Ronnie Sadka: Thank you very much. It's a real pleasure being here. It's the first time I'm actually presenting a paper on emojis and to prepare to really do a preparation session with my kids, they can explain me. I understand. Text. I don't really understand all the emojis, but we did that. And I think now I'm better able to help all of us understand more about that world. Just to give you a preview, right after the pandemic began spring 2020, get in. And I and a few colleagues, we wrote a paper. This was back, I guess, summer 2020. We wrote a paper about how read it. So how about Robinhood Traders helped support the equity market? There was no government intervention yet. The market kind of functioned. We had a significant drop and then started increasing. Where did that increase come from? Well, we checked and we we found out that a lot of Robinhood traders were trading the market. And we did an interesting test whereby we looked at states that went under lockdown and states that went under lockdown. The companies that are headquartered in those states, their liquidity improved. So we did some interesting natural like a test, and that suggested that really it was the retail traders that provided liquidity. But in that same paper we wrote in a paragraph that, well, if that's happening, it might be a case where if retail traders would coordinate, they might actually significant move markets up and down, and that actually might present a risk to financial markets.

Ronnie Sadka: So what I'm going to do today is try to help everyone understand. What are the implications of retail traders when they can coordinate? So theoretically, you can think about it in two ways. First, you can say, well, if you think about retail traders as noise traders, standard models of liquidity in financial markets would say, well, the more noise traders are. The more liquid the markets will be because there's more people to trade with. So if you have more, the more uninformed people you have in the market, then price impacts and liquidity prices will go down, liquidity will improve. Of course, the other side of this is if everyone informed, no one wants to trade with each other because you always think the other one next to me, the other party knows better than me. So if there's enough noise traders that improve the liquidity in financial markets. However, in this case with retail traders there, we might think that they're uninformed, especially when we look at some of their portfolios. If they coordinate, then suddenly they can move markets together. So what I'm going to show you in this paper, in this presentation is that that indeed happens when retail traders coordinate. They move liquidity in the market and in fact, they can actually take away liquidity. And the reason that's so important is because you see what happened after the GM frenzy. I'm going to talk about that a little bit. Suddenly you're an asset manager and you see this craziness happening in the market.

Ronnie Sadka: You don't understand it. Asset value goes up thousands, hundreds and hundreds percent. And during the course of a week, it can go up 1,000% going down. You don't know what's going on, what's going to happen. You're going to say, I'm not going to trade the stock. I'm just going to stay out. If enough institutional investors are going to stay out of the market for that stock, what's going to happen is going to be less incentive to produce information. So there's less incentive to produce information. There's going to be more information in between the stock. The liquidity is going to drop. So you could have a situation in equilibrium. We have more retail traders trade then our mutual fund managers. Other hedge fund managers say, you know, I'm out of this too risky and suddenly you have a stock, could be a large cap firm, but with less liquidity. I do think it's kind of interesting when you think about short sellers. There used to be this discussion of short sellers against CEOs. When the CEO, it has a lot of short seller activity. It's like the battle of short sellers, no better financial accounting. They check and they say, oh, you know, some issue with this firm. And then there's always a battle communication between CEOs and and and and short sellers. But now actually, it seems that there's some kind of battle between CEOs and retail traders and CEOs try to cater to Reddit, to Redditors, and then the Redditors win with the GME crisis.

Ronnie Sadka: Well, Redditors were actually against Wall Street or against Melvin Capital. Right. There's a lot of rhetoric about, oh, they should go down, down and let's trade the stock up to short squeeze them. Right? So there's all these effects in the market. I think it's very interesting both from risk management and investment perspective. So what are we going to do in this paper? We're going to look at the Wall Street bets subreddit. I'm sure a lot of you have heard of that. We actually have an agreement with with Reddit that we can drink their data and provide some indicators. So we we essentially every day get all the subreddits, the posting, the submit the submissions and comments for all in all of Wall Street bets. And I'm going to show you I'm going to create some indicators and I'm going to show you how to impact liquidity, how they impact price behavior around earning announcements. I'm going to show you some implications for asset pricing anomalies like momentum value, etc., and also show you some impact and firm value. The conclusion is going to be that the more social discussion you have, there's less incentives to produce information, and therefore it's going to reduce the liquidity of the asset. And I'm going to show you some implication for short squeeze, etc.. So let's look at the data. This is the number of users subscribers to the Wallstreetbets subreddit.

