Great article!! And i think it can probably be summed up with the sentence, unless your product has:
- a long history of data
- data that has low correlation to other available data sets
- a very small group of people who have access to this data( ie exclusive deals with a few funds)
Then its unlikely that you will be able to sell your data for any meaningful amount of money.
A few things the author wrote that I think deserve some attention.....
> You do hear the occasional story where a hedge fund paid a couple of millions a year to obtain a specific data set, and occasionally more. But there’s a reasonable chance that this type of price came with some kind of exclusive. Also, those contracts also probably have a limited shelf life, as the value of a data set decays over time, as per the above.
This is very true, almost certainly your data set is not worth millions a year to any specific fund. And if it is, its because you agreed to make your data set exclusive to a single fund.
Each data set requires its own special mess of tools for cleaning it, integrating it, figuring out how to generate a trading signal from it, putting it into production and then once its in production constantly monitoring it not only to verify the data is good, but to make sure it actually benefits the trading system.
I wrote this comment: https://news.ycombinator.com/item?id=13140352 a while ago and it got a very large number of up votes for what was essentially a rant that we're now drowing in data.
Hedge funds today share alot in common with AAA video games wrt to data. You used to be able to make a game with 2 people, then it was 10, then it was 50, then 100, now its 500+ for a AAA video game.
Hedge funds have the exact same issue due to the explosion of data. I know alot of funds that are skipping on data sets, due to lack of bandwidth to consume them all.
> First of all, unless it’s already commoditized, you probably don’t want to sell your data to the Bloombergs of the world.
Most hedge funds won't be your direct audience, they'll already have spent money to buy data from Bloomberg and Reuters. The few, say top 100, that might buy your data directly could be your market, but these funds don't exactly have huge sales teams ready to talk to you. Which means you need to hire someone who knows the landscape to sell your data. Your sales team might be the best group of 25 year old Stanford grads around but they aren't getting in the door of most quant funds without an introduction.
Also just plain dumb luck. stocks dont trade based on logic like revenue growth so if Trump tweets something negative about Starbucks and the stock goes down, but your data suggested Starbucks sales went up 300% last month, you are in tough luck.
this is also probably why your data is worth much much less than you think. there are very few things that correlate very strongly to stock performance. predicting human emotion is probably more important than predicting revenue growth or number of people who went inside a Starbucks last month.
If your data suggest that the stock is trading at a discount due to irrationality, then you have a perfect opportunity to buy lots of an underpriced stock.
I think the point OP was trying to make is that, that irrationality may not be easy to capture with automated systems, therefore the value of data is limited by your trading system/algorithm's ability to detect/predict irrational human behaviour. Or atleast that's what I read of it.
Yes, the grandparent comment makes a good point. You can have plenty of solid data and still make a losing trade.
But, given enough data-informed trades that would otherwise be proditable, the law of large numbers suggests that an overall strategy will still be sound. You can afford to have the market be irrational here and there if you really are correctly forecasting earnings in advance most of the time.
"Trump tweets a company name and the stock price decreases" is a signal. If you think about trading strategies as collections of consistent signals you can execute on, rather than assumptions that require market rationality, it doesn't really matter that a rogue event can cause an unpredictable response. The law of large numbers will generally protect you, given a consistent strategy and enough capital.
As an example, a hedge fund that consumes data capable of accurately forecasting earnings in advance can also diversify by consuming social media data. Matt Levine has written about profitable trading strategies developed with sentiment analysis on Trump's tweets. Instead of a hedge fund betting on the stock going up and getting blindsided by a random Trump tweet (or other irrational thing), it will use that as a signal if it can.
> The good news is that hedge funds are not advertisers and don't care about particular individuals
This sounds disingenuous; it's not hard for the author or anyone else to imagine how Wall Street could use personally identifiable information, and the author claims to be at least somewhat sophisticated in the 'big data' field.
