BNPL - Buy Now Pay Later

The trend coming from the USA will force the incorporation of AI in the process.

It is probably only recently that we have begun to hear of a new concept in the financial world: the BNPL - Buy Now Pay Later. It seems a rather unoriginal term, but for years financiers have been offering the possibility of financing our purchases in installments with interest or even with a 0% interest rate.

The novelty may be that in order for them to give you this financing they do not make any kind of study, no friction is generated in the digital purchase by the consumer.

This is what we want to achieve on behalf of the retailersIncrease sales, and reduce customer abandonment of the purchasing process.

The pandemic has led to an exponential growth in online shopping and in particular BNPL, especially in the U.S. market. In this market during 2021 more than 45.1 million consumers used BNPL, growing by an impressive 81.2% in that year. It is anticipated that by 2024 4.5% of all eCommerce in North America will be done by this form of payment. In some sectors such as travel, the use of BNPL has become the number one means of payment for the millennialsgeneration X and generation Z; always referring to the American market.

Accordingly, it is not surprising that in our country more banks and financial institutions are moving to offer to the retailers BNPL financing services. The problem, as can already be sensed, is the control of fraud and fraudsters.

In addition, this fraud must be controlled in the most transparent way possible from the consumer's point of view; a balance must be struck so as not to generate friction in the purchasing process that would break with one of the main objectives of the Buy Now Pay LaterMake the purchase very easy, decreasing the customer's abandonment rate.

Artificial intelligence (IA) provides us with the tools to "anticipate" new types of fraud, to switch from rule-based and therefore reactive systems to proactive systems. The big "players"see this technology as fundamental to their business: according to  Afterpay AI allows them to adjust the search to new vulnerabilities, which makes us highly effective against fraudsters.

Afterpay says that fraud remained below 1% in 2020; while Klarna claims that the protection in its Marketplace exceeds that offered by credit cards and large banks. This "success" in fraud control is due to AI's ability to infer new knowledge from known data.

BNPL must force a change in the way in which risk has been managed in financial institutions; we can no longer rely on reactive systems, where we wait to identify a fraud according to the match of a series of rules with the new purchase or financing request. AI can anticipate new behaviors and, at least, warn that there are operations with a certain probability of being fraudulent.

The latter can be complemented by "..." systems.adaptive onboarding"These modify the customer-to-customer sales method depending, among other factors, on the customer's risk. 

This opens up another potential use of AI in BNPL eCommerce, not only to control risk, but also to know the customer's propensity to buy, in fact, to know which are the best products and under what financing conditions for each customer. The ultimate goal is to maximize the conversion of eCommerce visits with maximum security.

You may think that in a BNPL model there is not enough information to train the models by machine learning, but in reality this does not have to be the case. This is a model designed for digital channels, and the use of these channels can be traced perfectly, you can know for each customer what they have done before choosing a particular product and buying it.

The digital channel has the advantage of speed and immediacy of operations, but we must have the technology to try to take advantage of the opportunities, minimizing the risks; and this technology cannot be based on the classic systems of segmentation or scoring. Solutions must be developed based on deep learning capabilities (Deep-learning)solutions capable of self-adjusting to new behaviors known through data.

At Gamco we have been developing machine learning models for risk management for years. We have had production deployments of systems during the pandemic and early confinement. The predictive models were created with pre-COVID credit customer behaviors, but thanks to self-learning, they self-adjusted to the new reality as of the second half of March 2020. 

On the other hand, Gamco has developed solutions for the commercial development of digital channels, estimating the potential of each customer, which products can develop that potential and which levers maximize the probability of conversion (moment, message, cross-promotion, discount or value-added promotion). adaptive onboarding to facilitate conversion, while securing transactions.