Session Notes: From Static Portfolios to Real-Time Decision Flows: How the 2AM Scenario Is Redefining Pharmaceutical Portfolio Strategy

Apr 23, 2026
From Static Portfolios to Real-Time Decision Flows: How the 2AM Scenario Is Redefining Pharmaceutic - Transcript
00:00:00
 
 : Not sure how that works. I'm here at Yeah. Yeah. parallel, but it looks like I'm But then I'm just like, ah, it works perfectly. Sound check. Hello everyone. We want to start with our we have Yana Vanguard from Captario giving us a presentation uh how the 2 a.m. scenario is redefining pharmaceutical portfolio strategies and you said it will be very live yes and not just slides so I'm looking forward to that. Thank you. Let's see if I work because I naturally speak loudly. So, can we turn it down a bit? Okay. So, welcome everyone. I will put my timer on because uh there is a lot to ever and uh this is going to be a combination of a presentation but mainly I'll be in a live demo and we're going to run through a uh scenario that I think everyone recognize more or less and um so a lot to cover and also by demo so you can judge yourself the probability of success but I'll do my very Yes.
 
 
00:02:37
 
 : And uh uh slow me down if if needed, but I'll try my best to keep a red throughout. All right. So, uh there are a couple of assumptions here. One assumption is that we are part of a a global project team crossunctional that shares information and uh and ownership and and accountability for the information and of course a lot of figures that will go into this uh the junk in Junk principle still applies. uh so it's really up to the quality of the information of course uh based on the expertise and knowledge and experience of the GP uh but uh having that in place uh the story is really about how you can leverage that knowledge uh there there are a couple of things um that I'm going to try to claim one is that the uh banning infrastructure which we're all used to not infrastructure and I think there is a general lack of decision infrastructure and also that decision work means comparing different options so that's what we'll be doing and uh it's it's also when you work with several options you can't go into detail with everything.
 
 
00:04:21
 
 : So how you get that because I I know that there is not a common way of looking at things and one say this and one said that. So um a way to get around that and then something that I personally learned the hard way uh working as a business consultant working from a company that uh it was a clinical initiative where we optimized the clinical design uh only to find that when that optimal clinical design project team they decided something uh which taught me that uh it's hard to optimize something unless you go beyond the next stage of development and actually end to end and look at it cross function. So that's what what we try to do and even if it seems optimal at a project level it may not be optimal for the portfolio. maybe the portfolio can absorb the risk that that seems uh apparent in in the optimal uh option for the project. So being able to move between the project and the portfolio stretch. All right. Yes. Can you take take it away a bit?
 
 
00:05:44
 
 : Yeah. Good. Yes. Like adjust it to I think yeah it's because I speak too loud unfortunately my wife so the 2 a.m. scenario and really we should see as efficient. So it's provocative. I know that this uh process sometimes also for good reasons take uh weeks and sometimes months. But in maybe this could be So this is um and even if it means cutting uh just a weekly forms. So the I talk about the decision infrastructure and the planning infrastructure. So the planning infrastructure is is uh well tried out and it works great and has huge benefits in agreeing what the future is expected to look like. But then something breaks that model and uh something like in the scenario that uh we're going to be looking at where a competitor gets out this information and we're shocked by the fact that it's so good place to data and that we because this uh a new work. So then we need to take the path of uh the uh decision and the decision.
 
 
00:08:13
 
 : Uh so so here is a a comparison where the plan is uh detailed. It's uh it looks at one planet and uh it's uh deterministic and static and this is natural and I think so I'm not claiming that that should be different but then on the other hand you have the decision and uh it uh has multiple paths that need to be evaluated. It's high level It's probabilistic and dynamic. So, and the plan answers what do we expect while the decision answers what should we do and what. So, let's then uh turn to uh the scenario. So uh this assumes that we have a plan and we are now dynamic model and we already have the model in place. So this is a very high level model that we're using. Uh but and if we would need to detail such as knowing how to be then there are more detailed uh but in this case we want to keep it up. So uh first we see to that the dynamic uh the equivalent of the static equivalent and that they match because then that means that we can work from the baseline dynamic from their fun and then we're going to evaluate the different options.
 
