From Product Design to Process Design

[00:00:00] Roberto J. Arbulu: This is going to be our last presentation today, but I assure you it’s going to be a very powerful message from Martin, who has been a friend of the PPI team for many years, even before SPS and PPI was actually created. And so it’s a pleasure to introduce Martin particularly with me. Martin and I have had the opportunity to collaborate for several years, on certification efforts as used so early.

[00:00:29] Roberto J. Arbulu: We travel to many places around the world dealing with multiple professionals in the industry. And one thing I think I can say is that the it doesn’t matter where you go, we have the same problems. Okay? The engineering industry, construction industry is the same. So let me introduce Martin formally.

[00:00:53] Roberto J. Arbulu: I don’t want to miss Any important details of his body. And so Martin Fisher is a professor of civil and environmental engineering and by courtesy, computer science at Stanford university. He’s also the director of the center for integrated facility engineering, sci fi, a senior fellow of the pre court Institute for energy.

[00:01:19] Roberto J. Arbulu: And the coordinator of the building energy efficiency research at the Pre Core Energy Efficiency Center. There are probably many other things about Martin, but I’m just going to let him.

[00:01:30] Martin Fischer, PhD: I have the list there.

[00:01:31] Roberto J. Arbulu: Let him take us through a very interesting topic this afternoon. Thank you, Martin.

[00:01:37] Martin Fischer, PhD: Thanks, Roberto.

[00:01:39] Martin Fischer, PhD: Yeah, it reminded me of one of the experiences we had in Stockholm on one of our journeys That led to some variability I don’t know, you probably remember the first time we taught the VDC program off at somebody else’s place, not at Stanford. And we this was a general contractor in Stockholm and this was March.

[00:02:03] Martin Fischer, PhD: We arrived late in the evening and then we crashed from the flight and then the next morning said, they said, yeah, you can just walk. It’s just like a kilometer, a little more than a kilometer. And we looked outside the hotel. Things looked okay. There was snow, but it looked okay. We that takes about, maybe we should allow 20 minutes or something.

[00:02:23] Martin Fischer, PhD: So we start walking, and then we come up just 100 meters at most to the main road. And we turn left. And then we look at the sidewalk, and we should have had ice skates. I don’t know if you remember that. It was just like the whole kilometer was just blank ice. And we were a little bit late.

[00:02:46] Martin Fischer, PhD: But we arrived. Anyway now that has been really amazing to work with the PPI team. And so I would like to talk to you about product design and process, from product design to process design. Actually, I would say from product design In one silo and some kind of process design in the other silo to concurrent and integrated product and process design.

[00:03:11] Martin Fischer, PhD: And we saw already examples right from Phil at, of that in a sort of little bit different industry. But we saw the importance there of thinking about the products that they need to deliver and then the process of, how well we get the order to them and then they get things to us. But also just now, from Juan, we saw at a very detailed level of The connection, right?

[00:03:31] Martin Fischer, PhD: And how improvements really came from using, I guess I’ll jump here at least several of the five levers that I’m sure you have seen over the course of today adjusting the product design, adjusting process design, thinking through capacity, inventory, variability levers. So I yeah, the message we have is really what we see as the foundation for practice of designing, building, and actually also operating the built environment because we find ourselves, I think, at a very special place right now with a tremendous opportunity.

[00:04:11] Martin Fischer, PhD: And so I agree with Phil that yes, you need, don’t need just concepts, you need implementation, but we need to have concepts. And I think that’s where we need to get better in our industry it’s a bit different maybe in other industries. We now have the opportunity to bring concepts that have been proven, tested, applied in other sectors, in other industries to our industry.

[00:04:38] Martin Fischer, PhD: And I’ll briefly explain why. So that’s really the next practice. So we have yeah, apparently playing the video is slightly more challenging. I just wanted to remind us that we have all experienced the benefits of, over the last, say, two decades, actually, frankly, even slightly longer than that, but let’s just say two decades, of using product models and sometimes connected with the schedule, 3D models and 4D models.

[00:05:05] Martin Fischer, PhD: It’s unthinkable to create a design and, build something where you don’t have at least a good visualization of what, We are imagining and what we’re going to build bring the team on the same page, use information that’s connected in the 3D model to create the building materials and so on.

[00:05:23] Martin Fischer, PhD: I hope you don’t have projects where you don’t do these things anymore. So we have seen the value of product designs and I on purpose picked an old video, but then it became actually difficult to embed in the time I had. This is a, snapshot of the 4D model from the Bay Bridge, which I like.