Ronnie Sadka: You can see the jump around January 2021. That's when it happened. The GME frenzy, significant. I mean, it's pretty clear you don't need to test this. There's a jump we're going to start we start looking at this data from about, I would say, October 2020. And we're going to look at we're going to look at the implications until until today. Now, what's so special about I just want to mention the Reddit data is a bit different than other social media like Seeking Alpha and others, because with Reddit data, it's like minded people. They just talk about what they want to do. There's no credibility stuff. There's no even Seeking Alpha. You need to be a contributor. You need to go through some a process. Not everyone can just say whatever they want. So this is a very special type of social media and what we're going to do and I'm going to use this data to provide indicators refreshed daily, and I'm going to show you what they mean. If you take a look, a macro look at the data, if you're thinking about, well, how many firms are actually mentioned on social media. So when you look at, let's say the Russell three, you get about 10% of the firms on any given day are mentioned in in on Reddit. So it's about 300 firms per day on average. You can see how it went up a bit. You can look at the the blue line went up around 21 and then it's now been steadied about 300 firms per day.

Ronnie Sadka: It's not the same firms every time, but it's about 300 per day. And when you look at the number of items we see per firm on average per firm, that is that it has some intensity, some mentioning in social media, you can see that it's above on average about 50, 60 posts per day. That's a very there's a huge skewness in the distribution. Right. So when the GME frenzy happened overall, there was about 1.5 million items that that week. When you look at the number of submissions now, it's about this last week I checked it was 33,000 submissions overall in the sample period is about the median is about 40,000 and the average is 60 per day. So so that just tells you that that it's skewed. The minimum point was during this time period was 10,000, 10,000 items. When you look at I looked at this yesterday. So as of yesterday, the highest social activity for firms. You see Microsoft was number one. I just had a brief lunch now and with a few Microsoft, Microsoft folks they mentioned, of course, the earning announcements was just happened. Google after that Tesla, Apple met us. You see, it's a lot of big firms, some small firms as well. But you have, you know, Coinbase, Boeing, they're all were able to capture all of that. One thing that was interesting in this project and Michael started that with the with the emojis, is a lot of things we've done.

Ronnie Sadka: We've been partnerships with Partners with States, which for eight years now, we've done a lot of NLP natural language processing for this. You need to do emoji, you got to do emoji emojis, emoji processing. So a lot of the results I'm going to show you are just going to be based on intensity. So is a firm mentioned, but at the end I'm going to show what happens if we just count emojis and the results are far stronger. So if you just look at the number of yellows and rockets, all that stuff, we're going to show you that the results are much more significant. All right. Let's take a closer look at GameStop just to give us. I want you to feel more comfortable with the data because after this, it's going to be more academic in terms of let's just run regressions the entire sample, but let's just get a feel of the data. What happened with GMI. So look at the top graph. The top graph shows you abnormal social intensity. So what does that mean, Abnormal social intensity? You look at the number of posts on a on we have it every day but look at a given firm check. The number of posts for that same firm. On the same day of the week, last week, the week before, the week before. So check the last four weeks.

Ronnie Sadka: What's the average posting on that day and that day of the week for the firm? And look at the difference between today's posting and the average of the last same day of the week for over the last four weeks. It's just to get an idea of sometimes the seasonality we uncovered in the data. So we're just trying to correct for that. Also, there's different levels of mentioning for each firm, so we're trying to correct for different levels and different days. So wait, do that and then you can look at the daily changes. In in the what we call abnormal intensity. So you can see the blue line just shows you how the significant increase in intensity for GMI and then it kind of drops and then now it's kind of level off. When you look at bit at spreads, what happens to the effective spread for Jimmy? You see that the pattern mirrors the pattern in social intensity. So our claim is going to be that the more social discussion actually took away liquidity. When you look at the bottom graph, I wanted to show that it's not just happening during that first quarter of 2021 when you look at the correlation. So you run a ten day correlation of spread and and social intensity. You can see each point here is every every day you look at the last ten days. So it's like a rolling thing. And you can see we did it weekly and you can see that most of these dots are positive.

