How Schedule Risk Analysis Optimizes Your Path Of Construction

Transcript

This transcript breaks down how superintendents and foremen are using Interval Planning to plan their daily crew allocation in line with the overarching project CPM schedule. You’ll learn how this new approach massively drives execution confidence and achievability of the original CPM schedule.

John Klobucar:

Hello, I’m John Klobucar with InEight, and I’d like to welcome you to the latest webcast and our “Best Path of Construction” series. Today’s webcast is titled “How Schedule Risk Analysis Optimizes Your Path of Construction.” Our presenter today is Dr. Dan Patterson, who is InEight’s Chief Design Officer. In this role, Dan focuses on expanding upon his vision of creating next generation planning and scheduling software solutions for the construction industry. Dan is a certified project management professional by the Project Management Institute. Now, if you have any questions as you watch this webcast, please email them to webcasts@ineight.com and Dan will do his best to answer them. Also, this presentation is being recorded and we will be sending you a link to the video in about a week’s time. Once again, we’re glad you’ve joined us and now let me introduce Dr. Dan Patterson.

Dr. Dan Patterson:

Well, thank you John and welcome everybody. So today’s webinar is a really interesting topic. In previous sessions we focused very heavily on how AWP, or advanced work packaging, can assist in forecasting both project duration and also capex cost. Today’s topic is adding a layer of certainty onto that forecast in the form of conducting a risk analysis to drive our confidence. In other words, okay, yes, we have a scheduled forecast and yes we have a capex budget, but how certain are we that we can achieve that forecasted schedule and full cost estimate? So what I’d like to do is share with you what I believe is really sort of the next generation planning process where we’re actually incorporating risk and uncertainty, the potentially bad things that can happen, and let’s actually incorporate those in our plan up front so that we’re actually generating a more achievable, realistic, full cost.

Dr. Dan Patterson:

So looking at traditional scheduling, I think it’s fair to say that traditional CPM scheduling, the focus is very much we plan from left to right. We start with a project start date, we then map out the subsequent activities, whether it’s engineering, procurement, construction scope, and from there we generate what we call a CPM schedule. And again, traditional planning has really driven us to focus on what we call a critical path. That is activities that don’t carry any float such that if they are a day late, they will have a knock-on effect of literally a day-for-day impact on project completion. Now, that’s all very well, but let’s take a moment to consider the noncritical activities in that CPM schedule.

Dr. Dan Patterson:

Water, for example, is noncritical activities that carry float, what of those noncritical activities actually consume the float as a result of scheduled delay, perhaps productivity isn’t what we thought it was going to be, perhaps there was an external event that prohibited our ability to execute that work on time. What happens is that float gets consumed to the point where that noncritical activity actually becomes critical and pushes our project completion date. So this concept of just focusing on the critical path I think has perhaps, it’s hidden some of the areas that we should have focused on during execution.

Dr. Dan Patterson:

I think also another shortcoming of traditional scheduling is the fact that because we plan from left to right, there is a natural compounding knock-on effect. So for example, again, take our engineering, procurement, construction, commissioning type project, construction and commissioning scope may indeed be very, very achievable and in fact may carry a low degree of risk. But if the proceeding engineering and procurement scope carries risk and that risk comes to being, then the construction scope will of course get delayed through knock-on effect. So the net result of that is by planning left or right and starting with a fixed start date. Project completion, our ability to accurately forecast and achieve an on-time project completion, our ability to achieve that, is challenged at best.

Dr. Dan Patterson:

And so I believe that traditional CPM scheduling really generates what we call a best case scenario, whereas in reality, shouldn’t we be planning and marching towards at least a most likely case scenario. So that’s really where I think AWP, advanced work packaging, actually massively benefits in terms of building that project schedule. So let’s start planning from right to left. Let’s agree on a project completion date and work backwards and again, I think one of the massive drivers of AWP is this constant thinking about constraint-free execution. Let’s plan everything and everything. Let’s plan such that execution is absolutely constraint-free.

Dr. Dan Patterson:

What I think that means is that we need to give more focus to aligning people, materials and space when planning those construction activities. More importantly, and this gets us back to the subject of risk analysis, it’s okay to account for risk and uncertainty when building the plan. Again, traditional planning, we build a CPM schedule and then perhaps we might go ahead and do a risk analysis. Well, that’s not right. We should account for risk and uncertainty as we’re actually building the plan itself. So plan to the most likely.

