From Charts to Decisions: When Onchain Data Actually Matters | Danning, Dingyue, James, Arnaud

Morning everyone. Hope you're doing well. Uh feels like a proper podcast studio here. And uh today we're excited to talk about onchain data. Uh my name is Arno.

I'm with Dune uh previously known as Dune Analytics, not to be confused by the movie. We have a leading onchain data platforms. We work with crypto companies, institution, government who need access to onchain data. Today I have an amazing friends and panelists. They are all expert in the field.

I was on a panel yesterday. Some people spend 10 minutes to introduce themselves. So I'm going to introduce you so we can jump straight to the topic. We have James who lead onchain data at Coinbase recently acquired from the spindle team. We have Danning who lead research at Panta Capital.

And then lastly we have kite who is the research and data at unis swap. So no need for introduction for those companies. And what we want to talk about today is really the dynamic between how to make decision with onchain data. You know most of the companies are uh data rich but often decision or information poor. So uh I'll start with you uh kite.

So I know you come from the academia world and what was the moment where you know you decided to move onchain and I know you've been some some interesting thing at at unis swap when it comes to uh user being exposed to sandwich attack and how those onchain data have drove a decision to make those product change. Um right right so the moment that I realized I want to do this on chain is ex exactly what you said that we have tons of data and as as a grad student um you were found getting data is so hard uh you either have to foyer uh and then wait for a very long process to get the data or you have to collect your own data or you have to pay tons of money to get data um versus if you look at all the DeFi data that we have um you can query all of them on dune um and you can just do your analysis and start researching stuff. So I I think that's the moment I was like I I really want to do this. I really want to do analysis in these onchain environments. Um right and about the default paper.

Um so so it it's been historically known in economics that the default option really shapes people's behavior. Um people really overwhelmingly kind of stick with the default options. You know from 401k uh enrollment uh to to menu design. So um the DeFi has no exceptions. Um so so we have these like slippage protection for example that there's a default option people really often stick to.

You don't want to set them too high such that you are exposed to uh Mav attack. You don't want to set them too low such that you have your transaction reverted um and and you have to pay gas which is very annoying. So we we were looking at this data and was like can we improve this from the onchain data that we observe uh from the gas that we see from the transaction size. So over the years we have iterated more more and more versions of this and then try to look at how can we better have our user protected from these onchain data

and I know pivoting for for you Danning. So I know you can rant on onchain data for very long time. I saw your podcast amazing with Il Dobby and and Boxer and from a from a decision-making standpoint like I'm curious to see have you heard a pitch from a company that tells a very different story from what you saw when you're looking at the data.

Yeah, I I think I'll give an example of like a whole sector maybe not exactly just one pitch but like uh this actually also mentioned at the data podcast. Um basically like deepen as a category was kind of like popular for a while. It was maybe a little bit overhyped but then it's so dead right now like the market cap of the whole category has dropped like by 99%. But what people haven't noticed if they checked on chain is that like their revenue generation actually went from less than 1 million to over 70 million among the major company for last year and a lot of them are moving those like accounting of revenue on chain as well like they're logging them on chain. So which is awesome change but like we're seeing then now like some investment are again going back to deep end and getting some interest.

So it's like kind of like a cycle of like you get informed by onchain metrics and you realize a lot of stuff are underpriced. So

interesting and I know for you James um so you work a little bit on a very unique problem to solve which is attribution right and how to work with um ads. maybe talk a little bit about that concept maybe for the audience who might not be familiar of Spindle obviously now part of of Coinbase and curious like how you working on that um from an onchain attribution perspective

yeah thanks Arno so uh maybe to set the stage let me uh lead with the spindle pitch which we've been building this product for over three years uh the inherent uh problem with web two uh advertising use uh to crypto companies is that it doesn't have the full picture of the user funnel. For example, you might know which user click on your ad, but you need the onchain data side to complete the picture of who actually completed deposit, who actually have the swap. All the data doesn't live on users uh their cookie, but they live on a blockchain. And what Spindle does is we establish uh this two sides data collection process. We have the SDK for the publisher and advertiser to send us their web two events.