Ronnie Sadka: So it's not just happening during the period of the first quarter, it's happening throughout. It's true that variations are lower, but the correlation is still there. It's pretty actually pretty stable. Okay, So now to more formally test this. What are we going to do? We're going to run a regression. We're going to run a regression using daily, daily data. You can see the number of days and firms at 900,000. So you run a regression, take the changes in the spread or the effective spread, and run a regression on changes in social intensity. And this again, we're doing it day firm. We're pulling everything together, but we're putting a firm fixed effect and a date fixed effect. And when you do that, the first column, you can see that in social intensity is positive. So this is significant. Square brackets are the statistics far higher than two, suggesting that there's something there. When you look at the third, the column number one in panel B, that's also significant. One thing we thought of, and we're going to talk to you more about this in a few slides, what happens around earnings announcement? Because that's when information comes out. Well, it turns out when you add in earnings announcement, dummy, you see indeed, that better spreads. And price impacts are higher, so there's less liquidity during during these times. We know there's a lot of trading going on, but information asymmetry, it's been well known that liquidity dries up around that time.

Ronnie Sadka: However, when you interact earnings dummy with social intensity, you see that that's positive and significant. So you see that that social intensity contributes to the drop in liquidity around earnings announcement. Okay. Now, let's try to understand a little bit more about the price informative ness. I'm going to try to make an argument that the prices of securities are less informative about fundamentals when there's more social discussion. Okay, so you think about it this way. You see the price of an asset, but there's a lot of social rockets flying around, moon yellows drop by. You don't know what's going on. What does it mean? You know what's going on. The price might be less informative. The price might reflect some more emotional behavioral of investors, but maybe less of discounted future cash flows. So there's different ways of measuring that. It's not easy to measure price informative ness because we don't have the right benchmark. So what we're going to do is we're going to do a few things that have been exercised in the literature and call them one. I'm going to look at abnormal turnover. So we're going to look at the daily. We're going to look at at earning announcement and we're going to look at the month, the 22 days before earnings announcement. We're going to look at two things. We look at the abnormal daily turnover and we're going to look at the return, the return drift before the announcement.

Ronnie Sadka: The idea is if there's more turnover around the announcement, there's more price informative earnings. Why? Because it suggests that people are trying to understand what's going on in the market. There's more trading. They're trying to produce information. So more trading suggests there's more information if there's more drift of price before an announcement also shows this is more information. It means there's people are trying to understand what's going to happen in earnings announcements. So part of the drift happens before part of the there's an announcement return, but there's a drift before that. So the more the drift before that suggested that there's more information that the price is, the more informative. So both of them are dropped, the reduced with social intensity in the last month. So if you look at the assets that have more social discussion before the earnings, have less turnover before earnings and they have less price drift before the earnings, suggesting that the price that the price informative before the earnings is low. So now let's check what happens during earnings announcements. So during earnings announcement, we're going to call we're going to calculate what's called in the in the literature, the earnings response coefficient. It's very simple. You look at the earnings announcement, you look what happens to the return around the -1 to 0, earning the two trading days around the earnings announcement and you run a regression on the earnings announced. With respect the difference between that and analyst median forecast.

Ronnie Sadka: So for every what we call here you a u. E is unexpected earnings. Unexpected earnings again is the announced earnings minus analyst median expectation. The you have to scale this we divide by previous quarter price. When you run a regression in the literature, people have shown me on the regression of returns in that announcement on unexpected earnings, the coefficient is positive. That's good news for the market because if there's if if the firm beat analyst expectations, the price goes up. It's like a measure of market efficiency. So let's suggest that the market is functioning well. What I'm showing you here, if you add in interaction with social intensity, that's positive and significant. What does that mean? It means firms that have more social media activity prior to announcement. Okay. Their price responds more during earnings announcements. To any outperformance in earnings that suggests that there's less price informative ness before the earnings announcement. Me Let me repeat this. If prices were informed, then you would see some activity before the announcement. People say, well, analysts have their conjectures. If the firm outperform the analyst, the price will go up, but it will go up even more. If people didn't trade before the announcement, some information is leaked or some information is provided to investors. Between earnings announcement periods, it's not that all the information comes out just during earnings announcements. What I'm saying here is that for firms that there's more social activity, it seems that more of the information comes out around earning announcement.

Ronnie Sadka: It's as if the the process of producing information prior to earning announcements is not that strong. It's consistent with the idea that investor or that that Redditors are producing their their their discussion. But when that happens, you put a higher higher the signal to noise ratio on these firms is in a way that you just kind of less believe this. You put a higher sigma on the signal. And so therefore when the company really announces what's happening, that's when most people trade and the prices reflect the information. So this this positive coefficient here, more more earnings response coefficient during announcement times suggest that prior to announcement time the price was less informative. So let's be a little bit more clever about this. What do I mean? One could argue that maybe what's happening is these firms, that they have less that they have higher earning response coefficient during announcement. These are particular firms that there's something about them and that's why Redditors talk about them. I'm trying to make the argument that what Redditors are doing, they're causing. Less price informativeness. One could argue the reverse causality, that there's something about these firms that attracts Redditors. And so something else is happening. It just is Redditors. Talk about these firms. Does that make sense to everyone? I'm just trying to tackle a potential reverse causality. So I think so. Here's the test I've done. It's a little bit different because I don't have the data from starting 2001 and we have 21.

