Snapchat does indeed currently run on GAE. In this article, I was trying to show cost of running efficient custom infrastructure to solve the problem. GAE's per-byte costs are higher than a custom solution operating at scale for a number of reasons, most importantly because GAE is a generic system and can't optimize for this specific use case.
As systems scale to massive numbers it's almost always more cost-effective to use custom solutions. (e.g. Twitter runs an entirely custom HW stack and this has saved them a ton of money)
These guys clearly do not care much about the costs right now... the more they grow the larger the drag is going to be, though. I have heard from a solid source that they're storing these images in BigTable. So, they're probably paying 0.18/GB a month (or maybe less with a deal) and then also paying for the reads and writes as units, and then also for the bandwidth (which is amplified by the # of recipients). I'm not sure if you could find a more expensive way to store images if you tried. Being in GAE means they wouldn't have peering arrangements, so they are paying top-dollar for GAE BW, where a FB or Google is paying for a fiber run and some rack space. SC are very green when it comes to the business of shuffling bits, and if they want to really play with the big boys they will have plenty of catchup to do. Google is making lots of money off them, which makes me wonder if that was part of the FB play.
This is the God damn age of creating value out of nothing! Snapchat is not worth $3 billion, it's been 'valued' at $3 billion but it is intrinsically worthless.
Yeah, I get that part. I'm just reacting to the idea in a headline, that just because something is VC funded and "dirt cheap" to keep running makes it profitable.
Another reason this kid should've taken the money and ran like hell. In another 12 months the kiddies are going to be on to the next coolest social media trend and they'll be left in the dust.
Maybe, maybe not. I'm sure people were saying the same about Facebook a few years ago. The fact that they can still easily choose who receives their "updates" makes for a longer lasting product in my opinion. The "My mom is on Facebook!" phenomenon has led to a dropoff in adolescent usage. Snapchat is a supplement/replacement for texting in my experience with younger family members.
I don't think they are running any ads yet but they could easily make money. A couple ideas:
-Ads that quickly flash for 2 seconds before the 8 second photo snap. "Transformers 5 in theaters tomorrow." Kids have very short attention spans as it is -- this is all you need these days. No more 30 second spots. You have to see it to see your friend's snap, no skipping or blocking.
-Celebrity endorsement snaps. You see a new snap appear from "selena91" or whatever. Open it and it's a quick video selfie of Selena Gomez telling you to check out her new movie/album.
The countdown refers to the amount of time you can view the snap (video/photo) before it is inaccessible. Snaps are immediately viewable (given a fast connection).
The advertiser KNOWS the whole ad will be displayed, vs creating a 30s ad that most users don't see most of because they skip after Xs.
The ad is over almost before the user realizes it's started.
No more conditioning viewers to instantly enter the "how and when can I skip this" mindset. (Tangent: none of the full-window "here's an ad while the page loads" ads work on me; I'm conditioned to instantly hit "skip" and make a point of not noticing the ad itself).
Knowing the ad will be very short, viewers are more likely to not attempt to skip it because it will be done (or very nearly so) before they can do so.
If a user is going to see a given ad frequently, best to get it over with and not turn it into a harrowing experience of repetitive tedium.
There's probably a worthy study of ad effectiveness between the initial awareness of the ad and the cognitive turn against it. Get it over with before the desire to skip begins, or a "I hate this ad" arises.
Just make sure the viewers don't explode from the density & pace of content.
I think it's "eventual profitability", as in the author explains how SnapChat's infrastructure costs are essentially nothing, so getting from some millions of users to profitability should be extremely simplistic without the highly redundant data store needs.
There is always the standard way to make money: sell all the user data to advertisers and anyone else who wants it. Want a list of people in city X that chat more than Y times a day with influential network of Z people? It's all for sale, just no one brags about it.
I'm not actually sure how valuable that is. What am I supposed to do with a list of supposedly influential names and cell phone numbers? Match them to addresses and send them direct mail? Sounds like a lot of work.
Correlate it. You've got phone numbers. That's a strong identifier. Phone/Power/Gas company will sell your their name and address. Other sites need a phone number and might have more details as well. There are a ton of big data brokers out there that do just this.
Think about it. If I know you, everything about you, I can sell directly to you. Not just things that are vaguely intersting to you, but to the most primal want you have. The key is getting an accurate image of who you are, tapping into the "I cant resist" part of your individual monkey brain. We all have somethings that we don't think rationally about,and if someone can map you down to you that point, then they have you. Your patterns on social media leak that data. The more sources, the more accurate they can construct "you."
The next step is biometrics. The kinect is a good example. Im betting that is looking to measure pupil dilation, heart rate, etc. if its not already. Looking at an ad and your heartrate goes up and pupils change? Time for these similar ones, ever deeper and deeper into you.
