 There's one thing you may have noticed. I already have a simulation ready for us There's one thing you may have noticed is that we have two measurement goals here We are measuring purchase value and purchase count This is what we've been using across all our examples And maybe you've noticed that the nose ratios for purchase value Always seems to be a little worse than they are for purchase count Here the difference is about five percent to twenty percent for purchase value and Under one percent for purchase counts, which is no small difference And that could be surprising because the actual underlying measurement data and Parameters that we've used are the same for these two measurement goals So why do we see this difference? And is there anything we can do about it? To answer this question, let's first take a closer look at what we're measuring here I was saying the measurement data is the same, but this is not exactly true Let's imagine we have a conversion count of ten conversions per bucket Our first measurement goal here is the purchase value So if we assume that people purchase shoes for about a hundred dollars on average Each bucket will have a value of ten times a hundred, so about a thousand And our second measurement goal is the purchase count So if for each bucket we have on average ten conversions Well each bucket for purchase count will have a value of about ten So what this means is that the purchase count buckets are way smaller than the purchase value buckets Now remember we're talking about scaling Scaling is the setting right over here It's turned on by default And scaling is a technique that techs and API users can decide to use So here in those labs it's going to assume that you have implemented scaling And it does it for you out of the box But what Nozlab does by default is not very smart in terms of scaling By default what Nozlab does for simplicity is it just splits The contribution budget which you can think of as scaling budget By default Nozlab is going to split this budget Into equally across our two measurement goals You can see it right there We have percentage for measurement goal one 50% and 50% for the second measurement goal But in reality as we've seen these two measurement goals are going to translate into Very different orders of magnitude in terms of data And Nozlab in the default setting doesn't really take this difference into account And as a result purchase count ends up benefiting from the scaling technique Much more than the purchase value does So this right here this explains what our purchase count Has much lower Noz ratios than our purchase value So what we could do instead to try and optimize that and rebalance that a little bit is Well help Nozlab make better decisions in terms of scaling So here what I'm going to do is to assign more percentage to my first measurement goal Let's say 90 and I'm going to assign only 10% of my scaling budgets to my second measurement goal Which is purchase value Purchase count sorry So here what I'm hoping is that this is going to rebalance out a little bit this this This inequality between these my two measurement goals This may mean that my purchase count Noz ratios are going to increase a little bit But I'm hoping that this would lower my purchase value Noz ratios So let's go ahead and try that we're going to simulate All right, let's compare what we have here 5 to 20% and under 1% and here We see that our purchase value Noz ratios have decreased as we expected as we wish to achieve Below 5% which could be more acceptable for my use case But I result as a result my purchase count Noz ratios have kind of increased by quite a lot So there is probably a little bit more tweaking I could be doing here and play around with my parameters But hopefully this gives you another view of how you can leverage scaling and use more advanced scaling configurations To ensure that your Noz ratios are acceptable for your use case across all of your measurement goals