 
00:10:13
 
 : So one option well the first one the bas is how what is the impact of the project given this new information. The next one is looking at uh mitigation. So we can accelerate and go head on uh with this competitor. we can choose to kind of leave that competition and go uh for efficient certification or we could expand into a different indication and uh maybe that's a more safe space for us. So we're actually going to do this together. Uh and we're now moving into uh the tool. So I talked about this uh model. The um the static one is pretty much just time series. So different accounts and different uh costs that occur over time. So time series while this is a dynamic model that when you push something in time everything else adjusts and that may have consequences for instance if you push and delay something then the likelihood of for instance a competitor beating you to market increases and then you'll see what kind of impact that has. Okay, so that's a dynamic model and just to have a quick look at it um uh from a development point of view very high level again and straightforward.
 
 
00:12:05
 
 : So there is our different places and we have a duration in months and uh I've actually put in a range here which is something I'll come back to sometimes when discuss details. One says it's 34 months, one says it's 38 months. And rather than uh having to solve for that and and uh land at number and and move on then we can also try you know different uh ranges and and even different uh point values to see to what extent it matters. So that helps us discuss what really matters. All right. And then uh a probability of success uh for each phase and then for commercial. So this has uh a bit more detail in it and the point of that is that we need to re-engineer the uh the figures uh that are in the time series so that we can reproduce them in a dynamic way. And in this case we have an empty model where uh we have uh available patients and then we have a class share and then uh we have a market share but we have market share we're first market and then another market share if we're second or or third.
 
 
00:13:40
 
 : Um all right. So let's then just check and see that uh these match up the static one and the dynamic one. So here you can see that uh the dark blue is the baseline static that is what we imported and the baseline dynamic is uh what I just showed you. So uh we've been able here to make these match and this is a typical thing for AI uh to do uh and let's even see from a value point of view. So now by the way uh we run these point 100,000 times. So it's it's a I don't know of any faster uh simulation than that. Uh uh here we can see side by side what the MPV is and what the time schedule looks like. Uh, and so here is something that may interest you. if we lose because we'll never end up here. We'll end up we actually make it to market and then we win. But if we win in average, how much do we win? What's what's the value there?
 
 
00:15:05
 
 : And if it ends up a loss, what's the average loss? Okay, so it matches uh fairly well. And from here on we'll move uh we'll use the dynamic equivalent of so follow so far. Yeah. Yeah. Um so let's then use that a dynamic baseline to see what the impact of. So uh uh I will this and call it uh new. All right. And then let's uh put in the hard fact of the competitor. So the competitor uh was assumed to be 50% uh of reaching uh a launch and uh uh I'll just say 75% now because they have such good data and uh we don't know when they're launching. uh we assume that it was uh June 31, but let's then make an assumption that it's uh somewhere in between uh 2030 and 2031. All right. And then uh let's see what the impact is. And I use a dashboard that I've created, but I can create any graph that I want. And if we now compare this with um it's this project and we have the basic So the first question is what is the impact of this competitive and this is the impact that's I'm now showing you uh averages and we can do so much more given uh the rich output but uh not to get it to be too much to consume uh I stick with the averages.
 
 
00:17:55
 
 : So there's a value loss obviously uh and uh there is if we look at the launch timing so uh it launches because we haven't accelerated it launches in April 2031 while the competitor as we stated uh uniformly uh launches somewhere in this range. So they beat us to market but it's not always they beat us because it's 75% of the cases where they actually uh launch in time. So we still have a chance to to make it first to market if they actually are. So uh so check uh on first one. So now let's move on to acceleration. So I'll go back. I'll create an option of uh the baseline new and let's call this exceptation. And uh so can we uh maybe move this down to 6 months earlier? So we say somewhere between 29 months and uh 30. So uh this is a triangular distribution. So the most likely is that it's 30, but it could be uh as little as 29. But also we could be delayed. So let's say 32. And then the question is uh what what did that cost us?
 
 
00:19:59
 
 : Well, maybe it cost us and this is again where we can have a range. Someone says uh 60, someone says 80, someone says 40. So no need to be exact uh because we can always use a range. But uh let's say that we increase this to to 10 and possibly it comes with a cost. Maybe we have fewer patients. Of course, we have more countries and uh sites and so on. Uh maybe we simplify the protocol a little bit. And that in turn uh will decrease our likelihood of reaching success with uh 2% with seven maybe. So this is the discussion that we would be having in the GPT uh where it's really focused on timing and we we keep the other things uh the same and we just then see what does that where does that take us and so now if we this again take this dashboard that we now recognize And we compare this uh with the uh baseline the new baseline that we created here. Then we can see that.
 
