UPD August 7, 2026. I left the original text below as it was: it is a snapshot of the 2014 CPI market and my reaction to a particular post and podcast. But the meaning of the term has spread out over twelve years, so first I will briefly pin down what people usually mean by fraud now.
Fraud in affiliate traffic is the deliberate creation, substitution, or appropriation of advertising events: impressions, clicks, leads, installs, purchases. The goal is simple: get paid for something that never happened, claim someone else’s conversion, or pass one traffic source off as another. Low CR, poor retention, or allowed incent are not fraud by themselves. Fraud begins when the source, events, or traffic quality are deliberately misrepresented to get paid or bypass an offer’s rules.
Fraud, Incent, and Bad Traffic: Where Is the Line?
These are all slightly different: incent, traffic that does not convert, and fraud. All of them can lose money for the advertiser, but their causes and consequences differ.
- Real traffic means a user became interested in the ad and completed an action without a separate reward.
- Allowed incent means the user gets a perk for taking an action, and the advertiser knows and permits it. If the offer accepts incentivized traffic, the source is disclosed, and the advertiser knows what they are getting, then of course it is not fraud.
- Low-quality or invalid traffic means events the advertising system should not count: accidental and repeated clicks, garbage, and artificial activity. Invalid traffic is therefore broader than fraud. You can see that even in the Google Ads definition.
- Fraud is deliberate falsification or concealment: bots instead of people, incentivized traffic disguised as non-incent, fake events, banner stacking, although that comes from a slightly different angle, or stealing someone else’s attribution through cookie stuffing and other mutations of the same approach.
If the advertiser understands where the user came from, it is an agreed buying model. Promise one thing and slip them another, and that is fraud.
What Mobile Fraud Looks Like Now
In 2014, the main horror story was mixes: some real traffic, some incent, and some outright trash. Mixes have not gone anywhere, but today far more schemes are called fraud:
- bot impressions and clicks that imitate interest in an ad,
- fake installs, registrations, deposits, and post-install events,
- click spam and click flooding, where a source sends clicks hoping to intercept an organic install by chance,
- click injection, where a click is inserted immediately before a real install,
- SDK spoofing and forged server events,
- multiple accounts and substitution of the source, device, GEO, or other data that determines the payout.
Here is the interesting part: a real user and a real install do not mean the attribution is honest. Someone can simply take another person’s payout.
How Advertisers and Affiliate Networks Find Fraud
I once helped build mobile antifraud, by the way. It was quite an adventure, with AI and all the blackjack, before all of that became mainstream.
There is no single magic metric. Different types of fraud are caught in different ways, or through a whole combination of factors plus behavior. CR proves nothing by itself, and neither does one IP, one phone model, or a short click-to-install time. People look at combinations of signals and cohort behavior:
- the distribution of time between impression, click, install, and target event,
- repeated devices, IPs, payment data, and other IDs,
- mismatches between GEO, carrier, time zone, language, and the declared source,
- an abnormal share of new devices or identical action sequences,
- retention, purchases, repeat deposits, and other post-install events,
- rejects, refunds, chargebacks, and discrepancies between the tracker, affiliate network, and advertiser.
Some things are visible in the first seconds, some in the first hours. Some appear only on D7, D30, or after refunds. So the universal promise that all fraud will be detected within a few hours is, well, not much of a promise. Some will be detected, and some will not.
What an Affiliate Should Check Before Launch
- Read the offer rules and the list of allowed sources before sending traffic, not after the hold.
- Get written confirmation from the manager if you use incent, an unusual prelander, or a mixed source.
- Pass proper sub_ids, keep your own logs, and reconcile postback.
- Start with a small cap and look not only at CR, but also at retention, revenue, rejects, and charges.
- Do not disguise the source. Well, you know. At your own risk, as always.
Bad economics does not always mean fraud. First separate poor quality from a technical rejection, and a rejection from shaving (in Russian).
Is anything valuable left in the 2014 text? Absolutely: the model buy incent, dilute it with good traffic, add trash, and dump it into a network does not become legitimate merely because it contains a few real users. There are still markets where it works, though. The unchanged original text is below.
Original Text from 2014
A lot of people wrote to me about the Habr post, the dark side, blah blah blah. They asked what I thought. I read the post and listened to the podcast. My impression is very mixed.