Dr. Dan Patterson:

Now relating that back to AWP, and I know we’ve touched on AWP theory in prior webinars, but you know essentially what we’re doing is we’re marrying up top-down front-end planning with field execution work planning. If we add risk or risk analysis to the mix, what we’re really doing is when we develop our top down level one, two, three, where we’re breaking the work scope down to construction work packages, what we’re doing and ending up with is what we call a risk-adjusted forecast. In other words, the durations for the planned activities actually have contingency built into them, and more importantly, and most importantly, the contingency is relevant to the degree of risk that is embedded within those construction activities. Now what that means is when we get to field execution and we’re actually executing the work and we’re actually out in the field is we’re essentially burning down on the contingency that’s been embedded in the forecast. And even if we do burn down that contingency, we will still achieve an on time completion because the contingency is acting as a buffer.

Dr. Dan Patterson:

So let’s spend a couple of minutes just talking through some of the modeling techniques. So as with any system, a risk analysis has outputs, has an analysis, and then obviously it has inputs. So let’s work backwards in that flow. Let’s start with the outputs and the insight. I think first and foremost, the biggest benefit that a schedule risk analysis brings to the AWP table is we end up with a risk-adjusted forecast, and that risk-adjusted forecast has an associated degree of certainty against it. And this concept of certainty is so important. Would I rather know that my project completion duration is 24 months instead of 18 months if that 24 months is presented in context of perhaps an 80% certainty? Absolutely, because my 18 month aggressive forecast, if I don’t know how certain that forecast is in terms of achievability, then I could come up with any number, it really doesn’t matter.

Dr. Dan Patterson:

So coming up with a risk-adjusted forecast, even if that forecast is longer than the deterministic, I think is incredibly valuable and it gives me a much stronger probability of achieving that duration. Now we’re going to talk a lot about contingency and I think it’s very important to understand that from a theory perspective, contingency is the difference between the risk-adjusted forecast and the non-risk adjusted forecast. But even more importantly, we can only report contingency in context of a given confidence level. So to give you an example, I am 80% certain that my project will take no longer than 24 months; that 24 months has embedded within it six months-worth of contingency.

Dr. Dan Patterson:

My contingency will vary depending on the confidence level against which I want to look at my project. I think another massive benefit of conducting a schedule risk analysis is the fact that the model doesn’t just come back with a single outcome. In fact, it comes back with a whole range of outcomes. And that range we typically call our P zero to P100 range. And I’ll talk a little bit more about that in a few minutes. But again, the value there is we can look up the degree of uncertainty of the degree of exposure again knowing that my project is perhaps going to be six months late, does it tell me much? I don’t know. I think I’d rather know that my project is going to be six months late and my range is so many weeks or so many months. So understanding the degree of wiggle or range is just as important as the confidence level.

Dr. Dan Patterson:

And then finally drive us. And again, I think this is so important with regards to next generation planning or risk analysis. Our exposure, our risk exposure is it simply because we have a very aggressive schedule or does the team believe our schedule is achievable, but the field execution, the work, the construction is at risk from discrete external events such as weather events labor disputes, so on and so forth? So differentiating between the two drivers of risk, exposure, uncertainty and risk events is very, very important. And the diagram on the right hand side actually I think is a good example of that. So the model at the top on deterministic schedule, we have at two parallel paths in our CPM network. We have task A and task B that in absence of risk and uncertainty are on the critical path. So the six days and four days respectively gives a total of 10 days duration. That 10 days is longer than the corresponding parallel path where we have task C with eight days.

Dr. Dan Patterson:

The eight day task of course then is carrying two days-worth of float. Now that’s interesting but in reality maybe that task C is actually on the critical pop and that’s what the risk adjusted example in the bottom right shows because in the scenario where task A task B have a combination of uncertainty and risk events, yes, the total duration increases from 10 to 17 days. However, because task C has a very high degree of uncertainty and it has large impacting risk events, the net path in task C there is actually 20 days. So what that means is the critical path has jumped from task A, task B down to task C. So not only has the duration increased, the critical path has jumped as well.