Ronnie Sadka: But what I thought about doing is the following let's run the RC test in 2021, but I'll run it for two different groups. I'll take the firms that I call poplar stocks. These are going to be the top 10% of firms in the top 10% of social intensity. And I'll take all the rest. These are the neglected firms. And I'm going to run the earning response coefficient on these two separate sets. The results I've shown before are the results pertaining to all these quarters in 2021, Q1, Q2, Q3, all the results have shown you this is consistent with that, right? The firms that are more popular, their irks is higher. Then I go back in time. I still choose the same firms, whether they're high social discussion. I call them popular and I look at the pre 2001 area 2021 period in 2020 using the same classification used for 2021. And I see that there is no difference between the arcs. You see, if there is something about these firms that attract Redditors, then I would expect to see differences in earnings before 2021, before Reddit was popular. But what I find is before Reddit was popular, there was no difference between the two groups. So it's must I won't I'm not going to use must be. It's more likely that Redditors are those that caused the I.R.S. to be difference between the two groups.

Ronnie Sadka: It's kind of a quasi placebo test, if you will. Okay. So I'm more convinced, at least myself, I'm more convinced that it's really the Redditors that are causing this. So I want to show you also some what I think are practical applications. So what I've done here is I looked at some asset pricing and what are asset pricing anomalies. When I graduated from grad school, there were maybe four values. Size, momentum. Remember all of that? Now we call it the Factor Zoo. There's 400 anomalies. Rob Stambaugh from Wharton published a paper, did it for a few years and looking at 11 anomalies, including the way we talk about the value and the size and momentum and quality investments. There's a few, and then you create it for each firm. He took 11 anomalies for each firm. He created this. He created a ranking for each of these anomalies with which each firm's sum them up. And then he has an overall score for per firm across all 11 anomalies that will indicate whether the firm is overpriced or underpriced. And the anomaly just shows you that the firms that are overpriced tend to underperform in the future and the underpriced tend to overperform, hence the spread between underpriced and overpriced assets. So what I did here. As I examined the impact of social media on the spread of overpriced underpriced. And my intuition would be firms that are more in the social, in social discussion. These are firms that the price might not represent or its less informative less.

Ronnie Sadka: What happens for these firms? It takes longer for the anomalies to converge. So if you think about take momentum, if you think about a firm that is, let's say in the high, let's say let's say a high momentum firm. Well, no. Let's take the low momentum firm losing stock. We expect this to drop more in the future. What we're going to show that losing stocks that have more social and more social media. What happens? Those don't drop as fast. So there's actually a spread between the two. So you can run a strategy of comparing overpriced assets across high and low social media. And that in and of itself will give you a spread. I'm going to show you that strategy. There's two there's a few indications for that. First, in the bottom table. I run Simple Pharma Macbeth regressions. These are the standard cross-sectional regressions. When I look at firm returns in the future and I run that regression on the scores standby scores for overpricing on the pricing, I do it in a weekly basis and I'm looking at it's like an event. I'm looking at every given point. I'm looking at what happens next week, the two weeks after three up to 12 weeks, so up to three months after. So in the table, what you can see is there's the overpricing score, which is you can see it's negative and for some of these is statistically significant.

Ronnie Sadka: Again, this is a one year plus a few months of data. But the and so we expect the overpricing score to have a negative performance in the future. But the point is, if you interact that with social intensity, you see a positive coefficient. What does that mean? Positive coefficient that was more social intensity. The overpricing coefficient will be lower in absolute value. So it's negative, but it's going to become less negative when there's more social media. So when you plot that bottom row I plotted above, you can see how that suggests that at the beginning there's less trading of these overpriced assets and then the market kind of catches up within a few a couple of months, the month the market catches up. So you can you can look at this in terms of just portfolios. And if you double sort portfolios according to overpricing and high, low social intensity, you can see that, for example, for overpriced securities, the high social media stocks significantly outperformed the low social media stocks. And you can see essentially those coefficients of showing the best regression are really those in the bottom right corner. It's like this the spread between high, the spread between overpriced and underpriced for the different subsets of high and low social intensity creates a spread of about 2626 pips. This is going to be in percentage this 26 pips per week. So the amounts are about 1% per month. If you run a momentum strategy on low social social media versus moment strategy and high social media, you're going to see that that spread.