[00:05:44] Martin Fischer, PhD: So this was done 13, 14 years ago, and it shows I mean you can see the snapshot already it shows the construction of the bridge. It shows the fabrication of the bridge segments in China. And it also shows when they’re being shipped, et cetera, different stages. So in one picture, you had really a lot of data aggregated.

[00:06:07] Martin Fischer, PhD: people to, understand. That was just a, quick reminder of the value of 3D models that is really unthinkable for us to do projects. More recently a supermarket chain in Sweden reduced their construction costs by 45 percent by getting out of the myth of every project is unique.

[00:06:32] Martin Fischer, PhD: And standardizing the 90 percent that is actually not unique. And then allow the project team to focus on the things that are truly unique. I think that’s an opportunity that I still see many organizations can take. Just as a reminder of the value of these you can call them product models, that was actually the original word before terms like building information model and so came up.

[00:07:00] Martin Fischer, PhD: But the thing that has changed in the last few years is this story here.

[00:07:12] Martin Fischer, PhD: So I think today this is, would be weird if you said this to somebody. That was not very data driven of you, but I think it’s not so far in the I think at Amazon, I don’t think that, I would imagine that you would say that, right? Yeah. But in our industry, yeah, it’s not, I haven’t heard it yet.

[00:07:40] Martin Fischer, PhD: But we are definitely getting there because we can, the ability to collect data on how our work is going has just exploded in the last few years. And I don’t see that slowing down and companies are connecting tremendous amounts of data, which is fantastic, but because we’re lacking or we’re not using key concepts process design and process modeling, production modeling.

[00:08:08] Martin Fischer, PhD: We cannot actually really, you make great use of this data. So I want to illustrate that and basically the message is yes, use these technologies, take advantage of the data you can collect, but in parallel upscale the conceptual thinking that you have around production processes and how to connect that with, product design.

[00:08:38] Martin Fischer, PhD: So I’ll focus so I’ll give a couple examples to hopefully bring to life what I’m talking about. First example is using the IoT crane hook from Versatile. So you might use it on your projects. Where basically the crane hook collects the data, everything the crane is doing. So now you get a second by second story for the whole duration of the project, what the crane was doing.

[00:09:06] Martin Fischer, PhD: Idle being rigged, loading this kind of, lifting this kind of load, etc. So that’s quite a lot of data that you’re getting. And so this is the first time we were able to work with a company. It was a few years back where we could see what can you get? What value can you get from this detailed data?

[00:09:25] Martin Fischer, PhD: So this is Clark Pacific for the construction of a parking garage near San Jose airport. I think 1200 car garage, something like that. And of course for a precast project, the crane is pretty important in terms of it going well or not. Up to that point, they only had average time. So I didn’t have time to have somebody there standing around and log all kinds of data.

[00:09:50] Martin Fischer, PhD: They just look at the end of the day, I’ll be put 45 pieces of precast in place. We worked 7. 8 hours. Okay. So many pieces per hour or so many, so much time per piece. That’s the data they had. Like I think pretty much everybody that we have found in industry. Okay. But now, they had this technology that gave them data about every step, every cycle, every component and be then curious, can you create a better life?

[00:10:18] Martin Fischer, PhD: Sheets and sheets of data. The the situation we wanted to help improve, or see if we could improve the data, is can you make a better prediction for tomorrow? Because if you’re the manager on that project, you live and die by how well, economically. Otherwise, how well your prediction is for how many pieces you can install tomorrow. If you order too few, because the yard is in Sacramento, the site is here in San Jose, so staff, they have to call the afternoon before, hey, send these pieces, these beams, these double Ts, these columns, and so on.

[00:11:01] Martin Fischer, PhD: And because you have to install them straight from the truck. You don’t have budget or space to unload them, find them again, and put them back up. So the inventory is at the yard, and then you have to just install them. If you are under obviously you over, you pay for the capacity anyway.

[00:11:18] Martin Fischer, PhD: If you don’t order enough pieces, if you order too many pieces, then you pay overtime. On average Clock Pacific is a very good company. They were two hours off per day. So you can see it adds up over the duration of a project. So the question was, could we by using the detailed data, help them make a better prediction.

[00:11:41] Martin Fischer, PhD: So here’s basically what they were using aggregate activity durations. So summary activity durations. And so basically we, studied the difference in how many pieces you would order if you used. The aggregate data, it takes 24 and a half minutes per piece or whatever the number might be versus per type of part.

