're not doing "real agile" and, therefore, your experience is invalid.

At the same time, I have now done software engineering for over a decade, in many roles and teams, and I have never seen Agile or Scrum to lead to the development of a good piece of software. I guess we were using it wrong.

[thimp]

I’ve seen agile done well once. Yes it does happen. The process was not described, spoken of or even considered. It just existed between a few like minded decent engineers.

Their manager got an “agile PM” forced on them and it broke.

The problem is charlatans, dictators and career ticket shufflers that deliver little to no ROI and generally abject chaos.

[hkt]

Exactly the same thing here. Once some interloper with paper qualifications and no tech background rocks up and says "this is how agile works!" you're doomed.

The only time I've ever seen this not happen was when we had a 55 year old delivery manager who viewed his job as facilitating team meetings in a way that meant everyone got the chance to speak and that consensus formed. Didn't try to dictate process or anything, just made sure everyone got heard and that decisions got made. He was patient enough to wait and not force the issues, too.

[praptak]

Organisations make it very hard to be that manager. On top of the human tendency to overmanage there's often push from top for "more visibility and accountability" which results in forcing stuff rather than facilitating.

[hkt]

Blame Mckinsey and their ilk: "what can be measured can be managed".

(I refuse to utter the above phrase without accompanying it with tjis: "not everything that counts can be counted")

[thimp]

It's worse than that. Any sufficiently motivated person can phrase data any way they choose, particularly when their bonus is tied to it. Nor does the consumer of the data always validate that data. I've never seen a manager read a report they requested. That is on top of your suffix which I agree with a well.

[nothercastle]

Not everyone deserves a voice in every conversation. We had an office assistant once’s that had really strong opinions about how to implement features. They were not well informed.

[Scubabear68]

The charlatans are a serious problem in software engineering today.

[gitfan86]

The good news is that it is easy for engineers to start their own companies and keep these people out at first at least.

You will still get cold emails and LinkedIn messages from these people but you don't have to respond

[thimp]

It's somewhat more difficult to start their own companies and eat though which is why a lot of us eat shit and work with project managers.

[whstl]

I also had perfectly operating squads crumble to dust because of PMs, more than once.

I still don't understand the role of PMs when there are engineering managers, product owners and self-organizing teams involved.

[nothercastle]

I honestly don’t understand how teams in large companies operate without pms. In the initial 5 man design phase sure skip them but when it need 10 groups to deliver some small change that they give zero fucks about the pm is there to keep everyone annoyed enough to do their job.

[whstl]

It works easily without PMs if you have engineering managers and product owners.

[DrNosferatu]

Quality engineering has to be *repeatable*.

If it’s not described, and only works via the “minds of a few engineers”, you get amateur treehouse-quality engineering like the 737 MAX.

I guess chance can make it work well like that “once” sporadically, but when you need to touch that code again to maintain it, we go back to the amateur treehouse.

[thimp]

I think that's a vast simplification. It doesn't have to be repeatable at all. Objectively repeatability is about elimination of waste and expenditure. The big one is qualification and testing because that's what filters all the crap out. Actually I've worked as an engineer where we had to throw 20% of the products we built out because they didn't pass qualification.

And of course that is where the cost savings are usually made because it is seen as a null function in some businesses. It creates waste and reduces delivery and ROI. Those are all symptoms of other problems in the business but that's how it is generally perceived. And sometimes clients are happy to receive muck if it's cheap.

But that's also how we get planes that crash and Microsoft firing their entire QA and delivering shit for the last half a decade.

And that in turn is because management and project management culture is data driven by people who have no idea what the fuck they are doing whatsoever.

[EchoChamberMan]

I once had to explain to a manager of hundreds of people why reproducibility in engineering (and science) is important. Not good.

[bruce343434]

What has been, in your experience, a repeatable process for quality engineering? How much does it depend on the kind of software one is engineering, or is it applicable to any kind of software with any kind of constraints?

[EchoChamberMan]

Repeatable processes require long term strategy.

[edit] Too many of these people want tomatos, so they plant a tomato, when really they want seasonal tomatos, and need to consider a farm, and take into account growth cycles. The term I like is "sustainable development."

[bruce343434]

I was hoping for more actionable, or at least concrete, advice...

[EchoChamberMan]

Some complexity is irreducible.

As mentioned by user "phkahler":

> 1) Come to work.

> 2) Look at the current state.

> 3) Decide what the product needs from you.

> 4) Do that.

> 5) Use git.

> Steps 2 and 3 may involve communication. Step 5 is tracking changes. Have a PM that tracks main things people are working on and estimated dates (not dictated dates).

> This is how my current job works and we are unbelievably productive.

[edit]

I like the book "phoenix project," and I think they wrote "Team Topologies" which I am curious to read.

[edit]

In general, psychological safety was the number one factor for productive teams, according to a massive google study. From this viewpoint, it's easy to see why there are so few productive teams, as there is little to no safety for most people. Remember though, "There are no silver bullets."