And then we have the blockchain ingestion engine to get the contract conversion events uh to get all the onchain conversion uh from the data. So some examples that we are really proud of Moro uh they really love us to run all kinds of campaigns through all the publishers that we have. They actually found that using our channel the ads perform sixx better in conversion rate compared to like a Google and Twitter. And because we treat wallet address as first party citizen, we know exactly uh what's happening for the wallet, how much uh how likely it is to convert and therefore we can better target and personalize the ads for them. GMX, a per DEX who has a lot of oning volume use to drive their uh trading competition and as a as a result with only 12K of spend they generate $25 million of oning uh volume and that's like $1 spent to get two over 2K of volume.

the the fact that we can even speak to this ad these stats shows like how we are completing this onchain data funnel and why uh this is a great solution for the unique problem of web three um token ad uh web three ads

yeah trading wallet at first class citizen I think it's a great way to put it um Danning you put a a tweet obviously you the Super Bowl was not that long ago I know you've been diving in prediction market in that oh yeah

and maybe you and talk to us a little bit about what have you learned and maybe what people um assume that may be wrong when they think about prediction market and onchain data.

Yeah. Um

bring the mic a little bit closer because I think

yeah there's actually two parts of this analysis. Um the first preliminary one I did was basically pull all the trades on poly market from dune it's on polygon um to look at like how does the volume accumulate over time for different type of predition markets because like going into this analysis like I feel like I was very you have to be very aware that predition markets host a lot of type of markets that are very different in terms of like how the information accumulates and aggregates. So for example, if it's like a political news or like Trump uh versus like sports betting, they they show very different like behavior in terms of like pattern. If you look on chain, um you see that like sports betting um bets or volume are coming in like steadily over time because people already betting on the historical like game information, knowing how these two team are comparing to each other. But then for the news one, it's so unpredictable or uncertain till the very last minute or any like news come out.

So you see very spiky and also like volume accumulates in the very last minute to arrive. Um so that was one insight that was very interesting. Um the other thing about Super Bowl particularly was we also then pulled data from Koshi um on their API to compare these two major two platform across all the NFL uh basically football games. And we saw that like u Kawoshi's trades actually leads by a median of 7 seconds compared to poly market. So like if you compare the uh price timing with ESPN like the sports uh timing number uh time stamp you see that like Khi's price like instantly reflect after like a scoring or like a touchdown or something but um poly market actually like lags a little bit and then you dig further you realize actually there's like a speed bump for poly market.

So if you're user there there's three seconds delay for it. So it's like super interesting like we looked into data first and then we realized a lot of like design of the markets are different.

Yeah.

Very interesting. If you are into prediction market go talk to uh dadding she can help you with your next bet. I guess that's the point.

Love to.

So kite you talked a little bit about split uh slippage earlier today. Can you give us maybe you know unis swap it's all about improving the protocol right constantly. Maybe can you give us another example where you leverage onchain data to let maybe like a new product design?

Um sure. Yeah. Um I I guess one of the example is if we look at stable stable coin markets. Um at first first glance it's really easy market if you to think about it. They're they're packag one to one and they you just narrow your range and then you're you have a a narrow range and then you you'll be able to maximize your um capital efficiency.

Um but if you really dive into on the onchain data of these trans transactions, you actually see a lot of like toxic flow. You'll actually see a lot of sandwich activities happening. Um so all of those what helps us better understand the market micro structures and then such how we design different hooks for potentially different usage of um uh the unisop liquidity pools.

