 Not sure, how many of you recognize this city, it doesn't matter of London. London. The red spots indicate where there have been lots of flicker photos. And the blue spots indicate where there have been a lot of quitter photos. So we just put them all together and said, what does the city of London look like if we do it this way? So there are the red spots, which are clearly the tourist spots. So that's the Tower of London. That's the working of others and so on. And the blue spots are where there's more news being made. This compresses how many hundreds of thousands of photos and hundreds of thousands of tweets just based on a geographic perspective. But it's also a ten year story. It's saying that's where the business district is located. That's pretty much where the tourist spots are located. And this spot that's a combination of both is Oxford Street, which is the biggest tourist and business area that's out there. In fact, he also told me a few spots that I hadn't been to that I should visit. In Rose Hill, for instance. What's a bright red spot there? Let me go there and have a look. And it has this amazing view of the entire city of London. Now, what I'm going to be talking about is stuff like this. How you can take data from crazy places. Put them together in a way that you would not have normally thought about. And try and get something new. Data visualization is really about telling stories with images. But about myself, my name is Anand. I have chosen the designation of data scientist at the start-up that we founded. It's Garanda. And I'll begin with an introduction to what data visualization is. How it started a bit of history. What kinds of visualizations can be produced. How you can learn more about it. That's for the first half. And then the second half, I'll go through some examples of how you can create data visualizations in JavaScript. It's actually extremely easy. Once you get on all the basics, you don't need a library. But I will be talking you through some libraries if we have time. Data visualization actually started having surprising origins. This is probably the first popular data visualization. Created by Florence Nightingale. Her contributions to the nursing industry. Actually, her contributions went well beyond nursing. One could argue that she is the founder of modern data visualization. This is a chart that she put together for Queen Elizabeth. The black areas show how many people died in the war by getting killed in the war. And this is month on month. The pink areas showed how many people got killed out of related causes. War related causes, but not quite when they were fighting. And the blueish gray areas showed you how many people got killed in the hospital. Just because they were not getting enough medical treatment. She got her funding, Britain won the war. She got her funding for hospitals. And Britain won the war. That's a map prepared by CP Snow. He was a mental practitioner. He was trying to figure out what was causing the cholera to take out. His theory was it was the pumps that were infected. And he went household by household. Mapping how many people had cholera infections in those households. And how far they were away from the various pumps and which pump they were taking water from. So each line is a number of people in that household who had cholera. And he brought this for a fairly large area and sent it down to this one pump. That was the main cause of cholera. Geographic visualization in the 18th century. There are many more. This is not something that is new. The point is, if they could do it in the 18th century without computers, you can do it very, very easily. The thing that's stopping us or the thing that we need is not as much an awareness of technology as much as an awareness of what data visualization is and what it can do. Let's take this table. I've got the sales and price of a product across four cities. So let's say it's a bar of soap that's being sold across various cities. And each of the cities, the branches of the cities have the flexibility to change the prices as they want and as the price changes, the sales wakes up. Bangalore, the average price was 9. The average sales was 7.5. Delhi, average price was 9. The average sales was 7.5. And so on for all the other cities. If you look at the variance, which is the square of the standard deviation for those that are familiar with it, that's the same as well. It looks like the same. The cities don't really seem too different. I'm going to plot the same data set. Delhi, the sales rises with price and then falls after a point. Bangalore, it sort of increases but it's not a very strong correlation. Hyderabad, you would argue that it's almost a perfect correlation except for one. What looks like an aberration. And Bombay, they never even changed the price bar in one point. And you would argue that that's not enough data to include anything. The scripted statisticians can be very misleading. There are a few points that you might want to take away from this session. And if there was one really strong point, I would simply say, don't bother with averages. Whatever the data, just plot it. There is far more that you can take out of a plot. No matter how bad a plot. But particularly something that shows every single point, rather than summarize it into a smaller data set. Just plot the full data set. See what you see. And invariably that will lead to better insight than the rest. Some of the work that we're doing involves pattern detection. And energy utility approached us and said, we know we have a lot of fraud happening in our meter readings. We've been pushing for automated meter readings, but there's a lot of resistance from the union and we need some kind of