Also, it's an industry that has an incredible track record of fraud, even fixing the bedrock interest rates like Libor, prides itself on an anything-goes atmosphere, and disdains regulation.
I could see them passing "insider consumer info" on black markets just like they pass insider information. How handy would it be for a life insurance company to know which of its customers is googling for "chest pain" or "$CANCER_TYPE symptoms". A car insurance company would love to know which drivers go to bars on saturday night, or which ones make it from point A to point B in suspiciously low amounts of time.
Wall street could creep into this by saying to either industry: "let our guys crunch the numbers and let us make book on what these rates should be...because market efficiency". It'll seem like a great idea at first, the data will be "anonymized" (except for when it's not, like when that guy pays a $5k to an analyst at the insurance company for the table mapping "anonymized" customer idns to ssns.
Suddenly hedge funds get really good at guessing which pools should be invested in, or which drivers should be packaged into the high-risk tier. A few years later there's a moderate scandal, a fine is levied equal to ~5% of the illicit gains made, meanwhile the funds have moved on to perpetuate their next round of fraud.
> "No man is better than a machine, and no machine is better than a man with a machine."
This concept also is popular in the military AI R&D world where the human-computer combination is called a 'centaur' (based on what I've read; I'm not of that world).
> When I was at Bloomberg, I used to cringe when startup founders would show up in hoodies, essentially destroying their credibility before the meeting even started.
I understand you can't roll in looking like a slob but I have to think times have changed and the casually dressed founder wouldn't have a problem, even with Wall Street types.
The cultural differences b/t the two industries are pretty interesting though...
And I'm sure they are devastated by this loss of accountability, since all they got in return was a 2x - 10x multiple of income from the average programmer.
If you have something of value and the people you are talking to are actually interested in making money, they will get past quite a lot.
(The lawyer question is more ridiculous, they have to work with the prejudices of the judges and lawyers they interact with so only in the narrowest of circumstances would a credible attorney walk into court in jeans)
I deeply enjoy these sorts of distinctions in decorum, and the undercurrents of what it all must imply. It's all this subliminal signaling that tunes the pitch of an atmosphere to cultivate ideas.
Consider the duality of U.S. Marines, contrasting dress blues against combat BDU's. Or any military subculture for that matter.
Snappy dressers are trying to send a message about the money they burn on dry cleaning, and how early they get up to look sharp for their morning cup of coffee. How new their razor is, how orderly their coiffure seems. They have the time for this sort of thing, and they might have the time to waste on you ...if you're lucky.
And then there's nerd signaling. The next twelve hours to be spent scanning text, perched atop a swivel chair in comfortable clothes. Waking up possibly before before noon, because yesterday's dose of caffeine and inspiration carried through until 4:30 AM, when someone got their head bitten off in a forum by some cranky sleep-deprived rage about abuse of notation.
Meanwhile, in a court room, lives are irrevocably altered. People being hauled before a judge, to have their lives taken away from them, because they were born under a bad sign. There can't be any flippant hairstyles among the insiders of such a system. The whole point of showing up is to impress upon outsiders that a stone faced institution housed within a forrest of doric columns shall weigh and adjucate a decision. Clear voices in respectable, classic outfits must deliver verbal pronoucements with the reverb of a cavern older than the dawn of humankind. Janitors have mopped blood off these floors. This isn't some cluttered bedroom filled with the detritus of decades old junk.
> Does that include lawyers and financial advisors too?
Why shouldn't it?
I have never seen a lawyer in court with jeans and hoodie or a MM in T and flip flops. However...
Every ambulance chaser lawyers ad I have seen has always had some guy in a suit.
Every big financial crook has been in a very expensive suit.
When looking for a lawyer or advisor I would much prefer to know their track record and referrals/recommendations by others than be impressed by their suits.
That said, because of these bad stereotypes, if I was on the other side of the interview table as a lawyer or advisor I'd be in a suit no questions asked.