 
00:21:43
 
 : Okay. So we broke down is reflected in our which I'm sometimes reflect. So this is the the principle and uh then if we get stuck there is no need to get stuck on anything because we can just uh if there are two different opinions of how long we can uh um or what the cost is or how much we can accelerate then we just try out both and if we can't agree if if it doesn't uh mean a difference then we'll just move it to the side or we remain with a wrench and move on. So, so that's that's the principle because otherwise we get stuck in any detail and we lose time. So, that's I guess the value of information. So, is it worth getting the perfect information of what the acceleration actually will be? Um, okay. And then the uh um 4 minutes. Oops. Then the uh subgroup. Yeah, you just was there but went away. That's good. Um, so then uh let's look at this.
 
 
00:24:07
 
 : Uh, let's see. So, it's the acceleration. Here we go. So in this case, we decided to uh oops uh we decided to go with a subgroup. Uh here it is. We decided to go with a subgroup and uh we increase the market share uh we increase the price you know the expected things and then again we can come up with something and uh and we can then there is the demo there. Um and then we can uh compare just like we compare everything else. Uh then for the portfolio because now we want to see how these options play out in the portfolio context. So here we have so now we have moved to the portfolio and you can see the uh basic portfolio default. So that's where we started and then you have the new baseline. So there is the loss and then we have the acceleration and then we have the stratification. And uh you can see some more details. You can see all kinds of different things.
 
 
00:25:58
 
 : And the interesting thing here is of course uh can we uh accept this additional cost? Uh can can we uh absorb that or do we have to make another decision in order to make room for that additional cost and is that worth it? So this is again how the uh the need to pay between the project and All right. Uh the indication expansion my last point is uh here. So in that case you have the original project that runs the original way and if that becomes successful in case then we actually start a new indication which you see below. And that new indication has a dependency that says if if we actually were successful in our original project, then we're more likely to be successful here. And we can actually uh spend less time on this portfolio. And so you would have to add those numbers, the value numbers. And you can also take this portfolio into the big portfolio and see also this option. So um with that that that was pretty much the story, wasn't it?
 
 
00:27:39
 
 : Um we made it in 20 minutes and I have no illusions that uh a team would take 15 20 minutes because I already uh I think uh what this can do is help the discussions on what really matters and it helps evaluate more options so that when you go to government and go with something uh three point out. Thank you, Johannes. Is it on? Very interesting tool. Thank you for the demo. Questions for Yannas? Yeah. Thank you so much for the demonstration. I just had a a question. So just to understand augmented part comes into it. Maybe that's part two. Part one, we're entering let's say shortening type making scenarios. Does the system set its guard rails so that you can't put in unrealistic timelines? Do you see what I mean? So that there are set guidelines so no one can go in and change timelines but that the system monitors and yes and protects realistic time. Yeah, I know exactly what you mean.
 
 
00:29:24
 
 : So uh basically do we have benchmark data to validate the input that we provide and uh so we have a principle and that is let me see if I can find that slide um here. So uh in the bottom we have a methodology. So we provide these uh models that go in detail or high level and those models uh assume input and that creates the scope for the benchmark that we collect and uh so we have this benchmark database in place but we're now integrating that with uh the tool so that we can do exactly this and and that also means that you can start from scratch and just use benchmark data But it would also potentially raise a red flag if someone can say, "Oh, I'll do it in two months when something takes two years." Exactly. So that would be signaled both to the project team but also to the portfolio because it's uh at least they are then aware and can challenge or maybe the the team can defend it but it's it's worth having the discussion. Yeah.
 
 
00:30:55
 
 : And and as for the AI, our thinking is that rather than use AI as a black box to do everything, uh we actually stay with uh these four pillars and we have AI that acts on top of it. Uh but that means that so I'll you can challenge me that by the end of the year we'll be able to run this whole scenario using just plots and that means both the the modeling aspect the data and the analysis. So that's what we're working towards. But that also means that we have full traceability. We have comments for I didn't show that but for the input so that you can defend your assumptions and everything is traceable and our thinking is that if we capture over time decisions and also follow them up uh we can have AI to learn about the decision and recognize the decision so that when you arrive at the same type of decision you can actually uh be better based on knowledge or of experience. One last question for you. I think everyone is completely flashed off the possibilities. So, thank you. Thank you. effects.
 
 
Transcription ended after 00:33:28

This editable transcript was computer generated and might contain errors. People can also change the text after it was created.