Word to the cloud. Okay, so we're recording to the car to the recording, I'll be held down no matter what happens. And I'm going to make you the host, which doesn't mean you have to do anything. That just means that when you ended it all ended for everybody. Okay? And then I have a way that you can see I'm up here twice. I've got it on my phone. So what'll happen is when I have to move, I'll switch to the phone, but I feel better being online here like this. Okay, so I expect a lot of people just going to come at the last minute because we actually did have more register this time than before. So it's been frenetic.
So Thorsten, Thorsten, that's like a Norwegian name. And I'm in Norway. And in connection.
No, just my parents. I'm from Germany. And my parents like Scandinavian names. So yes, they chose. Yeah, exactly like the summer for or champion. Yes. Okay, nice to be here today.
And where are you right now?
I'm waiting for my train in Frankfurt's Main Station. And so anytime it sounds, I will board and try to follow the webinar.
Okay, good luck. Otherwise, we're recorded. So. Yeah,
I know. I Yeah. I also could follow the session that you had on the fourth and actually, during my training, but I did my training class last two days, there were three instances where I could repeat an experience said youth that Elon shared on the fourth in his webinar, as Alan just talked about this and this pitch issue at the moment, so it's very practical for me. Okay.
We try to be practical. Yeah, I get I get very impatient with people who just say yeah, you should get more value. They don't tell you how. They'll tell you what it is. Now, by the way, your last name is it the same in German as in Norwegian? It means a mirror
No, I don't think so. I think this is from my very intense sisters and if need something like a Woodstock or school.
Well, the Norwegian it means a mirror. I can tell you that. All right. All right, mirror. Yeah, I will.
I will need to enter the train now. And I'll follow up just as I can. Thank you so much. See you later. Okay, well, we'll give it another minute or two people are starting to come in. Okay.
Yeah, yeah. Okay. We'll start in one more minute. I just realized something I have to get and we'll give people a minute to start or I'll be right back.
So anyway, hello, Susan. And hello, Tony. recognize you from last time. So first and just boarded a train in Frankfurt. Ah, but he'll be back with us. Right. And he's got a pure Norwegian name, but he's not Norwegian. His parents just like to Norse mythology or something like that. Or Stan. It's actually he said, Son of tour, but actually Stan means stone. I haven't told him yet. Okay, so the rock of tour, if you can hear me towards the will will sort you out. language wise, you should know what your name means. My mind is not glib.
Well, why don't we get started? We've got a few people are expecting a few more. We had a much higher registration today, but we'll see how we go. So you know. So I'm Al Shalloway, As you know, and Tom guild there. And I don't know that we need to do intros except Tom has been a big mentor for me for the last three years. And I've been around a lot, maybe for three or four years now. But he's sometimes been called the grandfather of Agile. So I'll say that because he influenced a lot of the people who were at the Agile Manifesto. If this is the first time you're either watching us or present, these are unscripted. But we have a theme and the theme today is questions to ask. And we're intentionally trying to keep them to be simple questions. These can be either to just kind of continuously He asked to get awareness or sometimes their questions that are just useful in general. So I thought we'd kick it off with Tom giving one of his I'll give one of mine. But I'm really hoping you guys have questions, not just about our questions, which are perfectly fine. But if you've had questions that you find are useful, that you tell people to ask in certain situations, or just to keep awareness, president of what's going on. So I'm going to turn it over to you, Tom, for you know, however long you want. And then I got a quick question, I will say this, I'm going to have to go in and out today, unfortunately, I've had something come up where I will be on the phone with LinkedIn. That's why I'm signed in twice, you may have noticed that that's my phone to backup when I have to get up. So I'll always be available. But I'll I'll be offline off picture. So Tom going forward.
Okay, my big question to everybody and everything all the time is very simple. Are you focused on the primary critical values of your stakeholders? And are you clearly focused, as opposed to vaguely focus being they focused is we need more security, clearly focused is we need to 95% detection of hackers within five seconds. There's no mistaking what that means. Okay, so focus on goals, people are focused on processes, people are focused on strategies, people are focused on just doing some work or writing some code. And they're not even indirectly focused on what should be their real goals. And I think that's why we fail and fail to impress people with work on our projects. That's my short question.
You're muted. Any thoughts about that? Or questions about that?
Well, a follow up question that comes to mind is how do you find out what those values are?
