 Okay, with this we're ready to move on to our scenario 2c. So, so far in all of the simulations we run, if you look at the data that's generated at the server reports that have been simulated, they're kind of homogeneous. Let me show you an example here. If we look, for example, at our data table for our purchase value for our last simulation. This here, remember, this is the the bucket name of our aggregation key and this is the value of that bucket. So, how big the blue bar is going to be in there. So, here it's actually quite straightforward to see from that example. The summary values we have in there by default in the way Nozab is going to generate example measurement data, they're kind of close. It's kind of homogeneous, right? We don't see a lot of variability across these buckets. But if I had very small buckets in there, for example, empty buckets, buckets with no conversions at all, this could greatly impact Noz. So, if I'm an API user, this is something I'm going to be interested in simulating. What happens if I have buckets with no conversions? In the advanced mode, there's actually an option to do that. So, let's go ahead and use that. This option is available here under conversion data. You can see here, I can pick a percentage of buckets with zero conversions. So, let's try that out. And let's pretend, let's assume that for that simulation, about 5% of my buckets are empty, that is, they have no conversions. So, now let's go ahead and simulate. Now, if we look at the results of the simulation with 5% of zero buckets, the first thing we notice is that our average percentage error metric is now infinity, which makes sense because we have zero buckets. And this is exactly one use case where RMSP, our second metric, which is also more stable, comes in handy because this value here makes much more sense to me. So, this is also higher than the value we had for the previous simulation with no empty buckets. And this is expected. We want the relative impact of Noz to be really high on small buckets in order to protect user privacy. So, if we think back for a minute about our orange and blue bars, if the blue bar, which is the true value, is really small, for example, you have one conversion by one user, then the orange bar, which is the Noz, will have a larger relative impact. And this is intentional so that we can protect the privacy of that individual that's alone within that bucket. So, here we had zero buckets, but hopefully this also illustrates the point for very small buckets and why the relative impact of Noz is and should be higher on smaller buckets.