Ronnie Sadka: So if you create that as an investment strategy, so take the the high social, high social intensity versus low social intensity only among the overpriced securities, you're going to find this graph. Again, this is a short time period. By no way am I selling this as a strategy. This is just to show you that it does seem to be the case that there's some performance here, that social the social media helps us understand among the overpriced and underpriced security. It helps us differentiate between them. You even get about a sharp of 1.5, even for talking about a time period of this is like 15 months. Okay. Now let's talk a little bit about evaluation in this literature evaluation. What a lot of people do is to say that prices are informative. What they do is they regress investment and Tobin skews Tobin skew is a measure of like a replacement cost is like a value of a firm. Normally when you run these regressions you find the coefficient here that 0.7 and in the column one at the top that's the typical coefficient. So investment insensitive to firm value. So that suggests that prices are informative. That's what in this literature, when you interact that with social intensity, you see it's a negative coefficient. So the sensitivity of investment to value is lower for firms that have a lot of social media.

Ronnie Sadka: Again, so this is suggestive that again, that there's less that these prices are less informative when in the presence of a lot of chatter in on Reddit. Now you can ask what about traditional media? Unless eight years of only getting traditional media across the world, many different sources. Well, what happens here? For example, you take Dow Jones. Take the same methodology we've done for social media, but do the same thing with doubt. So take a look at the Wall Street Journal, the veterans, whatever they all the media of traditional media. Dow Jones. Check the firms that are mentioned there create intensity measures and then instead of social intensity non stat add also Dow Jones intensity. For these firms and what do you see? So I'm doing here the IRC, the earnings response coefficient test coefficient, the first one above, that's just the earning response coefficient. We know it's significant positive. The question is does it change when you interact with the different types of media? So when you interacted with social media, we know it's positive and significant. But when you look at the Dow Jones, there's virtually nothing. So that suggests that our discussion about past informative ness really comes from this social media. And the idea I get here is that they talk about a lot about stuff, but it's not necessarily informative. So the fact that they're chatting more about the firm just creates more of a cloud. You're not sure what the prices mean.

Ronnie Sadka: It is not. The same effect is when firms are mentioned by traditional media. And I think it makes sense because you think about traditional media, they have some analysis behind it. Probably reporters that have done it for a while is trying to be more informative. But with Redditors, it's about rocket. Rocket, it's much more about feelings. So to me, that's more of a suggestive of a price distortion, which we can we can take advantage of, or we should at least be acknowledged. Acknowledge that. So now I promise you that what happens, I'm going to look at what happens specifically when look about look at emojis, and we track a lot of these emojis. So there's emojis for crashes and for lost moon, rocket squeezes, YOLO. And when we look at instead of looking at the number of posts or the intensity of postings on Reddit, we just look at the intensity of emojis on Reddit per firm. The results are far more significant. The key stat of 21. So you got a far more I mean, so this really shows you this idea of this people are just, you know, posting and rockets. I mean, that's really creates a situation in the market where it could be that just a lot of funds are pulling back. If they're pulling out of that, who's left to trade this? And then the prices are less informative. So this is what I want to talk to you about, the paper and the idea of price informative ness. But we created.

Ronnie Sadka: Many, many indicators that are provided to you through our partnership with State Street. And this just shows you an example of a dashboard dashboard we built with a ton of different stuff. I mean, you've seen the top social activity on the left. You see that. But we have top squeeze insensitive. What are the firms that have the most intense discussion on They're being squeezed. And then when you look at, for example, the squeeze pressure. So if you look at the middle middle panel all the way to the right, I really like this one. Why? Because on the vertical axis, you have the social intensity squeeze, the intensity of squeeze. But on the on the horizontal axis is the is the actual short float. So to me, these firms that have a lot of short activity and they're also there's a lot of discussion about short squeeze, these are the problematic ones. These are risky. And there's quite a there's a few of those there. So that could give you a measure to give you an alert that, hey, this firm that is on short now. Well, you know what? A lot of short interest, but a lot of Redditors are interested in this Alert. Alert. This might be a short squeeze. So maybe you want to be worried about that and maybe you stay away. Most of the time, they're not going to be mentioned in social media, but when they are, it could be a big problem. So that's why this is very useful. And again, there's other panels here. There's there's a there was a discussion tomorrow. I'm a co-founder. And Rajeev, we're going to talk Gideon and Rajeev, we're going to talk about narratives so we can look at narratives and social media and what people are talking about.

