How to win frauds in omnichannel?

Sep 25, 2019, 21 views

It is a fact that the electronic market is increasingly relevant to make up the company's turnover. It is estimated that in 2018 this sales channel will reach R $ 53.5 billion in revenue, a 12% growth in relation to 2017. The permeability of mobile devices and applications, as well as the dissemination of the market places and the capillarization of the logistics network of the companies make the clients delight in the virtual world.

To further improve the consumer buying process, the integration between virtual and physical operation with the customer advances, and also the policies and approaches differences for their multiple channels, now become a unique user experience, whatever the way of purchase.

The customer buys on the site, picks up the product at the store or receives it at home. He can also withdraw in closets closer to him. Or receive from registered couriers on a delivery platform. Does not matter! The omnichannel model came to stay. And with it, there are also opportunities for fraud.

In the US, it is estimated that businesses lost $ 6 billion in e-commerce fraud for fraudulent ownership of their business in 2017. A three-fold increase over the year 2016. In addition, there are studies that indicate that 5% all returns of goods are fraudulent, representing approximately $ 22.5 billion.

Collecting items in physical stores and cabinets increases the chances of these frauds because of the total anonymity of the fraudster, since, in addition to performing the purchases behind a computer or mobile device, he does not have to indicate any address for the delivery of the item . Another aggravating factor is the shorter amount of time between order approval and item delivery. When a potential problem is identified, the purchase has already been circumvented.

The solution to avoid these scams is to analyze the customer's digital footprints to identify the chances of him being a fraudster. The collection of large masses of data, coupled with analysis by a set of artificial intelligence models, as well as by parametric and non-parametric statistical algorithms allow us to trace an individualized profile of each client throughout their shopping experience.

Imagine a user who takes time to write his own name, his own identity and his own address? It is not common for someone not to know your personal data. He may be using data stolen from someone else. But it may also be that he is doing a registration for a relative, who does not understand much of computer. What if he is performing this registration on a pre-paid, newly purchased and unused cell phone? The chances of fraud increase.

What if the shipping address is in a state or country other than the geolocation of your device? Or is the registered email a random combination of letters and numbers, as if it was just created? These are some footprints that, together, can identify a fraudster at the time of his registration.

At first, these results may be false-positive. But if there is a machine learning model, which is machine learning, feedback with the result of that analysis over and over, the level of accuracy can be highly reliable.

Related Articles