proof. And we're going to be showing that the unions don't expect any sophisticated statistics to sink through. They're going to have to keep it as simple as possible. And we'd also like to find that evidence. So we took the data. We had a very large 180 megabytes, which had the iterating for every single person for about 12 months in that state. We said, let's do the simplest thing also. Let's plot a histogram. That's the histogram of the materials of every single person in the state. The number of people that had a meter reading of zero, one, two, three, four, five, so on. And you see some spikes. That's a perfectly long, smooth, long-armour curve except for those spikes. Those spikes primarily happen at 15 units, 100 units, 200 units, and 300 units. Which is exactly the number, the meter reading that people need to have to stay within this lab, okay? If you have one unit more than 100, then you get to the next lab and therefore pay at a higher rate. That is wrong. But incidentally, we were also talking to this guy and said, look, we can understand the reason behind why 50 units and 100 units is so on because if it's any higher than they can get at that rate. But one of those spikes at 10, 20, 30, 40, they said, okay, he said, take a guess. We had no idea. This proves something that we, for a long time, had a suspicion about. This is not fraud. This is just laziness. The guy never went there. He's just putting a round number for the meter reading. It gives us an even stronger case to push the automated meter reading. Tell him, look, you aren't even taking the meter reading. What's your problem? Then you look at, for instance, is this correlated? So, for instance, this is the same set of people that are having the meter, that have the meter reading of 100 and so on. So, for instance, there was this lady whose meter reading went to 100, 200, 200, 200, 200. I mean, that's really hitting your energy meter to the right level, either that or fraud. Incidentally, when he went through the list of names, he said this, unfortunately, not much of them. We can prove all the names on this list. They happen to be somewhat influential people, but it's good to know where this is coming from. And also, we looked at it from in different regions. So, we took one specific city, broken down into sections and tried to see this excess of fraud that you see. So, let's say this is about 40% higher. Does it vary from place to place? So, in section 1, how much is it in section 2 and so on? So, section 1, this excess was about 70%. This is roughly 70% fraud. 97%, 100% 6%. So, whereas some sections have relatively lower the fraud, 9%, 15%, 20% and so on. So, you also know where the problem is focused. But also, there was one anomaly. Section 5, that's a bit of a tip. All of a sudden, the degree of fraud drops, then stays a bit low and then spys up after a point. So, let's get set up your mover, pulls up, just looks through it and says, ah, Swamasundaram got transferred in, in June and got transferred out in September. Now, I know what to do with the pile of complaints against him. Stuff like that. Or another case. So, the second chart was prepared from the first one. We had a huge data set which said, there is a customer. In this month, they had this meter. And you also know for each customer where they live, in which section they live. So, the second chart is the fraud percentage. The second fraud is the excess, the height of this chart, divided by the height of this chart. Okay, thank you. Now, if you notice, this does not involve any sophisticated graphics. You can do this in Excel. In fact, we did do a version of it in Excel also because those guys then wanted to replicate it. So, part of data visualization is not about the sophistication of the graphics. It's really about the sophistication of your imagination, or not even sophistication. It's about the simplicity of your imagination. We will be covering how to do some of this in JavaScript, but just keep this in mind as well that you don't really need sophisticated graphics. Another one is, okay, this was with financial services provider where they said we want to understand some patterns in securities, be it currencies, be it commodities, be it stock indices. They put these together. Now, that's the set of currencies, commodities, stock indices. And what Excel shows is a correlation between a bank. So, for instance, 68% is a correlation between the Australian dollar and the European. And that's called slightly green because that's slightly positive. And where you see red values are negative numbers. What you also have here is a scatter plot. So, on any given day, and this was over a period of six months, on any given day, what was the Australian dollar price? What was the euro price? And you can see that in general it moves up. Now, why is this important? Because when you saw the earlier slide where you saw some patterns are not always obvious in terms of the number. Now, we've seen cases where, for instance, specifically, gold versus Swiss crack. Gold versus Swiss crack has this movement where initially it goes up and then it goes down. So, the correlation ends up being close to zero. But then, with gold versus some other currency, which one, was completely all over the map. That also had a correlation of zero. So, a correlation of zero can happen for two completely different reasons. There was a period when it was strongly positively correlated and then strongly negatively correlated. And then another is where it was just completely neutrally correlated. You've got to see some of this to understand the real pattern. So, we plotted that and found that there are, you know, there are blocks of correlated currencies. The Singapore dollar, Japanese