These are all accurate, except the first Waterfall one, who also doesn’t go to Mars.
Right. They design the whole rocket, spend years to build the rocket exactly according to the design doc, then the rocket explodes on the launchpad and they have to start all over.
We recently saw waterfall versus agile in actual rockets
Blue Origin spent years meticulously designing their rocket. They tested it on the ground. On the first flight it got to orbit, but the first stage exploded while re-entering
SpaceX started building their rocket out of carbon fibre. Changed to stainless steel. Started flying subscale demos, flew high altitude full scale examples to find if their aerodynamics was right, and haven’t actually tried for orbit yet
Blue Origin is trying for a last generation rocket (where the first stage is recovered) but bigger, SpaceX is trying to create the next generation where both stages are recovered
That’s why testing comes before launching.
The build phase took too much time, you now have 1 day to test all the features and design elements of the rocket, because launch day is tomorrow. Good luck!
So, we need waterfall testing to be separated?
Test-driven development: You spend all your time building a gizmo to tell you if you’re on Mars or not. A week before the deadline you start frantically building a rocket.
TBF the analogy is especially strained for that one. Per another commenter, Boeing actually makes rockets with waterfall, but test-driven only really makes sense for software, where making local changes is easy but managing complexity is hard.
Edit: Actually, there’s even software where it doesn’t work well. A lot of scientific-type computing is hard to check until it’s run all the way through.
That’s where digital twin engineering HOPES to bridge the gap.
There is definitely a contium of how long it takes to build and test changes where increasly abstract design makes more and more sense vs the send it model
Waterfall is more like: You want to go to Mars. You start to build the rocket. Managers that don’t know anything about building a rocket starts having meetings to tell the engineers who do know how to build a rocket what they should be doing. Management decides to launch the rocket based on a timeline that’s not based in reality. Management tries to launch the rocket based on the timeline instead of when it’s actually finished. Rocket explodes. Management blames the engineers.
The various methodologies don’t actually change what the engineers need to do. But some of them can be effective at requiring more effort from management to interfere in the project. Bad managers are lazy so they’re not going to write a card, so they can be somewhat effective in neutralizing micromanagement. I say somewhat, because bad management will eventually find a way to screw things up.
Waterfall: Boeing/ULA does this. Their rockets cost $4B per launch, don’t work, strand astronauts. Maybe the next repair/test cycle, if management’s dumb enough to keep paying them.
Agile at least launches something.
Yeah, but waterfall requires that management knows what they want. It’s impossible!
This but unironically
So often it’s patience from stakeholders to allow for time to actually design and build the things, or willingness to admit the actual cost, or an impossible grand vision with an unqualified/understaffed team, and of course reprioritizing constantly as if it’s easy to resume later without paying ramp up.
Don’t get me started on the constant detailed status reports…
Yeah, it requires replacing the “you test the rocket” with “you test the rocket and it fails or doesn’t meet the updated mission specifications” and the “you go to mars” with “you want to go to mars”
A software engineer was not involved in this if waterfall is painted positively.
I think the last time I heard an engineer unironically advocating for a waterfall IRL was about a decade ago and they were the one of the crab-in-a-bucket, I-refuse-to-learn-anything-new types—with that being the very obvious motivation for their push-back.
I know several who preferred waterfall, but the system I work on is a giant government one and when we were doing waterfall we were in specialist teams working on a small part of the system
At the same time we went agile management also said “everyone can do everything” so we’ve had to work across the entire system
For the rocket analogy: we started building a rocket under waterfall, but when we went agile we also decided that the rocket motor specialists could also work on fuel tanks and heat shields
And here I am, running projects for the past 20 years mostly using agile, and still very much unconvinced about its supposed superiority over waterfall.
Waterfall: Spend 10 years compiling written functional and technical requirements. Cancel the program due to budget overrun.
Yeah, waterfall would be “you collect requirements to build a rocket to Mars, 2 years later you have a rocket to Venus and it turns out they didn’t think oxygen is essential, they’ll have to add that in the next major release.”
Of course because they don’t like being held to estimates and deadlines.
…and when you agree to run it Agile, which calls for closer and continual communications with the users, the first thing they want is a rep to do it for them .
Yes, silly engineers that don’t like being held to unrealistic estimates and deadlines; typically the ones that arise at the start of a project where there are still who-knows-how-many unknowns to find.
Waterfall is the most effective tool for software engineering in a world where the whole world stops once you’ve planned and only starts again once the project has finished—i.e. a fictional world that doesn’t exist. Literally every waterfall project I worked on back in the old days was derailed because something happened that wasn’t planned for—because planning for everything up front is impossible and planning for anything more than a handful of eventualities is impractical.
Agile and subsequent methodology comes from realising that requirements will change and that you are better off accepting that fact at the time than having to face it once you’re at the end of the current road.
Agile does not mean engineers talking continuously to the users, engineers are hired to do what they’re good at: engineering. Understanding user requirements and turning that into a plan has always been product’s job regardless of methodology, in agile and similar it’s just spread out over the duration of the project, not front loaded. Agile isn’t “make the engineers do every proficiency”.
They forgot the bit where the Waterfall method blew through the budget and deadline about five times over.