Interesting. Um I know James you guys work on the builder code initiative you were telling me and the new standard the ERC 8021 right and I always thought like air airdrops has a pretty terrible uh attribution you it's kind of like spray and pray and hope that that it works how guys are you guys thinking at bay about attribution investment how you using the data perhaps differently um you know what's farming versus actual sustained growth Yes. Uh airdrops and all this uh base is exploring the launch of a token. It's a very interesting data problem and we're we're very glad the spindle team is uh leading the base growth team to uh investigate this initiative and uh we see this inherently as a data problem because uh when we you know look at the history of the airdrops and uh the token the chain grants that has used like reward incentives they rarely works. Most of them they can't really say hey uh this is the amount of money we spend this is the ROI return of investment and it's it's hard to get a positive ROI or even say what is the ROI is uh and we see that because uh we see that inherently as a data lacking problem and this is what builder code and ERC821 is trying to solve is trying to write more of the standardize offchain offchain data to standardize standardize it onchain so uh currently if you if a user does something on base maybe a swap or deposit on morpho you can see user interact with the unis swap contract or the moro pool contract right but you can't really know which wallet or which uh platform which user interface is the is driving this user is funing this user to go there and we want to know this because we want to reward such good behavior and penalize all the farming all the civil all the boss activity that is not really generate inherent value to chain but those good behaviors we want to reward by having this observability it is very important and this is what builder code trying to do it is a suffix uh written appended on the call data of each each transaction by getting this adopted to all the all the DAPs uh they able to say hey a unis swap wallet drive the user to conduct this transaction uh more for a gauntlet or coinbased wallet or a mini app inside the base have is driving this user to do this deposit, do this action and therefore we can more precisely target and um give outreward to either the user or the project and therefore complete this entire data flow funnel and as I said bring more the offchain uh very unstandardized using a lot of doom wizardry hackery data to actually a standardized way to get it on chain.

you can see it. Everyone can use the the call data suffix to to get that. So yeah, this is the initiative we try to push and we're trying to build the in data infra so that we can have a complete picture to run those campaigns and uh and potentially uh conduct a really good airdrop in the in the future.

I think this this is fascinating because I feel like in the past the only way to mark basically front end entry point is router. Yes. So like you can mark label all these router and say like this is from uh this DEX front end this is from this like maybe like wallet swap but like there's this problem who can never answer which is how much market share each wallet has but then now if you have a like a label in the transaction in the call data then you know

yeah exactly let me give you example uh our one of our advertiser cow swap wants to only pay out their ads conversion if the user is completing the cow swap transaction. in their cowat wallet and have their mu sound right everybody knows that but it's the data is not there on chain you cannot find the data on chain but uh builder code provides a perfect solution for that by append uh if their smart contract team which we are working on which we're talking to uh is willing to write that uh data suffix in the cloud data will be able to have the entire picture and know exactly how much volume they drive uh to the chain to the ecosystem uh not just ads but all all their organic growth campaigns as well.

Interesting. You were talking about some data that might not be available and I want to have your perspective on kind of a data gap like if you have a magic wand and says you know I want this data available today but it's not because it would really help me answer a question like what would that type of data be? Maybe Danny you can start.

Yeah like the wallet labeling of the that would be one good example. I feel like also this really hard problem of like how to label all the transactions across chain. Now there's all these interop and like bridge and like uh intense protocol that's going on and I know health is trying to build data set on that. Um um also there's actually this data set I think maybe a little bit more closer to the liquid team like the trading team for a VC you know like they asked the question about like looking at the token token price like we don't want to talk about price but still you if you look at the token price you pretty much know like when the price crash it's because there's an unlock of a in like a vesting of a investor. So like all these like vesting schedule and also how much distribution all these doesn't exist on chain and if there they can be clearly labeled I think will give also a much better picture of how like all these token performance are going on and why

how about you kite any any question you've been dying to answer and either you can or the data is not there like

I think if you ask any economist the answer will be I want the counterfactual I want as so so in the data we only observe what happened but we know what would have happened if it goes to a different route if it goes to a different wallet if it goes through a different pool. Um I think that will be huge unlock if we were to able to observe the counterfactuals of things we it will allow us to do a lot more casual inferences.

So it's like can give us a lot of benchmarking analysis I see.