yen, gold, Swiss crack and Chinese UR, they tend to move close together, very closely with each other. And this is another block. The S&P, the FTSE, the BSE Sensex and for some reason the Pakistani Rumi tend to move very close to each other. And also, when one block moves up, the other block moves down. So, when any of these currencies go up, these indices and this currency goes down. You can also see that and what that therefore means is if you're holding a lot of gold, what's your best hedge against gold? If gold goes up, what goes down? Your best bet is probably holding the frozen. Hold and I have an aggregate of the British stock index, that's your best bet. If you hold a lot of Indian rupees and want a hedge against that, not much that you can hedge, your best bet is probably the Japanese yen, which is going to drop by around 27% if the Indian rupee rises, but more importantly, the Indian rupee drops. It rises up a bit, but there isn't as strong a hedge against the Indian rupee, at least in the prevailing conditions. Stuff like this, I'm not going to go on more into the details. I want to give you an idea of just two things. A, data visualization is old and fairly simple, does not require much by way of technology. And that the things that you can do with it do not require much by way of tools. You can do them on Excel either. But let's dive in into, before I dive into the demo, just want to mention, in case you are looking for further reading, look for any book by Edward Taft. T-U-F-T-E is the be-all and end-all of data visualization. And any of his books would just read them. And these are relatively old books, meaning 80s, one of them was in 70s. And don't talk anywhere near about, anything about the technology. We'll tell you what kind of visualization one can produce. But now let's dive in into a demo of the simple data visualization. I pulled out from an ISP data about when people go to class. So what I got was Needs City, that's the third column. So Needs City, at what hour of the day and at what day of the week do people browse and roughly how many of them are there. Now let's try and see if we can create a visualization out of that. What I'm going to do is create a visualization that has the seven days of the week and the 24 hours in a day split up as a grid. And depending on the number of people in the cell I'm going to vary the intensity of the cell. Any guesses on how you would do this? Heat map. Yeah, this is a heat map, incidentally. Any guesses on how you would do it in HTML and JavaScript? D3 tables. D3 canvas. Rotovis. Ruffle. Simply RGB. Simply RGB. SVG. SVG. So two of those answers were technologies. Two of those answers were techniques and three of them were libraries. Let me walk through those. So we have SVG which is a way of drawing on the screen. We have canvas which is another way of drawing on the screen. We have Rotovis. We have Raphael. We have D3 which are like a piece. And we have tables and we have RG which are techniques that we would use to draw on it. We'll start with the simplest. I'm not even going to go as far as tables. What I'm going to do is make each one of these a div element. It'll have a width of 50 pixels. It'll have a height of 50 pixels. Depending on the R and depending on the day of the P it'll be positioned at some location. And the position is very easy to calculate. It's simply R times 50 and V day times 50. That is the X and Y coordinate. Width is 50-50. The color we'll have to work out. Now let us assume that the maximum range for this is 2 million visitors in that R. So we'll say 2 million visitors will represent bright blue and zero is white Let's walk through the code that would be this. The first thing I've done here is copied and pasted this next. So I just filtered Chennai and for Chennai copied this entire chunk even this So I just pasted this. Now I'm going to comment out everything that I've got here in detail. Let's see what this looks like on the browser. That's not what it looks like on the program. Sorry, I didn't do the wrong one. I'm going to remove all the scripts that I've put here and show you what it will look like to start with. Fine. So we've got this data set. Now we're going to pass off this into the key grid that you saw earlier. Let's walk through the steps for doing that. Now I've included two libraries. One is jQuery. Mostly because I can't live without jQuery but you don't really need jQuery for this. The other is a very simple library called Color. What I'm going to do is transform a number into a color and I'll walk you through that library. It's just code, 20-30 lines of code. That's very basic. Think of it as given a number from 0 to 1 it will convert it into RGB from any state to any state. You could just as well replace it with array times value, comma 0, comma 0. Now what we're going to do is take the preview which is where we've got all of the data and we convert the text into an array of arrays. So this data variable is an array that contains the lines split based on the new line and what we do is split it based on spaces and return an array. So the first array let me show you what it helps again that easy. Now let's see what it helps. Let me just show you what it helps. It's an array of arrays and each cell in the array contains the numbers or the columns that you call. So the first one was 0 which is the day of the week the second is the hour of the week the third is the city and the fourth is the day of the week and so on. So now that we've got the data extracted and incidentally the way I'm doing it is possibly the worst way of doing it. You'll typically source data as either a csv file or a json file. If it's json you don't have to do anything with a csv file there