[00:12:04] Martin Fischer, PhD: Right now we have the data. We can say beams take this long and this is the variability. Double Ts take this long and so on. Or if you go to every step of the process for every component. So for that we had to actually develop a a tool, a next day look ahead planning tool that was fully automated because.

[00:12:25] Martin Fischer, PhD: You can’t possibly, that afternoon, sift through all of the data and make sense of it manually. And then I know this looks a little crazy at first, but then we made the comparison. In gray let’s see, maybe I’ll have to do it like this. In gray yeah, it doesn’t show up, okay.

[00:12:49] Martin Fischer, PhD: Alright, the black line is, if you ordered exactly enough parts for the day, then the dot. Which represents basically if it’s above the line, it means you had to work overtime. If it’s under the line, you waste the capacity. So basically from left to I guess I should have started there. It’s the days and then each dot represents how many pieces grade on how many pieces they ordered.

[00:13:15] Martin Fischer, PhD: And you can see that on some days they were quite good on many days. They were quite far from the perfect order. And then you saw the color scheme here. Yeah, no, it didn’t translate. I’m sorry. Oops. That was the wrong button. Obviously. Okay. Yeah, sorry. Anyway the colored lines are basically leveraging the detailed data.

[00:13:43] Martin Fischer, PhD: And you can see that pretty much most of the days, the detailed data allows you to get closer to the perfect order. Not perfect. But in summary, we could show that the detailed data would allow us to reduce the forecast error by half. And we gave somebody two hours each day to do something else.

[00:14:05] Martin Fischer, PhD: Because we had to automate the lookup planning. So this is an illustration of the value, one example of the value you can get when you actually leverage the detailed data. But I think the actual value is much greater than of course this is, I think, pretty nice to have. But it also shows that we do need to develop some new tools to actually use all of the data.

[00:14:31] Martin Fischer, PhD: That we are collecting. Yeah, we are currently, we continue this kind of line of collaboration with industry. We’re currently working with HDCC on the construction of these kinds of towers in Honolulu where we use computer vision methods to find how long different activities take so we can then connect them.

[00:14:55] Martin Fischer, PhD: Into a production model. And this is the key point I want to make, right? Same thing at the four o’clock Pacific. The schedule was a kind of a summary level and now you have extremely detailed data. The two no longer connect before the average data was fine because that was the level of detail you had in your schedule.

[00:15:18] Martin Fischer, PhD: But now you have a lot more detail of what actually goes on and you have only some summary schedule. So if you want to make improvement because you now have the possibility for this feedback loop. You’re back to stories and anecdotes. And there’s the same here, where the schedule is, rebar in zone A.

[00:15:39] Martin Fischer, PhD: Yeah, but the story, if you want to improve it it’s, You have to look at a lot more detail, but now you can. If you’re successful in this case with the computer vision, but we also have data from other sensors. So for that, we need to develop and sorry, this, that this picture is so faint. I don’t know what happened.

[00:15:58] Martin Fischer, PhD: But we, basically using the PSO tool from PPI and SPS that allows us to really model how the process happens and then connect the detailed data to it. So that we end up with a process that model that we can look at and that we can compute. And this is really where we have to go because we’ve seen the benefit of having a product model, a building model, or whatever you want to call it that we can look at in 3d or 40 and the data connected to it.

[00:16:38] Martin Fischer, PhD: But we don’t have the same thing for our production processes, at least not in widespread use. And you might’ve said, yeah, up to recently, I’m all thing. That’s a nice theory. How am I going to actually implement that? But today, you can implement it. Actually you have to implement it.

[00:17:00] Martin Fischer, PhD: Because if you don’t implement this very, much more detailed production process models for your engineering and management work, and for the work offsite and onsite, you will not be able to take advantage of the detailed data you can collect. And I would not want to have to compete with somebody that is able to do that.

[00:17:25] Martin Fischer, PhD: It’s just maybe what I would say. We also the same applies for pre car off site factories. We’ve been able to do a study with a bathroom manufacturer thanks to Bochen. You can ask Bo Chen questions about it. But basically this is what they this was the plan for the project.

[00:17:50] Martin Fischer, PhD: They wanted to produce all the bathroom parts before they delivered the first to the site. We think it’s a little bit crazy, but they just didn’t want to get a call from the site saying, Hey! I need that bath, that pot and don’t not have it. So in in reality so then this basically they built up the inventory and then they built it down.

