I'm surprised more people aren't speaking against what you did... I guess people aren't too bothered about spam until it gets above a certain threshold, and I guess it's only Twitters money that people like you are wasting by abusing their services and breaking their T&C's
HN is the best response I've received compared to any of the other sites that this link was submitted to, to include Reddit and Digg. Two thoughts come to mind:
1. HN is where we just saw a thread where users posted about the most unethical things they've done to get a startup off the ground.
2. I think that many HNers could relate to me. I was a broke, down-and-out geek/hacker-type that came up with an idea and took it on for the technical challenge. And, hey, I made a few dollars off of it, too.
I wouldn't like to guess how many man hours it has taken them to develop the software and processes required to deal with spam on twitter? Then there are the ongoing costs of dealing with new spamming methods, and the increased hardware costs of having to deal with the extra load caused by the spam. And then there is the less obvious negative impact it has on their brand and image when their users are sent unsolicited spam.
But does it add to Twitter's costs in spam management? It's not nearly in the same league as untargetted, extremely high volume spam which follows a million people and sends out essentially the same ad over and over again.
To answer your original question, yes, I think Twitter isn't bothered by spam until it gets past a certain threshold. They're in it to make money, not to wage religious war.
"But does it add to Twitter's costs in spam management?" Yes, it all adds up. If nobody did what he was doing then Twitter wouldn't have had to spent as much money on the problem.
You're assuming spam is like trash: it accumulates from everyone sending small quantities. But spam is never a side effect. 99% of spam comes from a very few operators operating at extremely high scales.
So even if nobody spammed book recommendations with affiliate links, Twitter would still have to deal with the large-scale operators.
1. Are you claiming there are millions of spammers sending 10 each? If so, your definition is different from Twitter's. If it's just a future worry I'm going to cross that bridge when we get to it.
2. I don't know if it's easier, but it's certainly not easy. Otherwise gmail and twitter wouldn't have spam anymore. It's easy for spammers to pretend to be somebody else. Get a new username, a new IP address, forge headers.
Even if it's only 10 spammers, Twitter can't see that. What they see are a million different spammers. So they already have to deal with the problem you claim is harder.
You're pretty much taking the evolution of email spam and going backwards, unless you can point to a specific reason twitter is different from the general evolution of spam on every network up till now I cant see how thats right.
Given a million people sending out 10 spams each theres bound to be a very similar way they carry it out. However 10 spammers dedicated churning out the maximum capacity they can achieve will put up a fight no matter how you try to stop them.
Email spam originated with people like hood, generally not doing much damage and very easy to block and shut down. It eventually led to the current situation where vast majority of spam originates from a handful of spammers. So did comment spamming, and early social network spamming, they all pretty much followed this pattern.
In my view you're letting yourself get wrapped up in the minutiae of the argument. What hood did was wrong and no amount of justification will change the fact that he externalized the cost of commercial advertising. However none of us can look back at our lives and say we didn't have such morally questionable moments. We generally react like hopefully hood has done and don't make it a career; assholes keep going even when they only make $1 for every $100 they externalize.
"Externalized the cost of commercial advertising." That's a very succinct way to put it, thanks.
But I'm not sure how you conclude that that was happening. People who clicked on the affiliate links only paid him off if they made a purchase. If no purchase was made, no money changed hands between any players. Right? An affiliate link is not like an ad.
In fact, even if an affiliate link was like an ad, there's no data on whether the users clicking on the link are poorly targeted, or that the CTR will significantly suffer.
The only mention of spam I see in Twitter's ToS: "You may not interfere with the access of any user, including sending a virus, overloading, flooding, spamming, mail-bombing the Services, or by scripting the creation of Content in such a manner as to interfere with or create an undue burden on the Services."
http://twitter.com/tos
No mention of "unsolicited replies". Heck, for most early users, all @replies on twitter are unsolicited. Most of the time I have an exchange with someone before I follow them. Twitter's just designed to be more promiscuous than email. Anybody can see your tweets, so you're encouraging strangers with insight to jump in.
I think that makes what he did a grey area. Depends on how good his recommendations are, basically.
You make a good point, however Twitter's ToS was quite different back then. Their current TOS was updated: September 18, 2009 (this date can be found at the bottom of today's TOS)
That's around the same time my bot was suspended and Twitter's spam rates started to drop from ~10% to 1%
Hmm, so they made their ToS looser around the time they booted you? That doesn't make sense..