Can't hurt to look nice. Why not use every advantage you can get? Imagine some Wall Street type who can talk to engineers on a technical level. You're going to automatically trust this person more because they're meeting you where you're at... just like a suit-wearing Wall Street type will trust you more if you show up "speaking" their language in terms of dress.
Not to mention that humans tend to trust others easier when they are more attractive, e.g., dressed snappy [1]. Imagine that.
It can. You want to dress for the impression you want to make. For Wall St. financial types you probably do want to dress like a banker, since what they're really looking for is professionalism and reliability.
But you wouldn't necessarily want to wear a three piece suit while pitching a media company in Northern California, since what you want to say is "We're not staid, boring people who can't tap into trends and fads."
Interviews are the same way. Overdressing is as bad as underdressing.
> The financial services world has its own strong identity, handles massive amounts of money, and will not necessarily be in awe of your cool startup. When I was at Bloomberg, I used to cringe when startup founders would show up in hoodies, essentially destroying their credibility before the meeting even started.
Who wants wall street awe? Who cares about credibility? Run a goddamn business, not a fan club. If you cared about image, you wouldn't talk about what people wear. It's completely irrelevant to why I'd use your product.
Who cares about credibility? If you're trying to sell to Wall Street, you care about credibility. If you want them to invest in your startup, you care about credibility.
If they want to sell to you, wear what you please.
Except that Silicon Valley is equally superficial in different ways. Consider: founder fetishism, engineering fetishism (vs. sales/marketing/product), relative ease of getting funded as a Caltech / MIT / Stanford grad, favoring growth over revenue without coherent business plan, acting like fanboys towards YC or companies such as Uber, Google, Apple, etc.
> Run a goddamn business,.... If you cared about image, you wouldn't talk about what people wear.
Let's change perspective - what do you think, how much success will have business with web pages that looks like from 90-ties or products that looks like they are made in basement in spare time? (sure, there are exceptions, but well, they are exceptions :P).
Everything is about image; if I'm going to invest in you, you better be shaved with ironed shirt and good suite. Simply, it's telling me you are serious about your business.
I'll challenge this statement. I stopped shaving, ironing, and wearing more than many duplicates of one outfit explicitly so I could get serious about my work. Time is my most valuable commodity, and I intend to use every moment I have attacking problems I care about. I like to think that I don't look like a slob, but by the qualities you cited, you would disqualify me off the bat, and given the rocks I've moved in my industry tenure, that would have been an absolutely misaligned signal and you would have lost out.
Everything is about image _in some professions_ and _in some eras_. I will certainly stand on the side of trying to change that through my own actions, and given that the peers and mentors I most respect have similarly chosen to discard traditional models of image, I find myself in company I'm quite comfortable with.
(And as sister posts have cited, the sheer existence of successful sites like CL, Enom, and HackerNews (which for all its minimalism and polish is by no means a "modern image"; I say this with nothing but pride to be very clear) indicates to me that even in site design image may be a cost function but is no substitute for quality. I have had clients JUMP at products that as you say, look like they're from the 90s, because they solved _real issues_ that they needed solved.)
Image still matters. It's just that it's not the same image that matters everywhere. Wall Street? Bring a suit. Silicon Valley? Go for the hoodie, or the same outfit (copy) every day.
Both signal that you're part of a specific tribe. There's a reason major tech conferences are a sea of "jeans & T" people, and it's not just the comfort.
You're completely right that image is not a substitute for quality, but when you're selling something, you ideally want a bond with the customer. So, if you sell to Wall St., look like Wall St. If you sell to engineers, look like an engineer. If you don't sell, wear whatever you like ;)
That's why HN can get away with minimalist design - it's not selling anything.
"I don't know you, and you clearly don't take care of details in at least one thing you do, so why would I trust you take care of other details?"
That impression can be overcome, of course, through demonstrating skill in a particular field. But I think a lot of technical people miss the connection or somehow find the (implicit) question unfair.