If I can do that, I can sell that info to a vender with an actual product, making myself very wealthy just moving data around. Its even better if Im also that vendor, like in microsofts case with games. Then I cut out the middle man and vertically integrate myself to the moon.
Snapchat's biggest problem was never going to be whether it could get sales above expenses, but whether it can retain the user base in a era where users are fast to abandon ship for the next experience (and doing so is easier and cheaper than ever).
Agreed. Whether it is cheap to run or not is besides the point. There's nothing special about snapchat that a company like Facebook can't create in a couple months. The only thing they have of value is a very large, very fickle user base. Why they turned down three billion seems completely insane from my perspective. It's not even clear here that they can monetize effectively before they lose popularity.
Bad headline. The interesting point from the article is that metadata can be more valuable than data. Plus it's smaller. Therefore it improves both sides of the cost:benefit ratio.
Interesting article, terrible title. It's key to profitability will be actually generating revenue while preventing their users from fleeing to the next new thing.
>actually generating revenue while preventing their users from fleeing to the next new thing.
This is a huge question for even most of the big players. Facebook didn't exist till 2004, and they are already seen as nearly completely irrelevant by High Schoolers. I think we will only see an increase in the speed at which social apps get dropped for the next fad.
"In fact, the way the metadata is ostensibly used to target individuals and groups NSA agents deem to be a threat is not dissimilar to how advertising targeting works. But that’s another discussion."
This was an important takeaway, although in hindsight it's pretty obvious. Assuming (conservatively) NSA is approximately 5 years ahead of current commercially available technology, it seems there's quite a bit of headroom for targeted advertising methods to grow and get better.
Actually, the stuff the NSA doing is pretty old school for people in the ad space. Anyone doing online ads on social networks was doing this stuff even before Facebook started doing it.
Still a lesson here in running lean. Unfortunately most of the time when you get certain kinds of funding they cant wait to blow the place out full of desks for the dog and pony show that "something" is happening!
I know it's a little off topic by why on earth wouldn't one take a 3-billion-cash offer for a company without revenue? This simply amazes me. I must be missing the uber-motivated-entrepreneur gene.
Maybe the founders are already rich, they like the attention, and don't care about padding their bank accounts any more. They do (or should) have a duty to their investors and employees who might feel otherwise though...
Even in the case of rich founders, I think money is such a great enabler that one can take that incredible amount and move on to the next thing (or things). I respect one's decision to brush the money aside and move on towards the big dream but we're talking about 3 billion here :).
Before the $3 billion offer, when SnapChat raised $80 million at an $800 million valuation, the founders each took $10 million off the table. The investors probably wrote a clause into the deal giving them the right to approve any acquisition offer [1]. The investors want tens of billions.
Do we know it was just the founder who made the decision, or could it have also been VCs trying to stretch a double into an inside-the-park four-bagger?
I'm not hugely familiar with mobile, but I don't quite understand why Apple/Google can't just immediately kill Snapchat by rolling out an expiring photo message feature in their next update. As far as I can tell, there is nothing keeping users specifically on Snapchat since it pulls from your phone's contacts, and since the content expires it's not like you have a mountain of pictures stored in the application that you don't want to lose, like in Instagram.
What Apple/Google app? iMessage? Plenty of people on Snapchat don't have each other's numbers. One of the biggest issues would be that Apple never opened up iMessage (and Google hasn't opened up Hangouts) so there would be no cross platform snaps.
FB tried this with "poke" (remember that) and that didn't work so well. A service with a number of users is not as easy to copy as it would seem. Replicating the technology is one thing, getting all the users is another.
Unless they have some kind of vendetta against Snapchat, I doubt either company is interested in cloning a zero-revenue app which is completely orthogonal to their main businesses, and may in fact turn out to be a fad.
Apple's software & service products don't exist for their own revenue, they exist to attract more people to the proprietary hardware and to keep them there. Considering Facebook's & Twitter's iOS integration, Apple might just do the same with Snapchat and pay them some real revenue - making Snapchat utterly dependent on Apple without actually making the service Apple's problem.
Yes, but neither company is great at publishing cross-platform apps. I gave the "network effects" example to point out that "killing Snapchat" wouldn't be so simple as having built-in iOS functionality - because the people that Snapchat users like snapchatting with are already all in Snapchat and might not carry over to a new platform-only service.
Seams to me that his cost calculation is way off.(one should be able to pull of a service as snap chat where data persistans is not that vital, for a fraction of the cost he wrote)
In my experience long term storage reliability is where the biggest costs is in most IT related stuff that i have seen.
Snappchat never has to store anything for more than a couple of hours at most in the majority of all cases.
Could someone in the ad industry please confirm the key assertion of the article: that the targeting value of Snapchat's data is comparable to Facebook's targeting value? With improvements coming in NLP and computer vision (i.e., understanding images and someone's preferences), it also seems unlikely to remain true even if true today.