I'm not sure there's any disagreement here. I'm just saying: I think what you did would require human judgement to kick no matter how they write their ToS.
It's not just that they changed their TOS. They also started to take a stronger stance against spam. In addition to publicly announcing "we WILL actively suspend the accounts of spammers" they followed through and probably beefed up their anti-spam staff.
You seem certain of your views and I won't try to convince you. I just want to explain why I and perhaps others aren't 'speaking out against it.' We're entrepreneurs here, and we're trying to learn from each other about what works, especially what's in the grey areas. When you're little you need every little advantage you can get.
Every major piece of social software got off the ground by "spamming a bunch of people." But it was different spam. And the gradations do matter because they're orders of magnitude better and orders of magnitude less load on the system.
>We're entrepreneurs here, and we're trying to learn from each other about what works, especially what's in the grey areas.
speak for yourself. I really have no wish to play in spamish "grey areas"
Being an Entrepreneur does not mean you need to abandon your sense of right and wrong, or your respect for your fellow man. I don't know about you, but this isn't the first company I've started, and it won't be the last. Damaging my own personal reputation in exchange for some more success at this one business is probably a bad idea, long term, especially when the business is one that only works as long as you stay 'under the radar' (meaning, it doesn't scale, because the facebook spam people would stop you.)
There's spamming affiliate links, and then there's testing if the quality of my recommendations is good enough to create a steady long-term stream of revenue. It's the difference between a quick buck and a buck.
I'm honestly conflicted on this. On one hand, it's spam, as users are "receiving" messages they didn't solicit.
On the other hand, it's a kind of cool service, as an "enrichment" to twitter messages. If you could find some way of doing it for those who might find it useful without forcing it upon those who don't, I think this could be a great opportunity.
That's basically where I'm at now. Once my account was suspended I retired the spam bot. I even retired a similar project.
BookSuggest, in its web-app form, was the only way I could think of to use my existing technology in a way that wasn't, well... spammy. http://www.charleshooper.net/twitter/
The question I face now is: How do I take this application and deliver it to the user?
My initial idea, given the technology you have, would be to sell it to bloggers. So, let's say I've got a blog with a reasonable readership. I could go through and manually add affiliate links to relevant books, but that sounds like work to me. I enjoy writing; i don't enjoy jacking around with my templates to add affiliate links.
I'd suggest selling 'monetization made easy' movable type and wordpress plugins.
Just an idea; but it might make you money, and it wouldn't be spammy.
but that's the angle I would take. I mean, not only bloggers, but everyone who has content and users who wants an easy way to monetize.
hell, if you had an affiiate account with more than one bookseller, you could even choose the best deal (either the lowest price, or the highest commission, maybe chooseable by your blogger?) then take your cut before the blogger sees it...
Yeah... that's your freemium model. Give away a plugin that does it but with your affiliate id... give the blogger 50% or something. Then, for a higher fee, offer the plugin with the bloggers affiliate id, where they'd get the money direct from the affiliates, without you taking a cut.
That's an excellent idea, and one that I think xulescu was getting at with integrating the book recommendation engine into a bulletin board.
Part of the challenge is maintaining a large enough corpus/histogram to generate good results. Meaning, I'd probably have to host an accessible corpus interface (or make one downloadable) as part of this project. But maybe that's where the dollars come in?
yeah. if you do require a large corpus, that seems like an advantage to me... as if you do that, and maybe make more of the processing done server-side, it'll be harder to just reverse engineer your plugins, and even if people do, then they still have to come up with a good corpus, which presumably you already have.
The reason most people aren't speaking against it is that spam is just a class of information with a low utility rate, whereas useful information is information with a high utility rate. Sufficiently targeted spam is therefore not spam, but useful information.
I make about 400 EUR every month from Amazon referals without actively doing anything.
I own a comunity fueld site (hosting pictures) that receives about 70k visits (20k unique) per month. Its just a little hobby project that runs itself and has been unchanged for 2 or 3 years now. It's hard to find advertisers due to the sometimes offensive content, let alone making money from banner ads.
So I did the next logical thing: putting an invisible iframe on the site that loads Amazon with my referal id. Anyone who visits my site and decides to go shoppind on Amazon in the next few days, still has this referal id stored in a cookie. Note that I don't actually link to Amazon anywhere. Conversion is soley based on chance.
My conversion rate is about 0.5% and I've been doing this for about 18 month now.