And it turns out the customer actually needed a blender
This is why I always act as if neither exists
Waterfall only works if the programmer knows what the client needs. Usually it goes like:
- Client has a need
- Client describes what they think they need to a salesperson
- Salesperson describes to the product manager what an amazing deal they just made
- Product manager panics and tries to quickly specify the product they think sales just sold
- Developers write the program they think product manager is describing
- The program doesn’t think. It just does whatever buggy mess the programmer just wrote
- The client is disappointed, because the program doesn’t solve their needs
- Eventually Company decides “agile will fix things”
- Developers are told to work agile but the only stakeholder they talk to is the PO, who talks to PM, who talks to Sales, who talks to Customers
- PM&Sales don’t want to deliver an unfinished/unpolished product so they give a review every sprint, by themselves, based on what they think the customer wants (they are Very Clever)
- A year or two later the project is delivered and the customer is predictably unhappy.
- Management says “how could this have happened!” and does it all over again.
as someone who has made it through multiple ‘agile transformations’ in large companies: that’s how it usually goes.
however, that is the problem with people being stuck in their way and people afraid of loosing their jobs. PO is usually filled with the previous teamlead (lower management, maybe in charge of 20 ppl). PM & Sales have to start delivering unfinished Products! how else are you going to get customer feedback while you can still cheaply change things? A lot of the middle management has to take something they would perceive as a ‘demotion’ or find new jobs entirely - who would have guessed that with an entirely new model you cannot map each piece 1:1…
Given these and many more problems i have seen many weird things: circles within circles within circles, many tiny waterfalls… some purists would call SAFE a perversion of agile.
the point is: if you want to go agile, you have to change (who would have thought that slapping a different sticker won’t do it?). the change has to start from the top. many companies try to do an ‘agile experiment’: the whole company is still doing what they do. however, one team does agile now - while still having to deliver in and for the old system…
I’ve seen so many companies force Agile without changing the management layer and style. Setting deadlines while demanding that teams work Agile. Insanity!
That’s what we use in my company: a series of mini waterfalls
In terms of Mars
- Client wants a robot to go to Mars
- Project is budgeted and sold to send a Mars Rover
- Work starts and after successful test the robot is shown to customer. Customer states he wants to send a Mechwarriors in a drop ship, not a little Pathfinder.
- Panic, change requests, money being discussed, rockets are being strapped together with duct tape and the rover is bolted on an old Asimo that is being rebuilt into the smallest Mechwarrior ever the day before launch
- Mech Asimo lands successfully, stumbles and falls on a rock after three steps
- Customer disappointed
Seems like the author has never programmed anything
I’m glad I’m not alone. I couldn’t make sense of this comic.
I’m getting pretty old so I have experienced multiple waterfall projects. The comic should be
You want to go to mars You spend 3 months designing a rocket You spend 6 months building a rocket You spend a month testing the rocket and notice there is a critical desing flaw.
You start over again with a new design and work on it for 2 months You spend another 6 months building it You spend 2 months testing
Rocket works fine now, but multiple other companies already have been to Mars, so no need to even go anymore.
This is the way
This is the perfect waterfall analogy.
pretty sure they’re saying waterfall for building a rocket because that’s literally how NASA builds a rocket, including the software. It’s terrible for building anything other than a rocket though, because the stakes aren’t high for most other projects, at least not in the way that a critical mistake will be incredibly bad.
i take you have never heard of the V-model. basically you climb the waterfall back up to verify everything. most things that fly within the atmosphere are done that way. pretty sure NASA would do the same.
You can assume people here know what waterfall and the V model are.
Depends. I’ve heard management talk about agile and waterfall, but I’ve not heard even one manager say V model.
You’re right I haven’t heard of that model, but NASA has documented pretty well that it follows waterfall. https://appel.nasa.gov/2018/11/27/spotlight-on-lessons-learned-aligning-system-development-models-with-insight-approaches/
Oh yes, everyone know that waterfall works and the rest sucks, nice
A good team can make any of these strategies work. A bad team will make a mockery out of them all. Most teams are neither good or bad, and stumble forward, or backwards, doing the motions
Waterfall is much more expensive than any of the agile methods, even with good requirement gathering and management
In some situations with some people yes. It’s really hard to separate the project and team.
Usually, projects I have seen start with the best plans and methods, or at least vague good intentions, but later pretend they never met them. Like a cheap date.
There are some projects that naturally lend themselves to one approach or other, and they last longer following the original guidelines ; but if a project lives long enough these guidelines become the enemy.
I think the only projects that follow any set of guidelines for longer than a few years; they have a narrow purpose for being. Straightforward evolution or needs
If the shoe fits …
What’s not covered is the 25 years of R&D in advance of waterfall project starting, or that it’s delivered 200% over time and cost due to those requirements being insufficient and based on assumptions that were never or are no longer true.
But at least the rocket has an opulent ball room as per the design spec
The Agile Development here is the same result I’ve experienced for every one of these methods. Mostly because of clients/management.
That’s why agile was created. Because people don’t know what they want in panel 1.
More accurately the waterfall mission ends up on Phobos only to have to scramble to figure out how to land on Titan because the customer can’t tell the difference between moons
Kiiiinda true but only with boomer-era on-disk printed at a factory Waterfall. Also everything after agile is just copium for an over professionalized world in which craftsmanship itself had given way, undermining the very concept of expertise so everything is junior devs and now ai
Seems biased… What’s that logo they’re trying to hide in the top-right?
Toggl’s old logo.
Must be OP trying to hide it, Toggl displayed it proudly. The author used to work for Toggl marketing and ask can be seen from this post, did an excellent job. He still has a webcomic, it’s just not marketing for Toggl anymore. Here it is
As for bias - it’s a time tracking tool, but I don’t think they actually shill for waterfall, I think it’s just poking fun at the agile methodologies.