are enough csv to json passes. Please do not write code to do any of these transformations. That's important enough to say that please do not write code to do any of these transformations. In fact in general please do not write code. Code is a library what you want is a functionality. Someone has written it already and now I want to write it. End of rank. Then now what we're going to do is draw a box. Now first I create so now we create a grid. This grid is just one big div. That's going to have everything in it. And then for each row in data we're going to do the following. We create a div and we add a class called cell. Now cell is simply a div that is absolutely positioned as a width of 15. So instead of how to specify that explicitly I put that in the style right on top. That says that a cell is an absolutely positioned element with a width of 50 and a height of 50 and a nice black border. Now then we say position it so that the left is 15 plus and just leaving a little gap of 50 on the top left plus 15 times the row of 0. The row of 0 will tell me how it is positioned horizontally which happens to be the R of the div. Then the top position is just start at 50 and 50 times row of 1 which is the R of the div. So it's positioned each cell at the right location and then put in a background color. A background color uses a function that's defined in color.js and I'm saying gradient of row of 3 which is the number of visitors divided by 1 media. Because I happen to know that the maximum number of visitors in any hour is long again. Now you ideally ought to do this programmatically but I'm just keeping it simple. So therefore this number is going to be something between 0 and 1. And this reds is a gradient palette that has been defined. Now if we just did this and then ignore this line for a minute, data of value comma row and just storing against each cell what the original data is to look it up later. We'll come to that, don't worry about it. Take this entire div that we just created at that location and add it to this big grid. So one by one it goes through the row, it goes through the data and plugs it into the right spot. Now with that we have this little equation. You know this little thing that pops up and I'm going to show you how we did that. But that's what we have. Now let's spend half a minute looking at what it's telling us. Firstly it's telling us that people start well this is 1 o'clock, 2 o'clock 12 o'clock, 1 o'clock, 2 o'clock and so on 3, 4, 5 6 is where a few people wake up. 7 is when more people wake up, 8 is when they really start browsing and so on, so on, so on. It goes on almost till 10 in the night, at around 11 in the night is when their traffic starts dropping. And if you look at it, there's this like it starts a bit later here which is on weekends. These are people who like to sleep a little more on weekends. Now, we then took this to the next level. Many cities are different in their behavior. Let's take Bombay, let's take Chennai, let's take Delhi, let's take Bangalore and try to say how these cities differ in their browsing behavior. So there were two factors in which the browsing behavior was different. How early they wake up and whether they work on weekends. And it almost falls into a clean 2 by 2 pattern. Bangalore and Chennai wake up again. Bombay and Delhi wake up late. Bangalore and Bombay work on weekends. Chennai and Delhi don't. I'll leave you to draw the inferences, not committee or anything like that. But the way that was done was simply by taking not the absolute value of visitors per city but taking the difference in the percentage of visitors and then plotting that as a hit. That's available on I'll put you to the reference towards the end. Now we have added one little twist to this which is as you hover over these cells in Chennai, so for instance what 500,000 visitors came in at Tuesday at Chennai which lets you play around with the metrics see what's happening and so on. The way to do that is really straightforward. I'm not going to actually go through the code here but you just anytime you hover over a cell you find out the position of the cell and then move it a little bit so that you don't block the mouse and display it and set the text for this hover element and position it there and show it. That is pretty much it. So about 50 lines of JavaScript and that's an example. We have not used SVG. We have not used campus. We have just used disk. We have barely used jQuery. We didn't need to use jQuery. The thing that we've used is this gradient function which I'll quickly walk you through and you'll find that it's fairly basic. There's one helper function that converts numbers like ffffff and we have a gradient function that does the problem. So if you say gradient of 0.4 sorry for those of you who can't read it let me just talk you through it. If you say gradient of some number and you specify an array which says for 0 it represents such a color. For 1 it represents such a color. It just interprets the values linearly. There are far more sophisticated ways of doing it. There are good color models that one can use but doesn't matter. With visualization you have a lot of leeway. You can afford to get it wrong in many many ways and it will still be alright. You can also try very hard to get it right both are there. So that brings me to the last segment of my top. She'll take a couple of minutes in which I'll show you a few things that we've done except that I'm not going to. We're doing all this at the kind in on a browser and the examples you have had relatively small data sets. What about the processor complexity when you have really three data sets which normally is the case for engineers. So the question is what if there's