Okay, if I can give a short answer, and that anybody else. Now, one of the tricky parts here is I said, stakeholders, I didn't say users or customers notice that, okay. Now, some stakeholders are inanimate, and some are an obtainable, you know, you don't always wander to the CEO of the giant corporation, say, I need to interview you for my business analysis. So, okay. So sometimes you cannot simply asked, and you cannot even observe, but you might have to read a law on privacy, or something like that, right? You might have to read a speech held by the CEO, you cannot get to, and I've had to do that, for example, for Boeing to take an example. So, but the long and short of it is you have to find out what the stakeholder requires, what is critical for them, what is it they will buy into and bail out of if you don't do well enough? Okay. And so there are necessarily a variety of methods, what it starts with, who are my critical stakeholders who can shut me down, okay, who can badmouth me? And and, you know, you, you have 10 or 20, at least stakeholders, as opposed to the famous customer and user, which is too narrow. And you might have to discover some stakeholders as you go when they rear their ugly head and say, I'm going to shut you down make you illegal outcompete you. Okay? So it's an ongoing process of discovering stakeholders, and any one stakeholder can have multiple needs. And they can be the same need fate security as 10 other stakeholders but at a different level of a different kind. So it's complicated is my short answer.
Yeah, I wanted to throw something in there that Tom inferred, but make sure it's clear, inanimate, what's in an animate stakeholder. Could be other machines. Could be other systems could be the government.
Yep, yep. Yep. But I started off with some of the obvious like laws, regulations, policies, plans, contracts, believe me, they require things of your system. They give you requirements that somebody has to design towards and implement. And one of the reasons American Mehta's of the world keep on getting fined $800 million by the European Union is they haven't recognized that the European Union has laws that they have to respect or they will be shut down and fight.
I had a question that was going to be to ask in a certain situation, but something has been coming up for me as Tom and I've talked in the past. So I'm gonna ask this as a question to the group maybe to discuss, what's the difference between functional and non functional requirements? And is there really such a thing? It's double loaded here? Because I think that's a trick question. And is there really a difference between functional and non functional requirements?
I'm going to avoid answering that for the minute. But I'm
asking you all this question that Oh, yeah. Right. This? Yeah, not Tom. But yeah, everybody else? What's the difference? Is there a difference between functional and non functional requirements? And if so, why? What is it?
Well, one could give a, a cynical answer that non functional requirements are the ones you discover after you think you've got all the requirements.
Good. Well done. I
love that. Yeah. That's really good.
I'll try to go with functional requirements are often only valid or necessary for a part of the system or product and not for all where you could say non functional, non functional, are more generic Peric, they apply to most of the functionality of the feature like performance or something or maintainability. As right.
Yeah, that's good. I like that. I never thought that that's good.
No, that's good. That's a good observation. I sometimes push back and say, are you talking about dysfunctional requirements?
Like that, I
disliked the term intensely and never used it except in anger. And it's now in systems engineering. They're called performance requirements. Just to give you what somebody else arrived at, and the people who say nonfunctional, and I don't mean to you out who just said it, but the people say non functional, they don't know how to say, performance or quality. It's just all they know, is they're coder, they're a programmer. And it's not a functional requirement. Therefore, it must be non functional, but they don't even know how to describe it in the positive.
Yeah, if you were to if you were to say functional requirements are things that enable the customer to get something done, I would suggest having a 32nd response to every keystroke you apply makes it everything is non functional, pun intended. I'm actually don't believe there is a difference between functional and non functional requirements. They're like Thorson said maybe they have different characteristics. Maybe they're stated by different things. There is a performance requirement. And it's like Susan said, we sometimes we don't think about these issues upfront. Because it's hard to get them sated because the customer is not the customer assumes them unless they're things like security, and the product owner is talking to customers. So non functional requirements, or maybe assumed requirements that can't be specified well, that affect the whole system.
In my competitive engineering book, which you all have free digital copies of right. There's a glossary. And I define 700 terms very carefully. And functional. I'll simplify my definition, but it's very clear and black and white. It is what the system does or must do. What do can't simplify now, what are what some people refer to as non functional requirements, that's all the rest. Now, here's the tricky bit. It is not automatically the performance requirements because you can have constraints, like budgets, or legality, which are not functional requirements. But they're not performance requirements either. They're about I have a nine different categories of requirements that are not functions. Okay, but let's simplify because what people mean by non functional requirements is not all the other requirements that are not functions, they mean qualities. Usually, it's like security, reliability, etc. That's what they're referring to. And quality requirements is Describe how well the system will perform. It's a class of performance requirement. Things like speed and capacity, speak about how much the system will perform. So they're both subsets of performance. But one is, how much capacity be. And the other is how well, all the abilities in simple terms.