[Edit]

The goal should be continuous delivery imo.

[web3-is-a-scam]

I work at a company that did Agile/Scrum ok - it wasn’t perfect but we had regular retros and we tried to improve our processes and we were given some leeway from management to have some of the work required to support those processes into our sprints.

Then the CTO left, and was never replaced, then the layoffs came and a competent PM was replaced with a junior. Entire design department replaced with a single junior as well. Engineering gutted. Ok, lay-offs happen but at least I still have a job.

Then came “we’re waterfall now”. You know what didn’t come with it? Actual planning, documentation or specs or anything typical of “waterfall”. Estimates are now guessed before anything is defined and turned into commitments. You know what we still had? “Daily stand ups”, “backlog grooming”, two week “sprints”, and no retrospectives. No opportunity or agency to try to make things better - just endlessly stuck in my own personal hell.

Now we have projects that are supposed to take 2 months now taking 6. We have last minute spec changes that require massive rewrites because UAT isn’t done until the end of the entire project. “PM is slow to write out requirements” so we’re basically expected to just guess what they are based on very small selection of unfinished design flows.

And of course, executive wonders “why does engineering take so long to deliver”.

I started calling this “scrumerfall” or even ”fragile”. It’s basically the worst of both with none of the positives.

I don’t know what point I’m actually trying to make…just get like ranting and this seemed like a good time lol

[foofie]

> You can come up with a well-reasoned critique of Agile and there's always some Agile evangelist that pops up to tell you that you're not doing "real agile" and, therefore, your experience is invalid.

What I've seen the most is people putting up straw men in place of Agile, and proceed to attack the strawman in spite of being repeatedly pointed out they are pummelling a straw man.

[Jensson]

Problem is that managers actually tells you to follow that straw man.

[turkey99]

Scrum is easy to criticise. Agile less so. One such Agile criticism is about principle “that working software as the primary measure of progress”. I understand the sentiment but it’s not enough, it’s got to meet the customer’s needs. Scrum is far too sprint release coupled. Sure you could separate them, but are you really doing Scrum then?

[pydry]

"Agile" was defined so vaguely in the beginning that it became a kind of template for everyone to project their (often contradictory) ideas and dreams on to.

Scrum fixed that by being as bad as it was precisely defined.

I can get what the originators of agile were getting at but they explained themselves super badly.

[photonerd]

Scrum is supposed to be training wheels for agile, not the end point.

I think a lot of ppl forget that.

[pydry]

I quickly searched for evidence that it was supposed to be that and found nothing except one essay on medium refuting the idea.

One of the nice things about scrum is that it has an official source who defines it and you can look at it to see what it is and what it is not intended to be.

[whstl]

And Scrum is not even as strict as people paint it, honestly. It retains a lot of the good parts of Agile. It is a basic framework for a self-managing team.

The retrospective is where you can tweak the process, but you can honestly change things anywhere. Bad managers will resist change at any cost and will use Scrum as an excuse to resisting change, but that's true for any methodology.

The problem is it doesn't work when there's micromanagement, it doesn't work when there are waterfall-ish parts (eg: PMs not splitting tickets, QAs hogging releases). Scrum also doesn't work when the development team isn't empowered, or has no domain experts. But most methodologies also don't.

[SigmundA]

>You can come up with a well-reasoned critique of Agile and there's always some Agile evangelist that pops up to tell you that you're not doing "real agile" and, therefore, your experience is invalid.

Just the typical "No True Scotsman" fallacy [1], happens all the time when there is no good defense of the position.

1. https://en.wikipedia.org/wiki/No_true_Scotsman

[Viliam1234]

By the same logic, there are no actual quantum physicists, because various "quantum healers" outnumber them. So we need to admit that everything "quantum" is nonsense, otherwise we are committing the "No True Scotsman" fallacy.

Sometimes the fakes simply do outnumber the real ones.

[midasz]

What do you use that does lead to a good piece of software?

[PradeetPatel]

Having clear requirements, good communications with key stakeholders and empower the enginees.

[viraptor]

That sounds familiar... Oh ! That's where agile ideas come from! http://www.extremeprogramming.org/

[hu3]

That existed before agile.

[whstl]

Good on agile for restating what works and putting it in an fashionable, easy to understand framework, then.

But bad on agile for letting charlatans co-opt the term.

[watwut]

Agile does not really works. It sometimes works, exceptionally, usually the more you try to do the by the book agile the worst it gets. Every single time they make agile reform, everything stats to suck and everything turns into ritualized micro management.

[TheCoelacanth]

Three of the original signers of the Agile Manifesto are also the main people involved with creating XP[1][2][3].

The Agile Manifesto was just a bunch of people who were already doing XP, Scrum or some other lightweight process getting together and writing down the similarities between what they were doing. It's a blanket term invented to describe a variety of different processes, not an actual process.