Exactly. Yeah. The economist side of kite is building a lot of control or not control but natural experiment. Uh yeah uh everyone read the unis swap paper on the fee switch that's that's so good

exactly

and and danning from an investment perspective like how do you separate the real growth from like

farm metrics I think everybody asked me when I work at dune like what's the top metrics that you think and what you should not think about I think like

what's your perspective on what really matters and how do you separate the noise when you make an investment on panta

yeah I think like maybe a few approach here which is like we all know like there's token an incentive or like people are farming and so usually the protocol they try to raise money when they launch the token or right after like when their performance like TVL looks good and but we also know that it's probably not organic and um if you can maybe look at like the volume before and after the incentive stop right like whenever the time step like snapshot was taken by the airdrop like timeline after that how does the volume look like there's many other metrics you can try to wash the the bubbly part off which is for example like I remember there was a lot of per decks coming up and uh they have crazy volume but like if you look at the ratio of open interest versus volume then you realize like actually there's not much capital actually locked in this pro protocol they're just washing trading back and forth um I think often time it's a shame but we also fall back to like web two metrics which is like if you're building a consumer metrics how does your actual like front- end activity active user are but also you don't want to just look at user who are you know looking at the website but also more have like a action of querying something or like transacting something. So act monthly active user who is transacting something like that. Um but yeah I think like we all are constantly battling or fighting against basically the artificial int uh incentives. Yeah. So we talk about a lot of topic we talk about airdrop prediction market investment uh attribution.

Um maybe curious to hear from you James as you work you know Coinbase is you know clearly a blue chip across literally every product line probably in crypto and beyond like do you think there's an area of like onchain data that is a little bit underappreciated right now like what are people are not talking about and they should be because I think at Coinbase you really have the 360 like what would be that area that people need to pay a little bit more attention to now.

Yeah I think

you cannot say spindle or ads. Okay.

I I I I think the overall strategy of the the base growth team is to have this complete picture of offchain data, onchain data, we all treat them as events, right? uh inherently there's no difference between a user who click a button and do a transaction on chain versus user click a button uh of maybe ads maybe other things or in incentive campaigns on other platform surface of coinbase retail app or the base app right so we all treat them as events I think the mission uh of uh our growth team is to trying to complete this entire funnel of the user journey right now it's very interesting that we and deduce so much information from do analytics. We have done so many different tricks in using the router using the fe using all those onchain sloohs to get maybe 70 80% there in terms of understanding what the user actually does but without the offchain piece I feel like this is still a incomplete um cycle of the user journey and therefore it's hard to measure hard to reward hard to really do uh repeatable experiment on top of that and easy to game as a result I think in the age where we're moving more towards a gentic future where uh you know human clicking is going to be replaced by more like bots and stuff and some of the bots are not not only uh not even malicious but even necessary and very important for the ecosystem. This area this uh point stays even more true that we need to understand the entire journey of of how uh a user or a boss is like doing certain action either on the front end or the back end or the blockchain and we are in a very good position to build out all of that with the initiatives that we are doing.

Yep. And you were talking about future. We only have a few minutes left here. Uh maybe Kite I'd be curious to to hear from you 2030 right four years from now. Okay.

What are a decision that you think people made today and when we're going to look back we're going to say you guys were completely wrong. This was wrong because you guys didn't pay attention to the data.

Um I think something related to what James said. I think these like distributions you have different mechanisms designed which would cause different actions different behaviors. Um if you look at airdrops that's going to be different if you distribute if your token distribution is different. Um and I think not looking at the behavior after each different protocol designs that we see um is going to be costly and then I think people should pay more attention to those um distribution outcomes.

Yep. Danning, the last one I want I want a big no fluffy one like what would be something looking back that you say is we were all wrong or at least

I told you guys but you didn't look at the data.

Yeah, I so this might be a contrarian take. Um but I feel like one thing I'm questioning right now is like we we all hate speculation. We all hate like oh this is just memecoin gambling but like maybe speculation will not go away like it will be a forever needs by people like it the like we build all these rails in blockchain and right now it's kind of dystopian that only memecoin like uh trading bots or like prediction market which is also kind of gambling is taking off but maybe that's what people want and we should be proud of it

100%. Well that's going to be a wrap for us. I hope you enjoy nerding around onchain data. Kite at unis swap use onchain data to ship uh product changes. Danning at Panta does this to deploy capital and Jim does this to uh close the growth loop.

At Dune, we try to encompass all of that. Um if you're interesting in onchain data, you can access it through our platform. We have a booth over there. We have an amazing team around. So if you're interesting in accessing onchain data, please come talk to us.

And thanks for everyone in the room and watching online.

Thank you.

Thank you so much.

Thank you.