Actually, I'm gonna like I said, I might have to get offline for a couple minutes, I want to throw another question out, because this also might give you all other types of questions. So I'm really doing this to get you prompted. So these questions we just asked are more about things to things to, you know, think about? Kind of on an ongoing basis, or, but what about ones that happen at certain times? Have you found questions that under certain circumstances are useful? And here's an example. And I found this to be really helpful, as a practice by practicing means something you do without thinking automatically. It's got low cost and potential value, is I tell developers and product owners for that matter? Anybody talking? When you're asked to do something when somebody says, hey, I need this is you ask? And how will I know I've done that? That's it. End of story. I'm not talking about going up. Oh, how will they know I've done that? Because sometimes you'll get an answer right away. And it's not what you were thinking sometimes they'll get they won't know. But it's a useful question. How will I know I've done that? No cost 15 seconds. I've seen it save hours, just that simple question. Now, we can talk about the value of that. But are there other questions? Well, we could talk about the value of that for a minute and to throw out some things. But I really want you to think there are gonna be other questions of this type. And that's partly why I wanted this session was to think what they are. See if because if a product owner doesn't know what what it means to be done, then you know, they've not thought it through. They don't have a good requirement. And that's an issue. Now you saw that's another question. So other thoughts, other questions like that? Besides when, you know, in a situation, what would a good question be to ask?
Yeah, Susan, like just one thing about it's not you don't do it? unthinkingly you do it habitually? Yes.
In other words, you don't decide whether you're not you need to ask it. Because I've seen there are times when you know, you don't need to ask it. This is kind of in fact, the first place I use this was at an engineering company where they were building an engineering product, and the customers are engineers. So you'd figure if they asked us to do something we would know what to do. And I can remember, this customer, the engineer customer asked the engineer team, oh, I want this. And my thought was, oh, this is obvious. But ask anyway, that's what I was thinking I had my mouth zipped shut, you know, because I wanted to see if they would listen to the advice, and I figured they weren't going to because it was so obvious. But after about 20 seconds of other conversation, they said, and how will we know we've done this? And then the customer engineer gave us a completely different answer than what we were thinking of. Now. It saved a couple hours. I mean, it wasn't a huge savings. But 30 seconds to save two hours is not bad. Because what we would have done is built a screen showing an event said that's not what I meant, and then we would have built the right thing. Yeah, so I'm gonna leave I'm gonna be quiet now leave you guys and women to what questions have you seen useful like this? As Susan says a bit, actually, without thinking about it, just doing it. That's less why it's got to be low cost.
When I use a lot is Who else should I talk to about this?
Oh, that's a good one. Like and who needs to know about this really brilliant. I'm going to switch over to my phone, but I'm still here, but this will go away. The screen will go away.
Okay, I guess I sort of have a responsibility. Here are two things I put them in the chat. My first thing I put in but it accidentally only went to NBC. But I I put it in sort of again, the question of why is surprisingly powerful. Many of you will have heard of the Japanese five why's where you ask it repetitively. And what does that mean? That's that is seeking the core purpose, the real purpose, what people really care about, because too often, they're several levels down in the technology. And they forgotten what their corporation actually want it. And they're there. They're not going to deliver it either because they're too far away. Okay. So by asking why you get an answer, and that answer is one level up. And asking another wise a very good practice. Because if you get an answer, that's another level up, and it's more important higher priority, and worth knowing about worth stating, At some point, you get to a point where they say, I don't know, or, you know, because that's the way we do things around here. There's a point where it's obvious, you can stop, but then maybe you need to focus on that stopping point and get that straight. I'll leave that floating around, there's much more I can say about why and using an analysis of it is particularly important. It's all invariable that people are not very conscious or not conscious enough of the why they're down there. And so it's it, but they will answer the question. Or they will say, I don't know, my boss wanted to be we need to ask him something like that. Okay. Now, one more thing I put in the chat. And we need to recover it, make sure it's available. Everybody's a little booklet called 12. Tough questions. So without going to you can pick it up directly now from the chat. But the my answer is, I've got 12 tough questions for you. And they're all in writing. Other people?
So Tom, when you get Can you give an example. When you get to the I don't know, response to why? What example can you give and what of what you do and how that solves? The next problem?
Oh, right. Okay. It's okay, I have to construct an artificial chain. Let's just say that you're in a business and you get stuck at well to make a profit, or some business objective like that. Okay. Now, it turns out that may seem obvious to people that Well, that's it we're in business, this is capitalism. But it turns out if you were asking, for example, Elon Musk, he would, he would say, No, no, we need to make a profit. So we stay in business so we can get to Mars. Why? To save the human consciousness in case the planet, you know. So in other words, it's not at all obvious. The obvious thing that, you know, to make money is the real thing. And we've got to be careful about stopping prematurely at the wrong point. That sort of an answer, Karen, friend to be short, but if that's not good enough, you just get back at me and I'll try again.