Dr. Dan Patterson:

Now how we get to that answer is through the analysis. And again, traditional risk analysis relied on a technique known as Monte Carlo simulation, where the computer would somewhat randomly sample each of those individual outcomes from the uncertainties and the risk events, the idea then being the more iterations within the simulation that the computer runs the more the results converge to a given answer. Now I believe that was certainly a very useful, meaningful step in the history of risk analysis. But what we’ve realized recently at InEight is this concept of applying randomness perhaps isn’t the best technique. And so today we’re actually using a modified Monte Carlo simulation where we’ve actually eliminated the concept of applying randomness to those samples. And we’re actually using a much more pinpointed algorithmic approach. And again, more on that in a few minutes.

Dr. Dan Patterson:

So at the end of the day, once you’ve run your risk analysis, either using traditional Monte Carlo or modified Monte-Carlo, obviously interpreting the results is paramount. And the two reports on the screen here are really the two most important risk reports. So on the left hand side we have, I call this the “What” chart, what is my risk exposure? What is my risk-adjusted view of the project? And I determine that by using the right hand side scale where we have what we call a best case through worst case or P zero through P100 and everything in between. And again, the most effective way of using this report is to report against the given confidence level. So the example on the screen here, if I would to be 75% certain I would read off the corresponding project completion date against that given confidence level. The difference between my P75 and my non-risk adjusted of course is contingency.

Dr. Dan Patterson:

Now understanding what my risk exposure is, is useful, but equally valuable is to understand why, and that’s the second report. The drivers report on the right hand side. Again, this used to be called for example, a “Tornado” chart. Now what we’ve done is we’ve taken the concept of a simple Tornado chart and actually evolved it now such that we segment that chart such that the chart will actually report… Okay, so in our example here, installation is the biggest risk driver, but is it because of schedule uncertainty and aggressiveness? Or is it because of a discrete event such as the installation event here? Dropped object event? So understanding the drivers and more importantly differentiating between all those drivers risk events or indeed is it simply schedule aggressiveness?

Dr. Dan Patterson:

So we’re really taking a three-pronged approach at InEight to schedule risk analysis and applying that to improving the forecasting ability within the context of AWP. So that three-pronged approach, firstly comprises augmented, or artificial, intelligence. So again, really the concept here is rather than asking the project team and the project stakeholders to arbitrarily define risk events and try and quantify those uncertainty factors, what we’re doing here is we’re actually leveraging past project performance. So as the planner or scheduler is developing the CPM schedule, the computer will actually make suggestions as to risk events that happened on prior analogist projects.

Dr. Dan Patterson:

It will also benchmark and validate the durations that the plan scheduler are putting together. And if they are not in alignment, then that actually forms those uncertainty ranges that goes into the risk analysis. Secondly, we’re actually opening up the planning process such that multiple team members can give their buy in or push back to the schedule as part of that planning process. Now what that result is saying is in many cases you have differing opinions from those multiple stakeholders and that’s okay because those differing opinions themselves fold into those uncertainty ranges which goes again into the risk model. And then thirdly and finally risk intelligence. So really the biggest difference here from traditional risk analysis is we’ve eliminated the need and complexity to take a CPM schedule, export it from the scheduling tool, import it back into the risk analysis tool, try and capture those ranges and build a risk register and link the risk register back to the schedule, run the risk analysis and try to interpret the results.

Dr. Dan Patterson:

We’ve eliminated all of that complexity to the point now where in real time the computer is building a risk-adjusted forecast. And that’s quite remarkable because we’ve eliminated the complexity from the user. We’ve also now enabled the planner scheduler to see that P75, that P80 version of the plan that they’re developing as they’re actually creating the CPM schedule. I think this is also very useful for what we call a bid analysis as well. Traditional risk analysis is focused on post-project toward whether it’s detailed planning or indeed execution. We’re actually now offering the concept of risk analysis during the bidding phase as well and taking into account for example, the probability from a contractor perspective, the probability of not only all contingency getting consumed but even worse our margin. So we call that margin erosion analysis.

Dr. Dan Patterson:

From a methodology perspective we’ve developed a five step approach. So step one is to really interrogate the integrity and the quality of the underlying CPM schedule. And again, in days gone by, this would be a separate process. You would have to take the CPM schedule, run it through some form of schedule, critique and interpret the results. Well again, I think we’ve gotten smarter in recent years and said, “Look that’s great the results are useful, but wouldn’t it have been more useful to actually know that we had missing logic or we had a schedule burst or we had negative float as we were actually building the schedule itself?” So this concept of analyzing the schedule after the effect we’ve eliminated that we do it during the planning process.

