Artificial intelligence
How to use machine learning in ad systems? To search for traffic anomalies and bots. Our clients begin to save money within 24h after the initial setup – eliminating budget waste and increasing your return on ad spend.

How to use machine learning in ad systems? To search for traffic anomalies and bots. Our clients begin to save money within 24h after the initial setup – eliminating budget waste and increasing your return on ad spend.
All our research on bot identification with machine learning algorithms has been published to the public and is available at the link below. Of course, we are constantly improving algorithms, and if you are ready to dive into the world of mathematics, click on the button below.
A global database of bots known to us
Visits analyzed daily
Daily savings for our customers
Experts in the team
Imagine that a bot came to your website. The bot has a large set of unique parameters, such as IP address, session duration, number of page views, traffic source, browser screen resolution, etc. We collect all these parameters and use them to train a neural network. The more data there is, the more accurately the neural network will identify human mimicking bots.
k-nearest neighbors algorithm (k-NN) is a metric algorithm for automatically classifying objects or regressing objects closest to it whose values are already known. Sounds too complicated? Let’s break it down in simple English.
We collect an extensive characteristics database for each bot visit to any of our clients’ websites. The more data we collect, the better.
Prepare data. Group similar visits and determine groups that bots are more likely to fall into.
Block according to the received data of users that are very similar to bots.
Identifying bots in the traffic flow on your website is challenging since bots are constantly improving their algorithms, trying not to differ from people, which is why our defense consists of several stages. For instance, we analyze mouse movements.
It is challenging for bots to repeat human movements with a mouse on a particular website because each site has its navigation elements. In contrast, bots work according to a predetermined algorithm. If we know that there is a bot on the website with a 99% probability, we will record its mouse movements and train the neural network to find bots by similar moves. The more extensive our bot database, the more accurate the algorithms work!
HOW DOES IT WORK?We develop and improve all the click fraud protection algorithms on our own, and we don’t license other people’s developments.
We have 7 experts working in our service: mathematician, developer, project manager, tester, support technician, analyst, and boss 🙂
We regularly publish our research and experience on fighting against click fraud in the media, although sometimes we have to defend ourselves from negative commentators.
We accumulated a vast amount of data for all our time off work, which we use to train a neural network.
We are ready to provide you with an API that will find bots on your website on the fly so that you can make the logic of their processing.
We have been working on the market for several years now. Please, do not expect miracles from the protection system. We are still working on it every day, trying to keep up with fraud clickers.