All right, thank you.
I'm off screen, but I'm here.
Right? Was that owl? Talking?
Yeah, that was me. Sorry. Okay,
guys, we're a little bit unsure because I'm Torstein. Student. I'm
working hands free. So my camera's not working right. I am. Good bye. Okay, so what are the questions? I mean, I'm sure we can come to Tom and I can come up with them. But I think I'm sure you have something that's a useful question to ask habitually. I like that better than calling it a practice is, and that's a really nice way. Because practice gets misunderstood. What you mean?
Well, one, one I often ask is, what else happens along with this? Because a lot of understand their workflow, but you're you're now you don't
know good. You could use this it is good. And you could be implying many things. So we need to ask you, are you talking about side effects or parallel processes or something else? Actually,
all of the above because I find it that way. I look.
Good, all inclusive. Great. So Karen, hands up.
So I just thought of this. Why is what not working? Not working?
Yes. And people will have theories of that. And what you have to do is explore the hypotheses until something begins to work, I guess. That's all I can say at that level, Karen. But if it's not satisfactory, get back to us.
So thanks. They're gonna give a reason. That's the hypotheses. And then you have to try it out. Do you have a for example real life that you can give us?
Oh, I was just working with our dishwashing machine. And it just stopped with error message e oh one. And nobody knew what that meant. And that was not in the user manual. So we went searching, it said something about water in the pan below. And there was a user who said, I pulled it out and dumped the water and then things work. Now it turns out, my son yesterday, pulled out the washing machine dumped the water and now it's working. So that hype, but it wasn't easy. All we knew was its stopped and it said E oh one which was not very informative. new washing machine. Terrible. guy you want a real case? You got one that was yesterday?
dishes clean? Yes. Oh,
so great to have a dishwasher. Again, I hate washing up manual. Actually, my wife loves it. Because I dry she washes and we talk and that's like the old days.
Thank you.
We try. no perfect answers. But I think brief answers might be good. And then we'll refer to literature like the 12. Tough questions is good literature. Some of my friends say, Tom, that 12 tough questions is your finest moment. The best summary of all your ideas, so don't overlook it. Share it with a friend. Find out. Yeah, the tricky thing that I want to point out is, at one point, I was able to put the 12 questions on the back of my business card. And I would hand it out to managers. And they'd go to a meeting where they were bored, and people were trying to sell them some bullshit. And they would start asking those questions and they had so much fun. They would later meet me in the corridor and said, This is so much fun exposing these guys for what they are lying.
Anyway, you can look at the question. Yes, Tony. I don't know if this is helpful, but I'll just throw it out there. In my experience, like I asked the question, can you tell me more? Now that seems a bit vague. However, depending on the context, I will just keep trying to elicit more information. And in the language of the person I'm talking to because sometimes we don't speak the same language. Meanwhile, I'm collecting and collect them like the bee in the flower just collecting all the pollen. And then once I think I have enough pollen, I'll do whatever it is. That's a weird analogy, but I will take that and proceed accordingly. That's excellent. Tony. I like it. And you know, open ended questioning. And I take note. And I like the the analogy to be collecting colons. But, of course, there may be limits to it, but they'll probably say why the hell are you asking me I just told you is that some point? That's the point to stop, I think, at least recollect your wits and say, Man, why did they react that way? I was just trying to be polite.
I like that analogy a lot. It's kinda like the connector. You know, sometimes you can connect with tools, and boards, but sometimes it takes a person to be this pollinator of ideas. I think it's a brilliant metaphor. Agreed.
Let's see, Thorsten wrote something. In the context of improvement coaching, I really liked the approach of making it habitual to ask the questions of the Toyota Kata from Mike rather, pilot farts, you know, no Torsen. And, well, I've heard of Toyota kata Of course, it's useful if you can give a link which may be difficult on your train, but appreciate links, and I follow them up.
All right, I'll try to get get a link to you. But as you said, I'm on my smartphone here. So it's really difficult to have it. Yeah.
Well, if all else fails, email, Paul McGill. And
if you give it to me, then I can also put it on the page when this thing gets uploaded on the Tom Gill de resource page, as well as whatever Tom put in the chat. I'll put that there. Yeah, Toyota Kata is really brilliant. In fact, that makes me think about, you know, as a quick review, it's like you, you make a small the intention is to make a small change to see what happens, you know, there was an improvement or not. But it's not just running an experiment or inspected adapt, because it includes changing your underlying model as a result of the experiment. It's like, micro Plan, Do Study Act is another way to think about it. But a Kata means kind of hunting up actual action to learn. So that's why it isn't called Plan, Do Study Act, rightfully so it's just a small change to incrementally learn, kind of running under Kaizen in a sense and that way.