>if I'm going to invest in you, you better be shaved with ironed shirt and good suite. Simply, it's telling me you are serious about your business.
This is still not a justification for why we are wasting time and money on appearance. If it's just to show seriousness by cost, why not demand that everyone arrive brining a bowl of dollar bills?
When someone shows up in a suit to meeting here in the bay, I instantly assume that they spent so much time on appearance to distract me from the pile of bullshit they about to unveil.
you're wrong. that's public data and you absolutely can sell it. hedge funds are worried about insider trading so if you can make the case that the data you used to trade was available to everyone then it's cleared by compliance. source: I work for a startup that does that and has 1MM rar.
not according to the article. hedge funds claim they cqre about the means by which you obtain that data. for most of the methods i outlined, it is against the site's tos to obtain that data.
Throwaway for obvious reasons. I'm a quant at a quant hedge fund. I can confirm I want your data. The longer the history, the more I want it. Don't fuck with it much, give me clean but raw info. I don't want PII, but I'll take aggregate and anonymized. Just whatever you do, start collecting it NOW so in 3 years you have 3 years of history to sell me.
Majority are PhD, but definitely some have Bachelors/masters. Care more about quality of education/skills. 4.0 undergrad at MIT vs PhD at rando state U w unknown advisor... no question.
- Make it easy for them to get the backfile, and charge them a lot for fresh data (somehow startups are inclined to do this backwards. They need a few years of data before they even know it's useful, but they can't trade without recent data. Charge them accordingly)
- It's too hard to negotiate exclusives, but you can make them bid on exclusive access to delays. One second, minute, hour, week, etc. it doesn't need to be complicated but neither should you leave money on the table with too simple pricing
54 comments
[ 5.1 ms ] story [ 122 ms ] thread- a long history of data
- data that has low correlation to other available data sets
- a very small group of people who have access to this data( ie exclusive deals with a few funds)
Then its unlikely that you will be able to sell your data for any meaningful amount of money.
A few things the author wrote that I think deserve some attention.....
> You do hear the occasional story where a hedge fund paid a couple of millions a year to obtain a specific data set, and occasionally more. But there’s a reasonable chance that this type of price came with some kind of exclusive. Also, those contracts also probably have a limited shelf life, as the value of a data set decays over time, as per the above.
This is very true, almost certainly your data set is not worth millions a year to any specific fund. And if it is, its because you agreed to make your data set exclusive to a single fund.
Each data set requires its own special mess of tools for cleaning it, integrating it, figuring out how to generate a trading signal from it, putting it into production and then once its in production constantly monitoring it not only to verify the data is good, but to make sure it actually benefits the trading system.
I wrote this comment: https://news.ycombinator.com/item?id=13140352 a while ago and it got a very large number of up votes for what was essentially a rant that we're now drowing in data.
Hedge funds today share alot in common with AAA video games wrt to data. You used to be able to make a game with 2 people, then it was 10, then it was 50, then 100, now its 500+ for a AAA video game.
Hedge funds have the exact same issue due to the explosion of data. I know alot of funds that are skipping on data sets, due to lack of bandwidth to consume them all.
> First of all, unless it’s already commoditized, you probably don’t want to sell your data to the Bloombergs of the world.
Most hedge funds won't be your direct audience, they'll already have spent money to buy data from Bloomberg and Reuters. The few, say top 100, that might buy your data directly could be your market, but these funds don't exactly have huge sales teams ready to talk to you. Which means you need to hire someone who knows the landscape to sell your data. Your sales team might be the best group of 25 year old Stanford grads around but they aren't getting in the door of most quant funds without an introduction.
This is the website of one of the worlds most successful funds. https://www.rentec.com/Home.action?index=true
Hmm ... for anyone reading it, the article covers a lot more than the specifications for selling your dataset.
this is also probably why your data is worth much much less than you think. there are very few things that correlate very strongly to stock performance. predicting human emotion is probably more important than predicting revenue growth or number of people who went inside a Starbucks last month.