[00:18:11] Martin Fischer, PhD: The reality actually looked a little bit different, but the conceptual idea was what was shown on the first one, the previous slide. But in reality, this is how the demand from the site actually happened. So quite a bit of variability. But still you can imagine the inventory cost was really very significant in this case.

[00:18:31] Martin Fischer, PhD: And through, again, by building a production simulation model that you could validate with the actual data, we could then make, I think, a believable prediction that you could have reduced the inventory cost by 80 percent and still met the delivery requirements from the site. Not bad, as we say in my native Switzerland and but the yeah, because yes, it can be challenging to connect the off site and on site, although I think again with the transparency, the kind of transparency, visibility that Phil was talking about we can do much better, and we definitely, you could see the variability that actually happened in, in that situation on that one project.

[00:19:23] Martin Fischer, PhD: But we can be much smaller about how we actually handle it, manage it, how we can reduce it, and then how we can actually manage it, by getting a bit out of our silos. Basically, the key message we need to create a digital twin, or whatever your favorite term is, digital replica or something, of our production processes.

[00:19:46] Martin Fischer, PhD: And by that I mean something we can look at together, because we have to be on the same page as people. from different, with different functions and roles. It’s super important. And we have to be able to run simulations, compute things, connect with real data, etc. So I think that’s the task ahead for us.

[00:20:07] Martin Fischer, PhD: That’s what we need to do. If we do we will be able to leverage all five levers. If we don’t, then we will, the levers will be in the silos and the effect will be dramatically less than what it could actually be. Thank you.

[00:20:28] Roberto J. Arbulu: Any questions for Martin? Anybody? everybody. Please raise your hand so I can approach where you are.

[00:20:39] Martin Fischer, PhD: Was that just preaching to the choir and everybody going, saying yeah, we already got that. Or or is anybody going to take action? Amazon? No, actually, I don’t know where I got it. I it showed up in my Facebook feed and then I’m actually not sure who actually delivered. I’m sorry. I would have to look at my email.

[00:21:03] Martin Fischer, PhD: Yeah,

[00:21:05] Audience Member (2): Thanks again for the presentation, it’s always entertaining. In the crane example using the IoT sensor, how did you take into account the variability that the IoT sensor wasn’t measuring, such as, I don’t know whether it did wind speed, you can’t lift at a certain wind speed SIG operator, equipment breakdown, those kind of elements that may not have been associated with the IoT sensor, and factored that into the variability?

[00:21:25] Martin Fischer, PhD: Yeah. So that was So we, used actually the data from that project in my class that my graduate class called Managing Fabrication and Construction today as a possible class project. And we had quite a few students yeah, analyzing the data in all kinds of ways, which was cool. And they had just a presentation this morning.

[00:21:45] Martin Fischer, PhD: That’s why I had to be there and not here. And and that was exactly like, so they asked actually the people from versatile and they supported us super helpful. And they actually asked exactly this question, they said, yeah what could we do? What else would make it easier? And then the first answer from the students was yes, we can pull.

[00:22:09] Martin Fischer, PhD: So what we had to do is we had to create a data mashup because we had to get rather data for the location, superimpose it, connect it, and start to make sense of it would be super helpful if you already could collect some of this data like wind speed, precipitation, humidity, things like that, because yeah, and then there was generally the insight that we keep running into as much data as we can now collect.

[00:22:37] Martin Fischer, PhD: It’s still not enough. So we gain tremendously interesting insights. For example, they found that, they looked at the cycle times in the week before holidays, and they found that the let’s see what did they find, yeah, they found actually the Monday before, so the five days before a holiday, they were quite faster than typical, and then the four days before the holiday were actually quite a bit slower.

[00:23:13] Martin Fischer, PhD: So and now, but now you have data, right? Because you might have, you might believe all kinds of might, everybody might have their own story back to this point, right? But now you have data on some of the things. But then you wonder need to connect with the schedule. What were we actually trying to do?

[00:23:31] Martin Fischer, PhD: Were the load types different? Because we were approaching a holiday, maybe things, so you need to connect a lot of data. That’s what we keep seeing. There’s definitely, I think, an opportunity there for, Somebody in the consulting and startup world, tech world.

[00:23:50] Roberto J. Arbulu: So I do have a question for you, Martin, but before I do that, I’m gonna ask any other question coming from the audience.

[00:23:57] Roberto J. Arbulu: Anyone?

[00:24:01] Roberto J. Arbulu: One, two, three. So Martin, I think what you just told us is that if we don’t have in our organizations a production system perspective and we don’t focus on the production system do we have a choice not doing it?