Most things are dirt cheap to run. People just choose not to do so.
There have been countless services and websites over the last decade that somehow get or require massive infusions of money. Tens of millions of dollars, even. It blows my mind. Why does your website that mostly just regurgitates tech news in blog format each day (and usually only about cell phones and tablets, at that) require twenty million dollars of capital? You need a couple servers at rackspace and a couple people at home in their underwear surfing the web and parroting existing stories and news throughout the day.
When I hear "snapchat is incredibly cheap to run", what I hear is "it is reasonably priced to run, like most other services should be, but without the bloated and inflated needs that others somehow ladle onto their sites/services".
68 comments
[ 2.8 ms ] story [ 188 ms ] threadAs systems scale to massive numbers it's almost always more cost-effective to use custom solutions. (e.g. Twitter runs an entirely custom HW stack and this has saved them a ton of money)
This is the God damn age of creating value out of nothing! Snapchat is not worth $3 billion, it's been 'valued' at $3 billion but it is intrinsically worthless.
Of course $3 billion is completely ridiculous, but it's a little unfair to say they are worthless.
-Ads that quickly flash for 2 seconds before the 8 second photo snap. "Transformers 5 in theaters tomorrow." Kids have very short attention spans as it is -- this is all you need these days. No more 30 second spots. You have to see it to see your friend's snap, no skipping or blocking.
-Celebrity endorsement snaps. You see a new snap appear from "selena91" or whatever. Open it and it's a quick video selfie of Selena Gomez telling you to check out her new movie/album.
EDIT: To clarify, I mean that if the service is dirt cheap, it'll also be dirt cheap for competitors to pick up where Snapchat left off.
The ad is over almost before the user realizes it's started.
No more conditioning viewers to instantly enter the "how and when can I skip this" mindset. (Tangent: none of the full-window "here's an ad while the page loads" ads work on me; I'm conditioned to instantly hit "skip" and make a point of not noticing the ad itself).
Knowing the ad will be very short, viewers are more likely to not attempt to skip it because it will be done (or very nearly so) before they can do so.
If a user is going to see a given ad frequently, best to get it over with and not turn it into a harrowing experience of repetitive tedium.
There's probably a worthy study of ad effectiveness between the initial awareness of the ad and the cognitive turn against it. Get it over with before the desire to skip begins, or a "I hate this ad" arises.
Just make sure the viewers don't explode from the density & pace of content.
The next step is biometrics. The kinect is a good example. Im betting that is looking to measure pupil dilation, heart rate, etc. if its not already. Looking at an ad and your heartrate goes up and pupils change? Time for these similar ones, ever deeper and deeper into you.
If I can do that, I can sell that info to a vender with an actual product, making myself very wealthy just moving data around. Its even better if Im also that vendor, like in microsofts case with games. Then I cut out the middle man and vertically integrate myself to the moon.
And that's a good thing actually
Technically, yes, it's easy. The hard part is getting people to use your app, even when you're FB. They tried, and failed. [1]
[1] https://itunes.apple.com/en/app/facebook-poke/id588594730?mt...
This is a huge question for even most of the big players. Facebook didn't exist till 2004, and they are already seen as nearly completely irrelevant by High Schoolers. I think we will only see an increase in the speed at which social apps get dropped for the next fad.
This was an important takeaway, although in hindsight it's pretty obvious. Assuming (conservatively) NSA is approximately 5 years ahead of current commercially available technology, it seems there's quite a bit of headroom for targeted advertising methods to grow and get better.
So they can cash out
1. http://valleywag.gawker.com/source-snapchat-cofounders-unloa...
Hangouts is closed source, so is SnapChat.
iMessage probably not, but hangouts could add this in a heart beat.
Edit. Punctuation/formatting.
Edit: Turns out it's a lot easier than that:
https://developer.apple.com/programs/terms/registered_apple_...
"Apple may terminate or suspend you as a Registered Apple Developer at any time in Apple’s sole discretion."
Apple could kill snapchat at the click of a mouse and give any reason they want.
In my experience long term storage reliability is where the biggest costs is in most IT related stuff that i have seen.
Snappchat never has to store anything for more than a couple of hours at most in the majority of all cases.
There have been countless services and websites over the last decade that somehow get or require massive infusions of money. Tens of millions of dollars, even. It blows my mind. Why does your website that mostly just regurgitates tech news in blog format each day (and usually only about cell phones and tablets, at that) require twenty million dollars of capital? You need a couple servers at rackspace and a couple people at home in their underwear surfing the web and parroting existing stories and news throughout the day.
When I hear "snapchat is incredibly cheap to run", what I hear is "it is reasonably priced to run, like most other services should be, but without the bloated and inflated needs that others somehow ladle onto their sites/services".