MMO SMTB How many of you run Search Arbitrage?

[Glamoco]

Man, almost every traffic source has bots these days. The trick is to find a way to detect them, but that's a lot of work. To do it right, you need to know a lot about various things. You can use tracking tools and spy software to help, but they won't completely solve the problem.

Plus, there's always at least 10-15% of traffic that's from bots using very sophisticated methods. You won't catch them unless you use equally advanced detection techniques.

How do I know all this? I make my living by researching and engineering ways to stay ahead of the system and remain undetected. It's a constant struggle that requires lots of coffee and cigarettes!

We're a company that specializes in detecting click fraud, so we use some pretty advanced techniques to catch it.

Bots have gotten smarter, but we can often spot them by analyzing network activity.

Lately, a lot of scammers are using real traffic to get past bot detection systems, but we can catch this kind of fraud as well.

Most ad networks are pretty bad at detecting click fraud. Some don't even bother trying. There's one big, well-known ad network that's full of worthless clicks.

 

[XoInlove]

We're a company that specializes in detecting click fraud, so we use some pretty advanced techniques to catch it.

Bots have gotten smarter, but we can often spot them by analyzing network activity.

Lately, a lot of scammers are using real traffic to get past bot detection systems, but we can catch this kind of fraud as well.

Most ad networks are pretty bad at detecting click fraud. Some don't even bother trying. There's one big, well-known ad network that's full of worthless clicks.

So, are you saying that your company can instantly detect 100% of bots and fake traffic?

 

[Adtecarmp]

Taboola? I gave it a shot once, but it didn't go well at all. I think most of the visitors were just bots, not real people. How do you handle that? How can you filter out the bots?

 

[Cruistyzx]

Taboola? I gave it a shot once, but it didn't go well at all. I think most of the visitors were just bots, not real people. How do you handle that? How can you filter out the bots?

Every network has bots, some more than others. The real question is, what are they doing to stop them? In my opinion, beating bots is like a hollow victory because they and their creators are always changing and staying ahead of the networks. It's a bit like viruses. First, a bot or virus is made, then it's spotted, and eventually, it's fixed. It's a never-ending cycle.

 

[Glamoco]

So, are you saying that your company can instantly detect 100% of bots and fake traffic?

I wouldn't say we catch every single one, because there will always be a few sneaky folks who stay ahead. But we catch most of them, and we're better at it than your typical ad network.

For instance, we once spotted almost all the fake clicks for one of our clients. Nearly every click on their ads was fake for a while. But their ad network said they were legit and tried to bill them for it.

This happened because they're in the payday loans business, where fake clicks are pretty common.

 

[XoInlove]

I wouldn't say we catch every single one, because there will always be a few sneaky folks who stay ahead. But we catch most of them, and we're better at it than your typical ad network.

For instance, we once spotted almost all the fake clicks for one of our clients. Nearly every click on their ads was fake for a while. But their ad network said they were legit and tried to bill them for it.

This happened because they're in the payday loans business, where fake clicks are pretty common.

That's what I'm saying. Nothing is completely certain, especially in this line of work. Plus, there are a lot of false alarms, like real people acting like bots. I've seen plenty of stuff that doesn't fit neatly into categories.

There are lots of ways to filter out bad traffic, but it's not wise to depend solely on automated tools or software. Those tools are only as smart as the people who make them, and there's always someone out there who's smarter than you or me. Like you mentioned, you had one customer where you caught all the fraudulent traffic, but your algorithm only worked perfectly in their case, and maybe a few others.

What I'm trying to say is that catching fraudsters is like a game of catch-up. You and other companies like yours always have to keep improving. The best setup is to have teams, one focused on catching fraud (the red team) and another on defending against it (the blue team). They need to be experts in marketing and have different tech skills, including machine learning like in your case. But setups like that are pretty rare.

 

[Mamba]

I wouldn't say we catch every single one, because there will always be a few sneaky folks who stay ahead. But we catch most of them, and we're better at it than your typical ad network.

For instance, we once spotted almost all the fake clicks for one of our clients. Nearly every click on their ads was fake for a while. But their ad network said they were legit and tried to bill them for it.

This happened because they're in the payday loans business, where fake clicks are pretty common.

Which company are you referring to?

 

[Glamoco]

That's what I'm saying. Nothing is completely certain, especially in this line of work. Plus, there are a lot of false alarms, like real people acting like bots. I've seen plenty of stuff that doesn't fit neatly into categories.

There are lots of ways to filter out bad traffic, but it's not wise to depend solely on automated tools or software. Those tools are only as smart as the people who make them, and there's always someone out there who's smarter than you or me. Like you mentioned, you had one customer where you caught all the fraudulent traffic, but your algorithm only worked perfectly in their case, and maybe a few others.

What I'm trying to say is that catching fraudsters is like a game of catch-up. You and other companies like yours always have to keep improving. The best setup is to have teams, one focused on catching fraud (the red team) and another on defending against it (the blue team). They need to be experts in marketing and have different tech skills, including machine learning like in your case. But setups like that are pretty rare.