By the way, having said Plan, Do Study, Act, I can't resist. Can you see that the Diploma in the back of me? That is signed by W. Edwards Deming himself? 1983? Yeah, my buddy, I used to take him to the opera in London, and the ballet. Oh, wow. That is it is it. But I proudly hang his diploma there. So short, short story to give more information. At one point, I was looking at his plan, do study act cycle, which he then called Plan, Do Check Act, but he changed his mind went back anyway. And I said, you know, there's there's an obvious thing there like, where do you begin the circle has no obvious entry point. And I thought about it. And I drew my own conclusion. So I said, it doesn't matter where you start. If the cycle is rapid. It does matter if the cycle is long. And he said, That's right. deep knowledge about cycles and when to enter them. Go ahead, anybody. And you can put your questions in the chat, of course. I'm monitoring it, I would have difficulty in his car doing that.
Well, one thing I also like to ask when I'm talking to subject matter experts, is how often does this eventuality occur because subject matter experts are often focused on the exception. And that's what they deal with on a day to day basis. But that may not be 95% of what actually goes on.
excellent observation. I love it. Well done.
So So Tom, you said it doesn't matter if the cycle is short, but it doesn't matter if the cycle is long. So if the cycle is long, then what? Where does one start at the beginning?
Ah, okay. Now, I'm okay. In principle, when I, when I say in principle, in practice, there could be some reasons for doing other things, and I'll talk about them too. But in principle, I believe you need to start with stakeholders. I'm going to put a paper in here called Design logic, remind me if I forget it after I'll do it right away, though. It starts with actually in the design logic paper, we have to start with what I call the environment, define which environment you're talking about, then from that you can deduce who are the stakeholders in that environment? From that? It is what are their major concerns or values? Okay. Now, some of those values we cannot satisfy. They're too expensive. They're technically impossible. So their stakeholder values do not automatically turn into requirements. They only turn into requirements if it's cost effective for us to do something about it, which is very interesting point, because a lot of people well, the customer asked for it, therefore, it's a requirement is wrong. Okay, now, so there is a logical sequence. We also have something called the EVO cycle, which is a PDF cycle, but it's a little bit more detailed, and I will put a paper in on the EVO cycle, okay. But it's got things like stakeholders, values, designs, and then decomposition of designs and then implementation and feedback and learning. Okay, so having said that, I'm gonna get those two papers out and put them in the chat and let somebody else talk about what whatever they want.
Yeah, this girl, I want to give her another slightly different answer, that's a good answer to really answer what Karen was asking, but I'm gonna give a different answer. If the cycle is wrong, it means you're doing something if the cycle is long, it also means you're not doing something as well as you could. short cycles are very critical. And it's so when you when you have a short cycle, you have a lot of leeway for error, when you have a long cycle, there's a lot of danger, and a lot of risk. I want to say something else about learning cycles. Because this is I've been thinking about this a bit. Talked about single and double loop learning by Chris Archer, as single loop learning needs to get better at what you're doing. And double loop learning the you are looking at, you know, should you be doing what you're doing, or what are the assumptions underneath what you're doing what you're doing. And there's another way to think about this. That's more English, you know, without the jargony of single and double loop learning, we all know that the more feedback we can get the quicker feedback more than the more the quicker feedback we can get. While we're doing what we're doing, the better so we can stay on track with how we're working and what the what the success criteria of our stakeholders are. That's one loop that we want to make as fast as possible. But there is another learning blue that also goes to Toyota kata. So this is kind of a recurring theme here, that actually has to operate at a different pace. And that is checking your assumptions. So you, these are what are called heuristics, like we work on something, a particular way that we kind of assume how things are going to work. Like when you get into your car, when you open a door, there are certain things you do, they're just 99.9% of the time, you cannot be checking those every minute, you cannot be questioning yourself. So there's this second cycle of learning called checking your assumptions, questioning if you're doing it the right way. And they're the two things about that is one, you can't be doing it all the time. So that's a different cadence of checking. And the other one is you very often can't validate your own assumptions, because you don't see your assumptions, because they show up as truth to you. So anyway, I just wanted to inject that hoping we get another question from somebody else.