First, define the terms:
- Real traffic is exactly that: a real user who clicked a creative because they were interested.
- Incent is incentivized traffic: the user gets a perk for completing an install.
- Black traffic is a mixture of incentivized and cheap traffic that will never convert.
- A mix, or fraud (in Russian) in this context, is incent plus real traffic plus traffic that cannot convert. Incent is n%, and not very much.
Look up unfamiliar terms in the CPA Glossary.
I am writing based on the information I have. I do not have nearly all of it. As for the models he described, nothing strays far from the truth. A lot was left unsaid. Numbers, estimates, and volumes are a different question.
Are there problems with mobile tracking?
Yes. They are not as scary as he claims, but they exist. Or rather, it is like this. Tracking has almost no problems; advertisers generally validate installs. Nearly everyone can identify black traffic, and quickly. Some advertisers look only at CR. Some advertisers cannot identify mixes. Some networks cannot and do not want to identify mixes. Some networks know about mixes and are willing to work with them, but very few admit it openly. At the same time, there are advertisers and networks that are very good at finding fraud and applying sanctions, and there are more of those. Advertisers that understand incent and non-incent usually know how to separate and see them.
There are players with different levels of competence. One does not prevent the other. The market is large, wild, and dynamic.
Do people work with schemes like these?
Yes. With a huge number of conditions and caveats, and only with current, highly specific information. You cannot just take a mix and pour it in. They will catch it within a few hours.
If mixers leave CPI, will all traffic disappear, as he claims? Of course not. Not even close. There will be fewer installs, traffic quality will improve, and the cost of an install will rise a little. Again, advertisers are fine with incent for boosts and bursts. There will be no serious change, and if there is, it will affect only a few players that depended heavily on this business.
Are there many blatant mixes, specifically mixes?
Many. I do not have exact data. I heard estimates ranging from 20 to 80 percent. You also have to account for the fact that the share of incent within mixes, or fraud, varies widely. And remember that traffic suppliers are not the only ones cheating. Advertisers cheat too.
Are the daily volumes he talked about possible?
Mainstream traffic, yes. Mixes, yes, but it is hard, especially alone. Undiluted black traffic is highly doubtful, probably not. One caveat: either there are plenty of drops between peaks, or a team is working, also with peaks and drops but smoother ones. In general, I seriously doubt the claims about working alone. There is as much grunt work there as there is shit in Habr comments. Another point: advertisers willing to buy many installs usually validate them. That is precisely why you see a $300 daily cap. To reach $25k per day on mixes, you need enormous connections with a large number of CPI networks.
What is fraud, specifically for an advertiser?
Why are mixes and fraud bad for an advertiser? Because the advertiser pays for installs they can never monetize. There are situations where incentivized traffic is needed: boosts, times when you must get tens of thousands of installs at any cost in a very short period. What makes advertisers sad is getting disguised incentivized traffic passed off as non-incentivized.
Are the global market volumes he named reasonable?
If he means mixes and counts the total mix, yes. If he means only black traffic, that is unlikely, although of course I do not have all the information.
Now, about the people who asked me how easy this is and whether they should try it.
This business ruins your karma. The barrier to entry is very, very, very high. Much higher than in mobile affiliate marketing (in Russian). The model “buy incent, dilute it with good traffic, dilute it with trash, and dump it into a network” will not work. You will get a hold and a ban within hours. There are countless nuances. You need a specific financial setup and complex technical infrastructure. Nobody will give an outsider the same offerwall, for example. You also need current information, which is very hard to obtain once, let alone obtain consistently. People who worked in this area on the web can get in; so can people working in specific areas of mobile development, and people with the right insider knowledge and connections. An outsider probably will not get into it even with a large budget.
As for that particular guy in the post, my personal opinion is that he is a fake, a character. Listening to the podcast made that impression even stronger. What is he, a professional voice actor? Real people do not talk about their work that way. It was like listening to an audio drama. Seriously.
Why was it made?
I do not know. Buzz around the subject, laying the groundwork for an agency PR campaign, a scare story for advertisers? A filthy shitshow overall. No respect.
My overall opinion is that this whole thing is a shitty business that kills the market, and the post itself is a shitbomb that will help no one. It will only breed piles of speculation, fantasies, and flame wars.
I am not going to argue or comment on this subject anymore. I do not like it.
I will not give anyone information that would help them get started in this business.