large data and the browser can't handle it. The browser can't handle it. See he's also from leeway. Our typical data that we look at is 200 billion records and about 50 terabytes of data per day. And there is a lot of wealth of information that we can mine and visualize and generate trends. Yeah. Any reasonable humanly possible way to do it today? Absolutely but not on the browser. The question was if we have what 200 billion rows of data what can we do with it. That's nothing on the browser. A 10 second plug here. So the company that I work with, what we do is handle on the server set massively large data sets. The technology there is a custom build server that produces output in the form of HTML JavaScript. This is an example for us to see. This is a network of github works in Chennai. It's a somewhat larger data set and I'll show you some other larger data sets but the way it works is anyone who's following anyone else on github, they're one connected company. Now that's the network in Chennai. I have decided to settle in Bangalore and not exaggerating but truly for this reason. This visualization I mean I'll leave it to sort of implications. Firstly the network is twice as large and the connected component center is also is a significantly larger percentage. But the thing is this sort of thing which involves a reasonably large amount of data can be handled on the browser. I have up to a million rows I don't even think twice about on the browser. It's only when it starts getting to tens of millions of rows that you start worrying. At that point it needs to be done on the server side. But to be fair you have two bottlenecks. It is not the processor capacity offered that constraints you. It is offered the bandwidth. If you have to transfer a million rows it's going to take a long time. You don't want to do it. You want to summarize it and if you're summarizing it then there are a couple of ways of doing it. One is so let us say you want to create a graph like the one that I showed here. Now the number of data points here is massive but you don't need to show all of the data points out there. If you're showing complete the correlations on the back end send the data to the client and represent it on the client. That is what is possible. The other possibility is generate the full existing file on the server side and display. Both options work. We try both no issues. Another area that I thought I share with you is mapping or actually one of the quite interesting part that we shared that first. We would like to see who the fastest one-day class never had and didn't want to take a lot of time doing this. Just want to see the entire history of one-day interactions which incidentally if you take an excel and print out about 150 pages and see what can come out of it. That's roughly what it looks like. The size of each of the boxes represents the number of rounds that players score. The color indicates how fast players score. So can you tell me who is the fastest of you and how long did that take and who scored the most rounds. You don't really need this to say that. But also you can zoom in a little more. So really he's been doing great but let's look at his individual matches. So that was 124 at a whopping 206 strike rate. 143 guys feel I can't. It provides you that build-up capability as well and clicking on it will take you to the match that he was playing. You can see the statistics and so on. Like I said, about 150 pages worth of information that can be compressed into one shot. Again, here absolutely no fancy like grease. Nothing more than plain HTML and JavaScript. Another is maps. So that's a simple map tool where you have every state in India. Let's say I want to change Andhra Pradesh is slightly greenish. Let me make that red and the sand somewhere in that range as well. Let me make that red as well. So I can just put in numbers or copy and paste this into Excel, copy and paste it back and you've got the data plotted all of that. It's not restricted to the state map. You can have the district map or let's go to the districts within a particular state. Now what's happening here? Nothing more than one SVG fight which you can control through JavaScript. On change or on key of whatever you want. Find the item, change it. The possibilities are limitless. Since we don't really have much time for libraries I'll just mention one library that I think stands out about this. We've talked about, you've heard three being mentioned. Raphael Protobis and D3. Protobis successor is D3 and therefore between Protobis and D3 you don't really have a thing Protobis is D3. And between Raphael and D3 also you don't really have a thing Protobis it is D3. There is today one library that is clearly going to win the race for JavaScript at least for the next one year without it. It's a vague my post of the guy who authors it is going. It is already something extraordinary and will continue to be something extraordinary for a long time. I think it will get to the place where jQuery is without any competition. The library's name is D3. The coder network thing that you saw the social network that was done using D3 for example. But you don't necessarily need to use it. There are some complex visualizations that makes life that it makes a lot easier. So one thing more important visualization is if you have static way of visualizing that is if you are always plotting runs against pause phase that wouldn't actually be helpful for decision makers. They would want to move around the axis, move around on what you are looking the data at. So the question was how do we help people who want to play around with the data. Does D3 to a certain extent. But let me answer the questions. First at the business level