[1] https://en.wikipedia.org/wiki/Kent_Beck

[2] https://en.wikipedia.org/wiki/Ward_Cunningham

[3] https://en.wikipedia.org/wiki/Ron_Jeffries

[midasz]

What if the requirements are super clear but they change? When do you find out and what happens then? I've personally never had a project where the requirements were 100% clear and stayed exactly the same. Even when writing a v2 of an existing project from the ground up things tended to be unclear and subject to change.

[nudgeee]

The point is to use what ever works best for your team, not blindly follow some process that may or may not necessarily work for you.

Just as product requirements can change, so can the way we work. Just like there is not one singular product that solves everybody’s needs, there isn’t necessarily one process that does that either.

[lll-o-lll]

Sounds like you need a process that can be flexible, able to respond to change, some might say… agile ;-)

I kid, but I think the early proponents of Scrum and similar were trying to achieve a loose framework to do exactly what you’re talking about. The modern incarnations of these can be horrific, but the original intent was always to empower teams to make their own process. Ahh well.

Like so many good ideas (democracy, constitutions), you only really know how well they function once people are actively trying to subvert them. Scrum et.al. have failed in the face of corporatocracy. I honestly don’t know if decentralised structures (i.e. teams empowered to run themselves), can ever survive in large corporates. Which is a pity. Cities grow, but companies die. You have to jump off the dying colossus to find the new company that hasn’t yet succumbed.

[whstl]

The modern variations of Scrum and Agile exist to empower certain people to rule over teams without having to do much, if any, work.

The team is self-organizing but person X is the decision maker in the process.

There is an engineering manager, but they will get overruled by person X more often than not.

The framework is simple, but person X will add new rules.

Agile/Scrum says people must understand the domain, but person X is the only one talking to stakeholders since engineers are "not people persons".

[oneshtein]

> What if the requirements are super clear but they change?

You need to communicate with stakeholders then, to change requirements, explain mistakes, set new goals, and reprioritize tasks. Agile (SCRUM/Kanban/etc.) are designed with frequently changing requirements in mind. In SCRUM, this is done at sprint review/sprint planning stage. In Kanban, it's continuous process.

[saiya-jin]

Well then requirements were not 'super clear', they were simply incorrect and it doesn't matter in which way. A very bad start of any project, since beginning you are already putting out flames left and right

[SideburnsOfDoom]

Individuals and interactions, over processes and tools.

[slivanes]

By focusing on the product rather than the process.

[midasz]

How does that work? So you've already went through the steps of figuring out a process, or do you just pick one and stick with it - accepting shitty parts? What's the process?

[phkahler]

1) Come to work.

2) Look at the current state.

3) Decide what the product needs from you.

4) Do that.

5) Use git.

Steps 2 and 3 may involve communication. Step 5 is tracking changes. Have a PM that tracks main things people are working on and estimated dates (not dictated dates).

This is how my current job works and we are unbelievably productive.

[whstl]

I would argue that 3 and 4 already imply 5! :)

And IMO it also implies things like CI/CD.

[JackSlateur]

You failed as soon as you think "process"

A process is a hardcoded way of do thing, which is deficient because it cannot react to an ever-changing world

A better way to handle things is by defining "what" should be done, not "how" it should be done

[Kim_Bruning]

I have the opposite view.

Everything you do -up to and including taking a leak- is a sort of process. Perhaps you don't think that particular example should be a corporate process, but that's not my problem :-P .

You can compose processes to get things done.

Processes can be made mutable or flexible by eg. incorporating decisions or iteration. Especially iterated processes can be very powerful (you can get a lot done with them).

Thinking of things in terms of processes instead of in terms of component steps each time frees your mind to worry about other stuff.

Compare it with dividing a program into functions. Once you've wrapped a task in a function, you can then design using the function, rather than worry about how to write the code from scratch each time. Same with process thinking: you don't have to get bogged down in particulars all the time.

That said, maybe you're thinking of the "Befehlstaktik" vs "Auftragstaktik" [1] approach to doing things. In which case we're arguing definitions instead (which can be easily resolved).

[1] https://en.wikipedia.org/wiki/Mission-type_tactics

[oneshtein]

You should start with the developing of the product as is happens, then use Continuous Improvement process to improve the development process.

You can use SCRUM/Kanban/SaFE/LeSS, hybrid or other Agile methodology, or no process at all, but it will not help when people are not trained to be proper part of the process.

For example, a PM may replace SCRUM meeting with a status meeting, because it makes his job easier, while PM or other M should not attend a SCRUM meeting at all, unless called in by engineers. The proper moment for interaction between PM and engineer are beginning/end of a ticket, sprint review, and sprint planning.

[euroderf]

Last time I checked, Agile had zero to say about documentation. Is there any improvement on this front ?

[EchoChamberMan]

The second value of the agile manifesto is "Working software over comprehensive documentation." It is important to remember also "That is, while there is value in the items on the right, we value the items on the left more."

https://agilemanifesto.org/