But, given enough data-informed trades that would otherwise be proditable, the law of large numbers suggests that an overall strategy will still be sound. You can afford to have the market be irrational here and there if you really are correctly forecasting earnings in advance most of the time.
As an example, a hedge fund that consumes data capable of accurately forecasting earnings in advance can also diversify by consuming social media data. Matt Levine has written about profitable trading strategies developed with sentiment analysis on Trump's tweets. Instead of a hedge fund betting on the stock going up and getting blindsided by a random Trump tweet (or other irrational thing), it will use that as a signal if it can.
This sounds disingenuous; it's not hard for the author or anyone else to imagine how Wall Street could use personally identifiable information, and the author claims to be at least somewhat sophisticated in the 'big data' field.
Also, it's an industry that has an incredible track record of fraud, even fixing the bedrock interest rates like Libor, prides itself on an anything-goes atmosphere, and disdains regulation.
Wall street could creep into this by saying to either industry: "let our guys crunch the numbers and let us make book on what these rates should be...because market efficiency". It'll seem like a great idea at first, the data will be "anonymized" (except for when it's not, like when that guy pays a $5k to an analyst at the insurance company for the table mapping "anonymized" customer idns to ssns.
Suddenly hedge funds get really good at guessing which pools should be invested in, or which drivers should be packaged into the high-risk tier. A few years later there's a moderate scandal, a fine is levied equal to ~5% of the illicit gains made, meanwhile the funds have moved on to perpetuate their next round of fraud.
This concept also is popular in the military AI R&D world where the human-computer combination is called a 'centaur' (based on what I've read; I'm not of that world).
I understand you can't roll in looking like a slob but I have to think times have changed and the casually dressed founder wouldn't have a problem, even with Wall Street types.
The cultural differences b/t the two industries are pretty interesting though...
Would you be OK with someone in jeans and a hoodie representing you in court?
Would you be OK handing your life savings over to a money manager who was wearing a t-shirt and flip flops?
https://www.youtube.com/watch?v=yJFrkNY4n1g
If you have something of value and the people you are talking to are actually interested in making money, they will get past quite a lot.
(The lawyer question is more ridiculous, they have to work with the prejudices of the judges and lawyers they interact with so only in the narrowest of circumstances would a credible attorney walk into court in jeans)
Consider the duality of U.S. Marines, contrasting dress blues against combat BDU's. Or any military subculture for that matter.
Snappy dressers are trying to send a message about the money they burn on dry cleaning, and how early they get up to look sharp for their morning cup of coffee. How new their razor is, how orderly their coiffure seems. They have the time for this sort of thing, and they might have the time to waste on you ...if you're lucky.
And then there's nerd signaling. The next twelve hours to be spent scanning text, perched atop a swivel chair in comfortable clothes. Waking up possibly before before noon, because yesterday's dose of caffeine and inspiration carried through until 4:30 AM, when someone got their head bitten off in a forum by some cranky sleep-deprived rage about abuse of notation.
Meanwhile, in a court room, lives are irrevocably altered. People being hauled before a judge, to have their lives taken away from them, because they were born under a bad sign. There can't be any flippant hairstyles among the insiders of such a system. The whole point of showing up is to impress upon outsiders that a stone faced institution housed within a forrest of doric columns shall weigh and adjucate a decision. Clear voices in respectable, classic outfits must deliver verbal pronoucements with the reverb of a cavern older than the dawn of humankind. Janitors have mopped blood off these floors. This isn't some cluttered bedroom filled with the detritus of decades old junk.
Takes all kinds.
Why shouldn't it?
I have never seen a lawyer in court with jeans and hoodie or a MM in T and flip flops. However...
Every ambulance chaser lawyers ad I have seen has always had some guy in a suit.
Every big financial crook has been in a very expensive suit.
When looking for a lawyer or advisor I would much prefer to know their track record and referrals/recommendations by others than be impressed by their suits.