Thanks for getting back to me.

I understand your point, and I don't necessarily think that if something isn't perfect, it's worthless. Having around 95% knowledge of fake clicks is definitely better than having no idea at all, right?

If we follow your line of thinking, it would mean there's no point in using antivirus software because there are always new viruses it hasn't caught up with yet. But I think most people would say it's better to be protected against most viruses than not at all.

We were able to spot a lot of fraud for one customer because our search arbitrage focuses on specific ad keywords. So, click fraud isn't evenly spread across all advertisers. Some customers have more fraud, some less. This doesn't mean the click fraud detection tool failed; it actually shows it's doing its job right.

I'm not sure if the best solution is to hire your own team of cybersecurity experts (which would cost at least $20,000 per month) when you could just pay an existing click fraud service $600 per year.

 

[XoInlove]

Thanks for getting back to me.

I understand your point, and I don't necessarily think that if something isn't perfect, it's worthless. Having around 95% knowledge of fake clicks is definitely better than having no idea at all, right?

If we follow your line of thinking, it would mean there's no point in using antivirus software because there are always new viruses it hasn't caught up with yet. But I think most people would say it's better to be protected against most viruses than not at all.

We were able to spot a lot of fraud for one customer because our search arbitrage focuses on specific ad keywords. So, click fraud isn't evenly spread across all advertisers. Some customers have more fraud, some less. This doesn't mean the click fraud detection tool failed; it actually shows it's doing its job right.

I'm not sure if the best solution is to hire your own team of cybersecurity experts (which would cost at least $20,000 per month) when you could just pay an existing click fraud service $600 per year.

I think you misunderstood me. I never said we need solutions that are 100% foolproof. In our field, nothing is completely certain. We need to strive to improve every day. Any fraud detection is better than none, just like any antivirus is better than having none.
What I meant is that being one step behind fraudsters isn't necessarily bad. We don't need to catch 100% of them. Even catching 10% can make a big difference, but we must keep evolving and learning every day. It's a continuous process that never stops.

You're right about the keywords, but it's hard to tell for sure if someone looking for car insurance is a human or a bot. I've watched a lot of screencasts of both, and believe me, some humans behave as unpredictably as bots (I have some in my family!). On the other hand, some bots interact with websites very naturally, like real humans—they click around sensibly, scroll at reading speed, and close pages without clicking on ads.

 

[Glamoco]

I think you misunderstood me. I never said we need solutions that are 100% foolproof. In our field, nothing is completely certain. We need to strive to improve every day. Any fraud detection is better than none, just like any antivirus is better than having none.
What I meant is that being one step behind fraudsters isn't necessarily bad. We don't need to catch 100% of them. Even catching 10% can make a big difference, but we must keep evolving and learning every day. It's a continuous process that never stops.

You're right about the keywords, but it's hard to tell for sure if someone looking for car insurance is a human or a bot. I've watched a lot of screencasts of both, and believe me, some humans behave as unpredictably as bots (I have some in my family!). On the other hand, some bots interact with websites very naturally, like real humans—they click around sensibly, scroll at reading speed, and close pages without clicking on ads.

Thanks for your reply.

We constantly monitor traffic and find new techniques from time to time, but overall, we can detect it because (1) we see the traffic and (2) we see the clicks. The smarter fraudsters have stopped using bots, but we can still detect their clicks. Thankfully, most fraudsters use the same methods, so the very clever ones are rare. We do catch them eventually

We detect bots through network packets and browser irregularities. For example, certain functions aren’t available on headless browsers, and we can tell if they’re missing or faked.

 

[RaeLaugh]

This conversation got off track

Bot traffic is a big problem, and that's why the search arbitrage game is changing. Now, they can tell if you're sending traffic from an unapproved source. This game will change soon, and only a few skilled players will remain. It's already getting harder to get an arbitrage feed.

 

[Screevava]

These domains are usually hidden and not advertised openly. If you come across a search results page that looks like Google, Yahoo, or something similar, it might be one of them. Honestly, anyone can register a domain and get search results from an XML feed.

Hi Alex - I'm really into this topic too. Can I get in touch with you?

Hey Alex, can you send me a private message?

 

[Fishar]

Honestly, I don't think such material exists out there. I came across SA years ago by chance and got hooked. Everything I know about it, I learned from friends, peers, colleagues, and my own trials. The challenge is that platforms like Taboola and Outbrain have high entry requirements. Maybe you should start with search feed monetization and the "native to search" approach if you're skilled enough to build a site and attract traffic. Later, you can scale up to search arbitrage.

Edit: I just thought of another way to do it using social networks like Facebook. You buy ad space on Facebook and promote your search engine results page (SERP) there. Traffic from those ads can be directed to search feeds from Yahoo, Bing, or Google. I haven't tried it myself, but I believe it could work. If anyone has tested this approach to SA, please feel free to share.

Hi Alex - I'm really into this topic too! Can I contact you?