Home > Tracking Campaigns > Tutorials, Tips and Guides

---

When To Care - And When NOT To Care - About Bot Traffic (14)

---

There's been a lot of talk about bots on STM recently, and I want to correct a few misapprehensions that some people may have about bot traffic.

In short: bot traffic is not always a killer. It's never good, but you shouldn't regard "bot or no bot" as the ultimate decision-maker on a placement.

Why not? Well, let's look at 3 different types of placement.



John Connor's Nightmare: The 95% Bot Placement



When you run a bot test on a traffic source, you'll often come across placements which are very, very high percentage bots, up to 95% or 100%.

This is where bot testing (here's how to do a simple bot test) really shines. These placements are absolute garbage. They've got no chance of giving you any sort of a return - but if you're just testing conventionally, you can waste 3x your payout on finding that out, per placement. Bot testing lets you check a lot more cheaply on many sources - 30 or so clicks, sometimes less.

These placements are really simple. But...



I'm Sorry, Dave, I Can't Convert That: 50% bots



Placements with mixed bot and human traffic are much more complex. I've run bot tests in the past and discovered 60% bot traffic on profitable placements.

Why?

Because the remaining 40% traffic was competitively priced.

The first thing you need to do with a mixed traffic placement is to figure out if it can be profitable, based on the human traffic alone. Say you've got a payout of $12, and you're running on a native traffic source with $0.50 CPC. Now, say you discover that placement is 60% bots. That means that 6 out of 10 of your clicks will be bot-driven if they average the same click rate as humans. In other words, if you pay $5 for clicks, you only get 4 legitimate clicks, for an effective CPC of $1.25.

Are you likely to see a conversion rate of significantly above 10% (roughly the amount you'd need to break even)? That determines whether the placement is priced fairly, and whether you should continue bidding on it. In this case, I'd be skeptical: $12 payout implies a lower CVR. Native traffic can boost that, but still, it would look to me like that placement is being bid on by people who don't understand that they're paying for bot traffic, and aren't taking that into account.

As a general rule, the more brand-friendly the placement, the less likely it is that the pricing will take into account bot traffic. But that's just a rule of thumb based on the fact that brand advertisers tend to be less performance-driven, and more driven by being able to say to their clients that they've had "300,000 eyeballs on this ad" (even if 299,999 of those were less eyeballs and more HTML parsers).

So what should you do with a high-bot placement that you believe could be profitable? Answer - test it as normal, but ignore all metrics that don't make you money. Bots will play hell with your CTR - they'll either artificially inflate it or artificially reduce it. You might see click loss, you might see all kinds of odd behaviour. Ignore everything except how much you're spending, and how much you make. Conversion rate per click, cost per click, and your overall payout are all you care about.



The Promised Land? 0% bots



And here's the other thing to remember: just because there are no bots doesn't mean the placement will convert.

There's an excellent example of that on a lot of DSPs right now: Grindr. I've done multiple bot tests on Grindr traffic, and it's one of the purest-human sources I've seen (at least as of six months ago). Almost no bots, just tons of horny gay guys.

It's also on a lot of blacklists. Damn thing just won't convert for many people. Why? Well, just because the horny gay guys are human doesn't mean they're interested in your ad. The demographic is tech-savvy, cynical about advertising, and very very focused on one thing and one thing only: shagging. So despite being a very high-quality placement from the bot point of view, it's a terrible placement from the making-money point of view in most cases.

(Grindr can be an amazing placement, but you've pretty much got to tailor your entire campaign to it, and push something the users are actually interested in. Muscle, diet, dating.)

Other placements get great bot results, but won't convert on anything. That usually means they're heavy on fat-finger clicks: the ad's placed in such a way that lots of users click on it accidentally. Games are particularly bad for this. Some of them may even incentivise users to click on ads to get in-game currency - of course, the CVR of such clicks is terrible.



So just because you've found a 0% bot placement, that doesn't mean you've found the Promised Land. You could get better results with a high-bot but high-conversion placement.