That said, because of these bad stereotypes, if I was on the other side of the interview table as a lawyer or advisor I'd be in a suit no questions asked.
Not to mention that humans tend to trust others easier when they are more attractive, e.g., dressed snappy [1]. Imagine that.
[1] https://en.wikipedia.org/wiki/Halo_effect
It can. You want to dress for the impression you want to make. For Wall St. financial types you probably do want to dress like a banker, since what they're really looking for is professionalism and reliability.
But you wouldn't necessarily want to wear a three piece suit while pitching a media company in Northern California, since what you want to say is "We're not staid, boring people who can't tap into trends and fads."
Interviews are the same way. Overdressing is as bad as underdressing.
Who wants wall street awe? Who cares about credibility? Run a goddamn business, not a fan club. If you cared about image, you wouldn't talk about what people wear. It's completely irrelevant to why I'd use your product.
If they want to sell to you, wear what you please.
Let's change perspective - what do you think, how much success will have business with web pages that looks like from 90-ties or products that looks like they are made in basement in spare time? (sure, there are exceptions, but well, they are exceptions :P).
Everything is about image; if I'm going to invest in you, you better be shaved with ironed shirt and good suite. Simply, it's telling me you are serious about your business.
Everything is about image _in some professions_ and _in some eras_. I will certainly stand on the side of trying to change that through my own actions, and given that the peers and mentors I most respect have similarly chosen to discard traditional models of image, I find myself in company I'm quite comfortable with.
(And as sister posts have cited, the sheer existence of successful sites like CL, Enom, and HackerNews (which for all its minimalism and polish is by no means a "modern image"; I say this with nothing but pride to be very clear) indicates to me that even in site design image may be a cost function but is no substitute for quality. I have had clients JUMP at products that as you say, look like they're from the 90s, because they solved _real issues_ that they needed solved.)
Both signal that you're part of a specific tribe. There's a reason major tech conferences are a sea of "jeans & T" people, and it's not just the comfort.
You're completely right that image is not a substitute for quality, but when you're selling something, you ideally want a bond with the customer. So, if you sell to Wall St., look like Wall St. If you sell to engineers, look like an engineer. If you don't sell, wear whatever you like ;)
That's why HN can get away with minimalist design - it's not selling anything.
"I don't know you, and you clearly don't take care of details in at least one thing you do, so why would I trust you take care of other details?"
That impression can be overcome, of course, through demonstrating skill in a particular field. But I think a lot of technical people miss the connection or somehow find the (implicit) question unfair.
This is still not a justification for why we are wasting time and money on appearance. If it's just to show seriousness by cost, why not demand that everyone arrive brining a bowl of dollar bills?
When someone shows up in a suit to meeting here in the bay, I instantly assume that they spent so much time on appearance to distract me from the pile of bullshit they about to unveil.
Thats a shame as this makes this business model almost impossible for the average reader here.
There are loads of outside data that if crawled and gathered could be useful to hedge funds. things like:
- growth in number of Amazon reviews for a type of product month to month(against amazon tos to scrape them)
- number of job ads that mention a technology or product (against indeed.com tos to scrape them)
- google rankings of websites (against google tos to scrape them)
- growth in total yelp reviews for a restaurant like Mcdonalds month to month (against yelps tos)
which means this business is only for already profitable companies with large user bases.
and the rich get richer..
Don't care to tell me? Then don't respond with a comment saying "What else do you want me to tell you?" :)
http://www.dataminr.com/ http://www.discern.com/consumer-retail https://www.aggdata.com/
- Make it easy for them to get the backfile, and charge them a lot for fresh data (somehow startups are inclined to do this backwards. They need a few years of data before they even know it's useful, but they can't trade without recent data. Charge them accordingly)
- It's too hard to negotiate exclusives, but you can make them bid on exclusive access to delays. One second, minute, hour, week, etc. it doesn't need to be complicated but neither should you leave money on the table with too simple pricing