 Okay, this is the last part of our cluster analysis overview. Solver found optimum locations for each of the cluster centroids, for each of the offers and the relationships to the customers and how they've reacted for those offers. But it's hard to tell what's going on there. One thing we can do is to use conditional formatting, I'm going to select that first cluster, go to home, conditional formatting, color scales, and I'm just going to select a simple one which highlights in darker shades of green the larger the value. You can see for cluster one we have down here these particular offers 24 and 27 have very high numbers close to one. The closer to one that a cluster center is for that particular dimension, the stronger the relationship with how the customers reacted. So something's going on there with those particular orders and just glancing real quickly. I can see they're both peen on noir and that is about the only thing that jumps out at me as to why people jumped on those. Let's look at cluster two and what I'm going to do is just select that column and copy that format and copy that conditional formatting to cluster two, copy it again, go to cluster three, copy it again, and put it on cluster four. The reason you have to do it individually is that we want to locate the small and large numbers in each of these clusters as opposed to putting it as one big range. Although there are some more complicated and in-depth ways we can go further in this analysis and look at the patterns, you can just stop here for our purposes. What I'm going to do is to copy this range of information there, get rid of the extra work and just right-click, copy, go to a new tab, paste it in, and now I can do a little bit of sorting to help with the analysis. I'm going to go here to column H, cluster one, and go to data and sort from largest to smallest. It'll ask me to expand the range and it does and it organizes the information about cluster one from the largest value to smaller values. We'll look at these ones up here in green and see what jumps out at me. Again, we find it is peonor that these folks in cluster one seem to be interested in. They want things that are not past their peak but they don't seem to care which country. There's no pattern there with the discount, though I'm sure take whatever you give them, and they don't seem to care too much about the minimum quantity or even the, well I get to take it back, that July, October, September toward late summer seems to be the most popular time for that group. So you could structure your next custom newsletter for cluster one about peonor noir, publish it in that late summer, and make sure you're talking about wines that are in their peak. Now let's look at cluster two. I'm going to select column I, and I'm going to sort again from largest to smallest, and here we've got our dark greens up at the top there. Now let's look at the patterns and what jumps out at me real quickly. They like minimum quantity orders on these special deals, so you could focus on that. It doesn't seem to matter which month, which part of the year, which variety, oh Espen Monti, they bought a couple of times. The discounts, they like pretty heavy discounts. Those are all 45% or more, and they don't like things that are past their peak, so I would focus the newsletter for cluster two on items that have a minimum quantity, a low minimum quantity. If we look at cluster three, I'm going to sort that again, and here we don't have that much of a pattern except one real major thing right there, which seems to be that these folks in cluster three like champagne in December from France, that is prime, and a moderately sized minimum quantity and a pretty good discount. They like pretty good discounts, so use that information to design that special newsletter. Finally for cluster four, we're going to sort that from largest to smallest, and here though, remember we don't have anything really strong. These these values are all pretty close to one another, but looking across here, we can see that again it's pretty muddy for cluster four. They like champagne, they like moderately sized minimum quantities, they like big discounts, they like things from France, but again it's not super strong, so I might be a little bit wary of how much weight I would put on this information from this cluster analysis, and that brings up a point. We just arbitrarily picked K equal four for clusters. If you had a smaller advertising budget, you might reanalyze this for just three clusters. If you had more money, you might reanalyze with five clusters or six clusters, and then go through this ranking sorting procedure again to see what patterns jump out. Now this is a very simple problem that I've given you here. In real world, you would have more data points, more characteristics that you could put into your database and sort for. But the key thing is to take your customers, take your individuals, and put them into groups so that you can market to those groups, and that's what cluster analysis allows you to do.