Dr. Dan Patterson:

Assessing the risk register. Again, in days gone by we would ask the team, okay, here’s a blank risk register, we’d like you to brainstorm risk events, and then we’d go through the process of trying to convert qualitative identification into true quantitative inputs. And that was quite a tedious process. So now again, through artificial augmented intelligence, the computer will actually automatically build a risk register for you. Now, the teams still get the opportunity through the facilitated workshop to endorse or even push back on those suggested risk events, but it’s much more of a consensus-based and a collaborative approach to building that risk-adjusted schedule. As I mentioned with regards to the analysis itself, we’re still using a form of Monte Carlo in that we are running multiple iterations, but we’ve moved away from the concept of randomly sampling and now we’re actually using an algorithmic approach. So what that means is there is no push back or misinterpretation that, okay, well, yes you build a risk model but it’s largely based on randomness. No, there is no randomness in the analysis now.

Dr. Dan Patterson:

And then the fifth step is the final scenario modeling. And again, this concept of scenario modeling, don’t just build a single scenario. Let’s look at the best case, the worst case, and then let’s decide the confidence level. And in the case where it’s for example, 75% certainty, let’s run with that 75% certainty scenario. So I can’t stress enough the value of letting the computer mine historical projects to assist in the development of the risk models. So the way this works is incredibly simple but also incredibly powerful. When the planner or scheduler is building a CPM schedule, all they have to do is share some degree of context with the computer, whether it’s project type, geographical location, order of magnitude or size of the project for example.

Dr. Dan Patterson:

What that then enables the computer to do is when it goes off and mines those historical as-builts and those historical risk registers, it will give more emphasis to analogist-type projects and analogist data elements such that when it returns the suggestions, those suggestions are relevant to the project in question. So again, I think what’s really remarkable here is through machine learning we’ve actually enabled the software, the planning, the scheduling, the risk analysis software to get smarter over time such that if the suggestions that it makes, for example with regards to risk events from the risk register, if it is not making relevant suggestions because the team members will push back on those suggestions, the computer will actually self-adjust. It learns from its mistakes and the subsequent suggestions then become more relevant.

Dr. Dan Patterson:

So obtaining buy-in through human intelligence, again, in prior years, let’s take for example a facilitated risk workshop. The way we would conduct this is we would walk the schedule and we would ask the team to give us an uncertainty distribution or input. And we try to make it somewhat simple by saying, look, you can model the uncertainty range using either a triangular distribution or a uniform distribution where every value has an equal chance of happening. But again, people would really get wrapped around the axle on this sort of statistical approach to building risk models. Again, we’ve completely eliminated that complexity today within the InEight schedule solution because the way it works is the computer actually invites multiple contributors, project stakeholders, team members, to vote on the realism of the schedule.

Dr. Dan Patterson:

And you can see an example on the right hand side of that voting environment. It’s similar to a scorecard. Now, what happens behind the scenes, we have multiple contributors each giving their expert opinion, those expert opinions again, very well may differ. In the example in the left hand side, the foam in the planner, the design lead, they’re all giving different opinions on how long the activity duration should be. That’s okay because what happens is the computer captures those differing opinions and from those opinions it forms, in this case triangular distribution. So the mathematical modeling is still true to the concept of ranges and uncertainty, but the team members haven’t had to be exposed to that kind of complexity. They are simply giving their expert opinion. The computer is converting that expert opinion into a digital entity that can be used in the risk modeling.

Dr. Dan Patterson:

So the net result is, and again, an example on the screen here, we have our schedule that we’ve developed within InEight scheduled. You can see this in the full or graphically in the form of a Gantt Chart, but more importantly, you can actually see in real time the risk-adjusted version of the schedule. So the gray backdrop, Gantt Chart, there is the non-risk adjusted, the colored elements, the yellow, the orange bars. You can see that the impact from both an uncertainty and a potential risk event perspective, you can see the knock-on effect. In the grid itself, you can see not only the Delta with regards to duration and float, you can also see the knock-on effect from a date perspective as well. And again I think this is so powerful because the computer is doing the hard work behind the scenes. It’s running the Monte Carlo simulation. It’s generating the risk adjusted forecast, but from the planner scheduler perspective, it’s showing up in real time.