This is called cookie stuffing and it's not only against TOS but also fraudulent and very illegal. You're stealing money from Amazon. The owners of Digitalpoint just got charged with wire fraud for this practice. If ethics don't concern you, you should stop simply because of the low relative reward and high risk.
It would surprise me if Amazon doesn't have an effective automated way of detecting this. It would be easy to detect based on user behaviour when they hit the Amazon page, and Amazon isn't known for throwing money away like that.
hoop saw similar things with his affiliate links; a lot of the sales didn't come from the books he recommended, but once people were on amazon they'd end up buying something.
That's much different though. Those people were at least manually clicking on the amazon links and not having their cookie stuffed secretly in the background. I have a low-traffic book review website that makes $2/month from Amazon affiliates, and around Christmas someone bought a Kindle and I got credit for that, which was a nice surprise. All I ever linked to were books, but if someone starts browsing around Aamzon then you'll get credit for whatever they end up buying.
yea I run a couple book sites and one time a person clicked on one of my book links and ended up buying a couple thousand dollars worth of camera equipment. Jackpot!
All I can say is, nice work. It's creative. I've made some decent money off affiliate programs and twitter. What you're doing is similar in nature to what a lot of people are doing (human and bot).
We recommended these books based on the following terms:
schon, only, when, that
It's really weird that any of the suggestions made sense when words like these aren't blocked. The "schon" is German, but the rest should really not play into any ratings
Apparently my corpus isn't a large enough size to naturally filter out useless words like the ones you listed so I have added stopword filtering. Give it another shot and see how it works for you.
While I haven't seen any numbers, I think that there are alot of these bots. I have even less of an idea if any of them are contextual alot. Right before I was suspended, I believe that there was a report that came out that suggested that a large percentage of Twitter's traffic was caused by spam bots like these. I'll see if I can find the report and verify the numbers!
What I do know is that there are certain keywords that put you in line to be tweeted at for some product or website. Some keywords have more bots trolling the search API than others.
UPDATE: I wasn't able to find the report I was looking for, but I did find the following blog post. It contains a graph that shows that SPAM used to account for about 10% of Twitter's traffic. This number is now below 1%.
That is a great idea and I do not find it too spammy. Twitter could implement something like that themselves as advertisement (simple context aware text ads).
It appears that Twitter is starting to move in this direction. I recently read that they now embed what they call "Promoted Tweets" (another word for an ad) in their search results. I'm wondering how long it will be before these start showing up in our timelines.
I did this a long time ago, almost the exact same thing, but I actually parsed trending topics and searched Amazon for ANY product, not just books, based on the content of the trending topic. I started out with books, but moved on to all products. It was a cool project and netted me some cash but not worth carrying on so I dumped it about a year ago.
Was actually incredibly easy to do. I'm sure there are loads more out there.
They started out like "@{{user}} You should read {{book title}} {{shortened book url}}" but many people would take offense if the recommendation wasn't a very good one or the book touched on a sensitive subject (like weight.)
I had much better results with "@{{user}} Have you read {{book title}}? {{shorted book url}}"
Interesting idea (to suggest based on already written content).
Maybe it would be considered less "spammy" if it would be just semi-automated, i.e. more like an "expert system". Also, it it would work for popular forums like PHP BB or others, would greatly extend it's usage (so not just on Twitter).
That's a great idea. I hadn't thought about plugging it into a bulletin board system, but did think about trying to plug it into facebook. A recommendation engine plugin for phpbb, etc could definitely make the forum operator some additional revenue.
Sadly, I lost my corpus some time ago so until I pull more data from Twitter (I've got a script doing this as we speak) the suggestions are going to be a little weak.
The book suggestions are faded unless you mouse over them - if you click on them, you will be able to see a description of the book (as provided by Amazon.) I'll experiment with different design elements to see how I could make this more intuitive. Any suggestions?
Great to see a response. Regardless of your previous bot, seems like you have a good application now and it is worth getting off the ground. I will try it again in a couple weeks once your Corpus is rebuilt.
As far as the design, I would suggest the book icons come through in full color, and when you mouse over you could highlight them, or even better provide a short summary (just like how Netflix provides mouse over summaries of movie suggestions).
Looks much better now =)
Only other suggestion I would give is to improve the UI on that landing page. I'll be interested to see how this turns out for you. Cheers.
I have just added some stopword filtering until my corpus becomes more fully developed. Hopefully this will improve the results of the recommendations!