Bingo, I got my two papers up for everybody in the chat. And for those who don't aren't seeing the chat for any reason, we'll get them to you that the designed logic paper, it goes into depth about the question Susan asked, like, you know, what is the sequence? You know, if you don't have ultra rapid cycles, which is a special case, but we know where where should you logically start? I tried to answer that question in depth like I've never seen it answered before. Even that insight, I used to think you always started with the stakeholders. And that was a great assumption. But in fact, starting with defining the environment is precludes and helps you answer the question, Who are my stakeholders for this environment, therefore, it is logically necessary. Anyway, it's a short paper, and it's right there right now you can be able to get it. And the second paper is one I did last year, about the EVO cycle in some depth, which if you'd like is a practical extension of plan do study at its simplest thing I can say about it. Enjoy the papers, you can get them right now. Anybody that says a bad link, they're shut out.
What I found about the EVO cycle in this paper there is that it can be applied in the large like for your, your eternity of projects, which tend to continue on after initial deliveries, and they have a life cycle. And then and then it down to extremely short cycles. like Elon Musk has a one day 10 hour cycle for his mobs. And they try to accomplish something specific in one single day. A quantitative goal, by the way. So that's a pretty short cycle, actually, in Lean Startup Eric Ries his book, that you get situations where they're throwing changes into a real system 60 times a day. So that's a very short cycle. You know, if you'd like 1/60 of a day is the average cycle Frenchie. Each and but there's so in other words, we have ultra short cycles, and we have lifetime long cycles. Okay, and everything in between. But this one EVO cycle I call it can I feel serve all purposes? Have fun
open for questions and remarks. Don't be shy
Well, this is possibly more of a technical question, but it has come up in so many projects.
Susan, can you talk a little louder? It's hard to hear you.
Okay, I was gonna say this is a, it might be a more technical question. But it comes up so often in projects where there's a new platform for supposedly more or less the same stuff. And there's some capability in the new platform that you didn't have in the old platform. And when is it best to explore that? When you're in the middle of this? Is it when you're in the middle of this big effort? Or is it is it after you've replac formed and then you do the other stuff, because I've seen problems both ways.
I'm fanatic on doing minimum change when you're transitioning to new platforms. By the way, I've practiced this since 1960. In projects, you know, just get the damn thing over there and don't muck about with new stuff. And when the system is stable, I'm sure you're very well aware of this, you sound like you're when the system is stable, then you can experiment with new stuff, and not before. Otherwise, you risk everything.
Yeah, I take a slightly different view, but consistent with what you're thinking about. First of all, you're in this situation, when you have the system, you're when you have your application, know what system it's on. That's a design error right there. In other words, any code on in an application should either have functionality, or it should relate to the system you're on, it should never have both of these. This is a core design principle that we wrote about in our book design guidance explained. This also goes as far as objects should either have functionality or know which object is in play, but never both because that coupling causes problems. However, chances are from the way you ask the question, and the reality of the situation is that this coupling is present in systems. So I agree with Tom, you do not want to add new function while you're getting something converted. Because you want to get first the same functionality across in a new system. And then you want to make changes, the human mind works a lot better that way. And they're less complicating things. But I would add this as a possible thing, but it's context dependent. There are times when you can add something like a facade, not functionality, but something in the design of the code that isolates or separates the new functionality that's available. So then when you get to wanting to add it, it's easier, but you don't add it until you get the things stabilized first. So I don't think I'm inconsistent with Tom's comment. I think I'm just adding a wrinkle to it from a design approach.
Thank you.
And it looks like Karen has a question, but she's on mute. Yeah.
So Tom, you said you can only experiment after the system is stable. So what do you call what you do? Just stabilize the system? What is that? You know? What, let's try this.
Okay. There are two questions. One, what actions do you take to make a state system stable? The second was, how do you detect whether the system is in really stable? Okay, so the two part questions. The first thing I do for stability witness says we've already talked about it. Don't add complexity before you have proven stability. Okay, don't just you know, keep keep change to a minimum, whatever that means. And, you know, I've got some dramatic case studies of doing that. And I've done it again all my life. And it always works very well. And people who don't follow that always seem to get terrible problems. Now, next question, how do you know whether a system is stable? For example, it could be apparently stable. And but maybe it isn't. Maybe problems are hiding and cumulating. Okay. Now there's an obvious answer called, you know, testing of some kind. The other is just wait and give it time. You know, don't don't assume because everything's okay. On the first day, there'll be okay on the second day, and at some point, you may be willing to try and then the experiments you try should be safe experiments. That is, if you thought the system was stable, but it really isn't, then you need to be very sensitive to your experiment, and what side effects it has, like on performance or bugs or anything like that. And be prepared to instantly retract your change. If things aren't working. By the way, I described this process in my software metrics book, 1976, you do a change, and then you look for feedback. And if the feedback is negative, you have to retract it. That's my earliest definition of what is today called Agile. And that is published in 1976. In software metric, I have the book right here. Okay, so I've always been interested in this question of stability and how you deal with it. And 1976, I had 18 years of experience, I joined IBM in 58. So I probably knew what I was talking about after 18 years. Maybe, maybe I'm getting dementia today. And I don't know what I'm talking about. Be careful, maybe three. Don't Trust me. Trust but verify.