All of this will become more important as advertisers become more bot-savvy. Expect to see mixed-traffic placements get a lot cheaper, particularly once brands start throwing fits about paying for bot traffic. And given the current press interest in ad fraud, expect to see that happening sooner rather than later.

What do you think? Any tips on making the most of mixed-traffic placements, or any questions on the proper care and feeding of bots? Let me know below!

---

Thanks for yet another amazing post caurmen!

So basically: Use the bot test script to weed out the 95%+-bot placements, then test the rest and cut as we usually would when they convert badly.

I'm building this into my campaign strategy to save some money on testing.



Amy

---

Good article Caurmen!

I like the picture of the promised land :-)

---

This is exactly the reason, why I hate BOTs so much as an affiliate. If it was just bots=cut, no bots=keep ... that would be fine. But the fact that bots pollute pretty much every placement makes the optimization part just harder. Bots click banners and then LPs, so they skew up the data and make it harder to make qualified decisions.

Great article Caurmen!

---

As a side note, I also accidentally let slip one of my favourite strategies here - find a placement that everyone else hates, but that has considerable volume of real people. Then find a way to make it convert.

One-placement campaigns aren't quick roads to five figures a day, but they can be very stable, meaning you can leave them ticking in the background whilst you work on more stuff.

---

Hi Caurmen,

If possible, could you elaborate on this one? Not sure if i'm thinking the same thing as what you're saying.

I recently ran a camp on Zeropark where one placement amount for half of the conversions for my sweeps offer but also drained most of my budget as well. Which resulted in a -70% ROI (for this placement only). It was obvious that this placement has a lot of traffic and conversions rate is high.

Newbie question: how do you tackle this issue? do you blacklist it or try find an offer that may works better for this placements (I previously blacklisted this placement )? Not sure if I'm being clear here or if I've communicated my concern/question clearly.

---

@ekentran1558 - We're not quite talking about the same thing here: I'm referring to specifically picking a single large placement and tailoring an entire campaign to it.

Having said that, what I'd do in your situation would be to break that placement out into a seperate campaign, and start testing landers and offers like crazy. I'd also (sorry to be a bore on this one) run a bot test on it to get an idea of what percentage of the traffic you're seeing is actually human, and I'd start experimenting with bids to see if I could improve the ROI that way. And of course I'd drill down into the stats to see if a specific carrier, device, or other factor had a significant effect on conversions.

(Keep in touch with your AM to find out what the lead quality's like during that process, too. No point working endlessly on a single placement if it'll just get you kicked off offers.)

However, if after a few rounds of testing I wasn't seeing improvement, I'd probably write that placement off. Some placements do have a habit of converting reliably, but not enough to be profitable.

Hope that helps!

---

Thank you for taking the time to respond and clearing things up for me, Caurmen.

---

@Caurmen, thanks for the info.

Lets say you found 1 placement that has a lot of bot traffic. You decided to skip this placement. So my question is will you retest the placement again in future? Is there any possibility where the placement will come back to live with a lot of human traffic?

---

If it was really a LOT of bot traffic, chances are its a placement run by some fraudster and it's highly unlikely that it could become a good placement in the future.

---

How can you do that? from what I have seen trying to focus on only few widget_ids/Pubs, let alone one is impossible and will result getting no traffic. I'll be happy to know how can this be achieved?
Unless I got something wrong here?

Thanks

---

This is not possible everywhere, but there are niches where it can work like this : for example adult, there is a handful of super large sites with millions of daily visitors - think sites like xhamster, pornhub, xvideos ... any one of them can burn $XXXX a day even in smaller GEOs.

---

Ok thanks, so it's basically directly with them, not through the traffic source (mgid for example), correct? as these are not revealing the pub that stands behind the widget_id.

---

Nope, you simply use the traffic source but make separate campaign for just that 1 placement. Technically, you don't really need to know what site that widget is placed on, you just need to know it has huge volume. But like I said, it's pretty specific case and only applicable with really huge sites, like the adult ones I mentioned.

---

Home > Tracking Campaigns > Tutorials, Tips and Guides