Ronnie Sadka: And you can see their a list of things in the doughnut and the on the bottom right dump and drop and crash all that stuff. So let me let me conclude. I have to say, it's been very exciting to work on this project. It's a kind of a different thing for me. It's a different technology. I'm kind of older now. I don't know what it's talking about, but it's all these emojis. And, you know, you take a look at the social media, the type of stuff they write. They're just in the words. It's a very colorful I'm not allowed to talk about it here. So it made me, you know, I can't say the words they use in social media, but the emojis they can use. Right. And I thought this was just very insightful to see how many people are really doing that and posting all the time and commenting and re commenting. So my my conclusion here is I do think that intense social media hinders is informative. So people, when they see a lot of social discussion, perhaps institutional managers just pull back and then the whoever's left is creating a situation where there's less liquidity, there's less, the price is less informative, the less reflect what we think as traditional pricing is discounted future cash flows. So therefore, it's I think it's very important to try to understand why these discussions emerge. But what we did here is really provide you some tools and indicators that could help you mitigate the risk of suddenly being significantly mentioned in social media and what that means for your asset returns. And that also could give you some indicators for potential investment. Happy to take any questions. Oh, I think a proper clap.

Michael Metcalfe: Your’re actually right. Do you want to take a seat? I've got a I've got I've got a few questions here, but very, very happy to. Very happy to take any questions from the audience first, if they're if there are any. We have one here left. Thanks for the presentation. I was turning in your analysis. You look particularly at the Wall Street best section of Reddit, but as you get your splinter of communities off of Wall Street, best also discuss, you know, meme stocks a similar way as Wall Street bets. Have you guys looked at incorporating those communities into DSA as well?

Ronnie Sadka: We haven't looked. We haven't looked at those communities for this. What we have done is look at some other communities when we think about politics. So we've been supporting efforts around, you know, the geopolitical what's happening in the US and the midterm elections coming up. And we do find some really, really interesting results. So for this analysis, we really focused on the Wall Street bets. But but the point is that now we have the technology of identifying, you know, alerts for other types of interesting events that happened in financial markets. And definitely that's doable.

Michael Metcalfe: Yeah. Thank you. So we've got a really good question actually coming through the app, which is a very funny one, which I'll touch on as well. But so from a sociological perspective, what may look like a loosely unorganized mob may actually be led by leadership who may in fact be more structures than is apparent. Does your analysis suggest any detectable organization?

Ronnie Sadka: We thought about that. We I don't have a good answer. I don't. It could be we thought, well, maybe, maybe. Be careful. I mean, I'm being recorded. Some people in some firms might have the incentive to provide, you know, to to to put a lot of emojis on or to comment and to create some kind of coordinated trading. You can think of, like, just like the flash trades in high frequency, maybe something like that is going on. I wasn't able to detect it yet. It's possible. I understand the incentive, but I can't. Any comment, anything about that yet?

Michael Metcalfe: Okay. And this might be a question for tomorrow's talk, actually. But why don't just throw it out here as well? Are you able to get to sentiment through emojis or just the intensity of social activity?

Ronnie Sadka: Yeah. So I showed you just intensity, because I think intensity can be measured very clearly. It's whether you're you're are you are you mentioned or not mentioned or how many times you mentioned. So that's for us. It's just like intensity in the media. In the traditional media, we use that as a first case. We also did some analysis looking more particularly in the sentiment. So you can look at sentiment of positive emoji sentiment versus negative emotive sentiment. And you can also look the the regular NLP analysis on the postings. So we've done that and it's quite, quite interesting. The the negative sentiment is a little bit more indicative. So we have done that. It's not in the presentation now, but it's definitely doable. We've done it and I think there's going to be more to come on that.

Michael Metcalfe: Okay, brilliant. Well, unfortunately, we're out of time. I do just want to share with you this one last question. I'm going to have to show it to you. You can't get it on the camera, but it's the emoji with the glasses followed by a thumbs up and a line going up. I think that's the speaker feedback. Right. But anyway, thank you again, guys.

Ronnie Sadka: Thank you.