[00:24:18] Martin Fischer, PhD: You always have a choice but if you want to attract great people and be competitive in a few years, I don’t think you have a choice. Would be my take.

[00:24:31] Roberto J. Arbulu: I think we, we see the same thing because with all these data and the production system and the data is actually from the production system is not from this, from the schedule. Yeah, so we need to move from the schedule to the production system.

[00:24:46] Martin Fischer, PhD: And to your point without, what the students also saw is if you don’t have a production system, if you don’t have that map, then you don’t actually know what data you really must have and what data you must really connect.

[00:25:06] Martin Fischer, PhD: And then you’re back in the world of just in case, let’s get all of this, then you’re back in the world of drawings. Let’s just make a whole bunch of drawings because somebody might need them. Or in case there’s an issue, we can say no, it was on the drawing, right? But it’s not productive.

[00:25:24] Audience Member: The everyone’s special. So my project is completely different to every other project that’s ever happened before and nothing’s the same, right? That’s not true, but how the examples there were quite modular type items or, multiple floors of a building for example.

[00:25:51] Audience Member: How do you take that same principle and apply it to these more, less repeatable type projects?

[00:26:08] Martin Fischer, PhD: Yeah. I think the, so of course we should think about the product and see what can we, I don’t, I really don’t care that much for that word standardized. But for lack of a immediately better word that comes to mind modularize, systematize. Systematize might be better but I think there is actually more opportunity to systematize our organization and processes.

[00:26:37] Martin Fischer, PhD: And, so I should have maybe told the story that I briefly showed just to highlight the value of Of a digital model connected with information, a 3D model in the, supermarket chain in, Sweden. Of course, they did more than that. And they really first standardized as much as they could the decision making processes for each of the different stores.

[00:27:08] Martin Fischer, PhD: Because it was exactly the same story, right? How do you want to compare a store in downtown Stockholm with one at the outskirts in the north of Sweden in the little town of Lulea? Like it’s a totally different story, right? But the reality is actually about 90 percent is the same. From a decision making process and and, an information flow perspective.

[00:27:32] Martin Fischer, PhD: So they organized, they standardized, systematized that first and then through that they were actually then able to recognize. What can they model rise in terms of the, physical product? And I think that’s still an opportunity we have, even if we build extremely unique things, because you’re not reinventing everything from scratch every single time.

[00:27:54] Martin Fischer, PhD: Occasionally it happens, I understand, but No, we’re not. Yeah, generally. Yes, you are, but you don’t have to, I would say. I think it’s rare that you would have to.

[00:28:06] Audience Member: We like being special.

[00:28:08] Martin Fischer, PhD: Yeah. No, I, I know. I, know. Yep. Okay.

[00:28:16] Todd R. Zabelle: Just to add on to what Martin’s saying, I think one of the secrets that we use that I’ll put out there is, for, something to think about, Bruce, is to get to the lowest common denominator.

[00:28:26] Todd R. Zabelle: Pooling, welding, fixing, rigging, and so the lowest common denominator is the lowest common denominator, and I think that’s where you were headed, Martin, right? Yeah.

[00:28:37] Martin Fischer, PhD: And I’m feeling sarcastic. Then I say then if everything is unique and you make everything new from scratch and you, don’t, it doesn’t matter.

[00:28:44] Martin Fischer, PhD: You could hire anybody. You don’t need experience.

[00:28:47] Todd R. Zabelle: And you wouldn’t have trades and crafts.

[00:28:50] Martin Fischer, PhD: Which I understand is an extremely sorry, sarcastic comment and not true, but obviously there’s something there that is common, right?

[00:29:03] Roberto J. Arbulu: Any other question, comment? You have one. It’s a simple one because the concept of a production system typically might get associated with something that is repeated multiple times. Something that produces a lot of things off. But, which is right, but what if we have a one off situation? So we have applied these same structures and frameworks when we have a one off situation.

[00:29:35] Roberto J. Arbulu: And it’s even much more powerful. Because we have only one chance to get it right.

[00:29:44] Roberto J. Arbulu: And we still have a production system. If we have one chance to get it right, what are the odds that we can get it right? And so it’s not just related to repetition. Although that’s probably where we’re going to see a lot of value. Okay, so we have my production systems that are both types.

[00:30:06] Roberto J. Arbulu: If that makes sense. Okay, so thank you, Martin. Appreciate it.