which is there are two kinds of people. People want to play around with the data and people want to see the results of the data. And the two are completely different audiences. Visualization is something that is a visualization area presentation. Analytics is what helps you find the answer. And the people that are doing these two jobs have a very distinct tools. So your question was will D3 help me explore? Panseries? Yes but I would not suggest it. If you are using SAS or R for instance, you have far more powerful tool to data exploration. If you say you have far more powerful tool for doing data exploration. Once you know that this is roughly what I want, then you spend a few minutes thinking about what is the best way to picture it. Which is where a language like D3 will help. But just have a look at the interactive capabilities at D3. It can do enormously. It can do a lot. Problem is the predecessor to the D3 as well as that one which is interactive visualization. Both of them are offshoots of Stanford University experiments. So you seem to be fairly advanced. You seem to have fairly advanced capabilities in your companies. Do you plan any tools or set of wrappers around the existing tools that would be available to the community? Questions? What open source tools are we going to develop? The India district map that you saw that's the only district map that is available that conforms to the 2011 census. That's open source and like licensed via ESG and like data assembly. And incidentally the WTFBL license for those of you who want that licenses. That is an example. We plan to build a series of tools. Where we're struggling with this to build something like Tableau for instance. Tableau incidentally is an exploratory data visualization tool. You can do a lot of private property, create some really amazing visualizations. The problem with creating something like Tableau is it caters to an audience that has a rough idea of what they want, how to create visualizations and so on. See we found that you can create extraordinary visualizations using Excel. You can create extraordinary visualizations using HTML. The reason why people don't do it is not because the tools give you the reason is because the people are not clear. So in order to create an end user tool that creates great visualizations we have to get over the hurdle of that training we've given up or not given up. That's too ambitious for us to start. We are focusing on small domain specific tools. Map very straightforward. So then we've created a new construct called a caramel map in which let me not go into it. But it's a new type of visualization and that's something we've got to do to pair on that will be live soon. We've got a cluster's catapult which actually you saw the hierarchical visualization cluster this one. Now we've applied it with a good effect on a number of cases. For instance we went to a telecom company and said take all your products. Brought it this way. They said that's interesting. I've got one product at 111 MRP which is very tightly correlated to 899 rupees MRP. Then they think how well it's said of course we'll do R up to 200 as the 200. The guy comes in and gives the 200 and we won't chop people or won't have change. So this makes a lot of sense for them to be confident. So which means that if you push me, if you do any marketing on the 100,000 rupees MRP product it will automatically increase the sales of your 899 rupees MRP product. But at the same time it is conflicting with your 222 rupees MRP product because roughly similar price point, roughly similar proposition. So they would see if you have, if you do any marketing on iROB used to product it's going to kill your cycle larger product. Larger both from a price point perspective as well as from a volume of sales perspective. So that was one example. Another example we took this and analyzed competition. How much of your traffic is going to compete with me versus compete with me and so on. So if your traffic with compete with A rises traffic with compete with B rises and so on. And you can almost see that the world of mobile income is grouped into three types of segments. One, CDMA operators, two, GSM operators and there is one company which I will not name which is bit of a young company which is completely distinct from all of these. This tool really helps to group together various kinds of anything. Here we have grouping securities, you can group products, you can group companies, whatever. So this is another tool that we are trying to get out as an open source. We are doing a bit of it but the thing is our aim is not to create a general purpose data visualization tool that anyone can use. It is to bring more of more confidence in the view. So you have a question. This would have to be the last question for running short time. Okay, when it comes to performance, which festival is BGO or Canvas? Canvas but there are two reasons why you wouldn't want to go for Canvas which is the primary one being A if you try taking a printout, the resolution will fail you and if you try zooming in the resolution will fail you. And also actually thirdly, if you want to interact with it, you can't interact with an element in Canvas as easily as you can. I over over an element, I don't have an object to attach an event handler to to see what that is or could not. Obviously in the SCG has sort of become the de facto standard in terms of guarantee of distribution. So with that question, let me just end with how you can reach me which is on my side here. Just Google SR I will, it's away with lunchtime. I'm going to be around for the rest of the day. Feel free to reach me with all my contact information on my website or right now. Thank you very much.