Thanks.
I would especially say the people who have not asked questions have first priority in asking a question. So Feel invited?
People are so so afraid of dumb questions.
Hi, Tom. So I would love you to talk a bit more about the requirements. You said something earlier that it is not a requirement because it is requested for. I know that most times, it's very business analysts, you collect requirements from stakeholders, and which is later converted to a scope statement. And then to filter those requirements. We have to look at the budget if we have the budgets available to carry out those requirements. So I would really like to talk more about those requirements, because I feel that when a stakeholder makes a request, then that should be a requirement.
Okay, well, thank you. I'd love to say a few words. Again, I'll try to be brief. The dangerous I could talk for an hour on that we certainly don't have that time. Now, let's imagine the following requirement. I want very high availability for my system. And in fact, we have a case study from AT and T the telephone system, where they had 99.90% availability at one point. They decided they could make a lot of money by completing more telephone calls. And they went for 99.98 0.08% For doesn't sound like a lot. But you know what? 3000 people used several years to deliver it. Now at&t could afford it with their high income. But if it were you, maybe you couldn't afford, you know, 3000 engineers working for five years, because there's a law as you move inequality and this is why the so called non functional requirements can be quite important. As you move inequality towards its ideal state 100% availability, you've also moved the costs to infinite time and infinite money, because that's what perfection costs in our universe. Okay, so there's a very high curve, you know, from some comfortable point that was easy to reach, even small increments will cost you more and more, until you get to what economists call the point of diminishing return on investments. Some of you heard of that, right? In other words, you can do they can want, they will want it, that's for sure. They'd love it, if you'd pay for it for them, if they don't have to wait five years to get it, right. So people cannot just demand what they want. Irrespective of the costs, there has to be some notion of they will pay for it, or the corporation will pay for it at the very minimum. And now you don't know what it will cost. For example, I talked to the man who did that AT and T project. And he said, You know what, Tom? Years later, I now know how to get to 99.998. And it ain't easy. Okay, so this is where people are at high quality corporations are working and have to work. So you know, one of the tricks is, before you can say yes to your client, you have to know what it's going to cost, obviously, and somebody has to put the bill, obviously, but you cannot know what it's going to cost until you've got a design to get there that will work. And some people we know are working in state of the art, space and military is always pushing state of the art. Okay, that means they're gonna work with designs that nobody knows what the cost is, because nobody has done it. Okay, so you can't just commit to the unknown and get infinite costs and infinite times. That's called project failure by deadline within budget. Okay, so now that's my short answer. And I'm pleased to go into more depth by email or in writing, or orally, when if when we have time, but we don't have to think about it. I will thank you very much.
Thank you for your question. I love I love answering that question.
After, after 10 seconds, those who have asked before should not be shy this because I invited those who hadn't asked before.
Okay, there's a whole category of what are often called transitional requirements, things that you need to get from the old system to the new system, but you don't need after you're on the new system. And those are often handled very badly, especially in in regards data. Any tips in that area?
Well, that that's a lovely, an interesting question. Could you help me out a little bit with an example of such a requirement? So I can feed off of that?
Well, you may not discover it till you're trying to transition. But sometimes the reputed data in the old system isn't terribly clean.
Right? Usually not. You're right. You think you know how to
transform it, and you find out as you're trying to transform it that you didn't, because there's this, this thing that, you know, it doesn't meet some minimal thing that you want to shove into the new database. One way of discovering that early appears to be to try to do a few mock transitions, but is there a is there a better approach than brute force?