Dr. Dan Patterson:

And not only that, as you can see in the far top left, the risk drivers and my risk exposure again is being presented to me in real time. And taking that risk driver, that traditional risk Tornado chart and segmenting it and differentiating between discrete risk events and aggressiveness of schedule. What that allows me to determine very quickly is, should I focus my mitigation and risk reduction efforts on those risk events? Or is it simply the fact that as with those green bars in the tornado chart there, it’s simply the fact that our schedule is too aggressive, it doesn’t comply to standard historical benchmarks we need to revisit the schedule? And again, in days gone by, the revisiting of the schedule would be very much a manual process. Within an InEight schedule, the computer will actually give me the opportunity to generate and adjust my durations based on my confidence level and generate that risk-adjusted forecast directly within the schedule itself.

Dr. Dan Patterson:

And just to close off on sort of next generation risk modeling. In the bottom left here is an example where we have a cost-risk analysis and this is really valuable. Again, during the bidding phase, if I’m a contractor and I’m trying to ascertain how much contingency and associated margin to apply to my bid, again, historically I would apply, for example, a fixed percentage of contingency and then a fixed percentage of margin. Well, our thinking now is very different because what the risk analysis actually comes back with is a range of contingency and a range of associated margin. It will tell me even in the worst-case scenario, how much margin erosion I should expect as a result of risk consuming the contingency into that embedded margin.

Dr. Dan Patterson:

So some really, really powerful developments with regards to usability and the ease at which project teams can now develop risk-adjusted schedules. Again, behind the scenes the mathematics, the multicolor simulation the modeling of those uncertainty ranges, the quantification of risk events with their probabilities and their impacts, I would suggest the mathematical modeling really hasn’t significantly changed because that’s never been the root cause of the challenge of building a risk adjusted schedule. The challenge has been the human factor is being, well, how do we make this meaningful to those team members? How do we enable again, a planner or a scheduler to build this risk-adjusted schedule without spending days, if not weeks on trying to build this thing? Well, we’ve achieved that through this one-stop interface.

Dr. Dan Patterson:

And then likewise know with regards to field execution because the contingency has been embedded within the schedule, the team, when they’re marching to the CPM schedule and they’re doing the daily planning and doing the interval plans, the steps, those integral plans and steps actually have appropriate contingency built into them. And again, I think what’s really, really exciting with regards to the direction that we’re headed with next generation scheduling is because now for the first time field execution as we execute the work and we status those interval steps, the computer is actually taking the actualized steps and feeding them back to the CPM schedule. So what that does in a subsequent run, when we run our next risk analysis, the risk analysis will actually take into account those actualized productivity rates. It’s a continuous feedback loop.

Dr. Dan Patterson:

So to close out on what we’ve walked through, again I think the first generation, if you like a form of risk analysis and risk modeling, was absolutely a step forward. I think though the challenge was the risk analysis was very separate from the planning process. And again, it wasn’t the mathematics that was the challenge, it was the fact that the inputs and the range types were very complex and typically a project contributor or stakeholder really didn’t care about that statistical complexity. And likewise the outputs, the results, were hard to interpret as well. So I think now in the year 2020, we finally got a bit smarter where we’ve said, okay, we’re going to enable the planner scheduler to see the risk-adjusted version of the schedule in real time. We’ve eliminated the need for a separate CPM scheduling tool and risk analysis tool. It’s now a one-stop shop.

Dr. Dan Patterson:

The team is absolutely engaged in the model development process through that human intelligence, through the capturing of expert opinion. The fact that the risk register, the project risk register is now embedded within the planning process. And all of this is driven and endorsed and enhanced through this concept of AI, or augmented intelligence. And I think again we’ve made great strides in making the outputs more meaningful. Again, I think reporting that I’m going to be six months late on a project that doesn’t help me. Reporting the range of outcomes and the drivers and again pinpointing those drivers as to is it schedule aggressiveness or is it indeed external discreet risk events gives me a fighting chance when it comes to mitigation.

Dr. Dan Patterson:

So I believe that this concept of next generation risk analysis, again when you incorporate that into next generation planning in the form of advance work packaging, this is a very, very exciting time full for project planning and project management. And I believe that this is again a massive step forward as an industry in ensuring that our schedule forecasts are more achievable during project execution. Thank you all very much.

John Klobucar:

Thank you Dan. Again, if you have any questions, please email them to webcasts@ineight.com. To learn more about InEight as well as our solutions that get you on your best path of construction. Visit ineight.com and click on the “Request a Demo” button. And if you’d like to see a schedule of upcoming webcasts, just visit ineight.com\webcasts. Thanks for watching, this concludes our presentation.