Okay, I've got a feeling like this question has many simultaneous aspects. And there's a time problem here. So, you know, if I were consulting with you and answering it, we'd have a one hour discussion. And we've got eight minutes left, so I'll have a stab at it. I've written a paper, which I'll put up here about data quality, and how to measure data qualities. And there are I've got at least 10 different ways of measuring the quality of data like, is it up to date? Is it correct? Is it correlate with other things to give you some idea? So I believe we need to state requirements for the quality of data, like 10 different aspects of them? And the first question we could ask is, you know, is are the existing data in the database? Is this or is this up to spec here? Yes or No? Does it meet these requirements? Probably not. Okay. Then the next question is, if not, is it safe to transition at all? And if yes, how are we going to deal with problems in the data that we know we have, so that'd be the beginning of it. Now, what I'm going to do is I'm going to grab the paper, which I just held at a conference last Tuesday. And and I'm going to give you the paper so you can see how do we quantify and thereby get some control over the data. And then I'm going to shut off my microphone while I.
mean everybody should be quiet. Well, I got it. But by the way, there's a fun thing with this talk, which is that my son, this is a conference where they do not take videos of the 57 speeches held there. They're all sorts of TED Talk length, 15 minutes. And but my son was there, and he took a video, and just go straight to the video, because it's kind of funny. We got some good laughs. And my son was very critical of me and always has been said, That was a great talk, dad, then I know, I must have done something right for a change. So, but go straight to the video and have fun. And you can pick up the slides with the details. By the way, this is an AI question. You know, how would you measure the top 10 properties of data? Some of you can ask your favorite artificial intelligence, which I did. And you get good answers. My favorite person in the data area is a colleague of ours. Scott Ambler. Okay, you can find him on LinkedIn. But he has had the courage to write great depth books on data and, and data qualities and data management, like, you know, as Susan is hinting people are very good at algorithms, but they're actually very bad at the data part of it. And that's why we get so much problem with data. But I recommend Scott Ambler on the going into depth, he has books and papers. And he's on LinkedIn all the time. So I recommend him. He also has dad jokes if you'd like that.
Yes, I've been referring to data as a second class citizen for some time now.
With reason, unfortunately. But actually, the the conference I was just at is sort of a it's called a called Big Data conference has been going for about five years. But they they noticed what happened with musk and his self driving. You know, he was his self driving was totally focused on writing algorithms to understand what is a pedestrian and a child and movement bah, bah, bah, bah, bah. And it didn't work very well. And it couldn't handle all the edge cases. Suddenly, somebody came along and did what a lot of people knew about and said, No, no, this be driven by driver performance data, and have an extremely simple algorithm that just learns, suddenly, his driving algorithm works, you know, a hell of a lot better. So there's this transition to having far more respect for the data than the algorithm. That's the shift. We're in into artificial intelligence right now, as exemplified by Tesla's transition.
All right, nice. Thank you.
That's what my talk is trying to address in this context.
Quick question at 1253. There's an option for a full summary of This through the AI. I have not used that. Is it safe? Not safe to click on that through otter? And if not, what, what is the way that we can you know get a transcript for credible knowledge?
Oh, dear, sorry, I'm not sure what exactly what you're referring referring to the last two links I put in this a slides and
this is from this isn't 1253 A BZ auditors pilot to everyone? It has meaning. If you scroll down, it says seek full summary. And link. And I haven't done any of this yet. And I'm wondering, click don't click Find you know, somebody said zoom as a trans transcription and team so I'm just wondering what's the best and safest way if anyone has any input for that?
Tour Stein, can you answer he's offline. He's the man. There. Yeah, that's that
you're breaking up our daughter AI? That's
a no, that's, I use other AI to that's hotter. That's an otter AI with transcripts be useful for people? I can put them on the Tom Gill Resources page if they are. Oh, I would like that. Okay, I'll start doing that. I can make it that's otter. It easy. Just creating a transcript of the system. Audrey is fine.
You know, we hang on your every word Tom and owl.
Oh, okay. Time to wind down. Is that one last question you wanted to get in? This is your last chance today.
What about when's the next session? I'll be better prepared with questions. And
well, I forget what the next session is. But it is up on the site. It's two weeks from today. This one was done as an experiment. But I liked it. Actually, I learned a lot from this. So what we're going to do is have sessions every other Friday. And every other session will be some topic, you'd have to look on the website to see what it is. But it's it's out there. Now. I forget what we said we do. But it was selected. And then every other one will be these questions assuming people like these questions, so give me feedback if you want to continue with this open really open ended just asking questions, because I think there's value here I think we I think we step over the easy stuff and then complain how hard it is.
Okay, is it are we due to stop on the hour owl?
Yeah, we can. There's no other question, then what why don't we just stop, I'll put this up later today. And you can get the success and for the resources, Tom Gilbert resources, there should be actually just go to I'm sorry, success engineering that works on the front page. Take resources rather, and then take the manual, see what's there. But I'll I'll be posting that on the Tom guild Resources page later today as well. can always look there. So thanks for being here. And we'll see in a couple of weeks, some of you sooner. Thank you. Thank you. Thank you.
Nice to see you all.
Very much also. Next time, guys. Bye bye.
Okay, I'm the only one left almost Tim and NBC. Okay, a busy pilot and
