Voice of the Industry

Automation: an enabler to run a marathon or sprint

Wednesday 17 February 2021 07:38 CET | Editor: Simona Negru | Voice of the industry

Catherine Tong, STRATGranat, explains why the automation of fraud decisions is needed to drive accuracy and why companies will always benefit from having an automated backup plan. 

2020 has been the year where we have all been asked to become athletes, as speed, agility, and focus have been important in our professional lives – also discipline and determination in many of our personal lives – but in general, drive and resilience have been important to keep us going, as well. 

In ecommerce, speed and agility have long been capabilities which most businesses have strived for. However, what has taken many companies by surprise is that what felt like speed before has now been turbo-powered and projects expected to take 12-18 months have been delivered in the last 3 – proof that with the right focus and collaboration, great things can happen. Never before have companies been required to embrace digital transformation so aggressively to at least try and keep up. 

In the fraud and payments space, this has certainly been true, particularly in industries such as retail, where volumes have moved from physical stores to ecommerce, and many new pressures are faced with one common theme – the need to automate. Businesses have simply not had the time and often money to increase or retrain people, or to change systems, thus those who already had a strong sense of automation have been more agile and able to adapt. 

For fraud and payments, machine learning is often at the heart of automation 

The automation of fraud decisions is made up of various algorithms making the binary decision of whether customers are able to complete their purchase or not. It is theoretically possible to fully automate this process, however, many would argue that a human is needed to drive accuracy in the true ‘grey areas’, through a manual review or as a sense check where purchasing behaviours suddenly change. In this instance, automated machine learning does most of the heavy lifting, and gives a company the ability to scale the human element through new hires or retraining existing staff, meaning continued employment for those who would otherwise have lost their jobs. The manual review team and fraud manager play an important role as they remain the eyes and ears of sudden fraud attacks, but their scale is likely to adjust, as machine learning removes more of the genuine transactions from analyst workloads, allowing them to focus on the true grey areas, but also to manage alerts triggered by the machine learning if there are spikes in decisions not previously seen – the ‘sense check’. 

Understanding the type of models used and how they behave will also impact the ability to automate. Some machine learning models themselves have also been subject to a decrease in accuracy and therefore required retraining. Other models, which avoid the concept of ‘over-fitting’, will naturally adjust to these changes, but require a human ‘sense check’ of what is going on in the wider industry or world. Again, the human touch should not be entirely eliminated. 

In payments, automation is more complex as it is not a binary decision. For those who leverage machine learning to route their payments traffic, they want to optimise acceptance rates and minimise their processing costs, by moving quickly and with minimal human intervention.

This means that there has to be some human intervention and -although some tools are marketed as giving automation through machine learning -they are in fact human-dependent rules in the background (to manage volumes sent to different vendors, for example). 

For both fraud and payments, companies will always benefit from having an automated backup plan. For example, for payment processing outages, an automated backup plan will be having the ability to automatically switch payments traffic to a new provider and ensure that customers and associated revenue are not lost. 

Testing out new features and functionalities also offer a great use case for automation 

With A/B or multivariate testing, making adjustments to the checkout page or payment offers can take a lot of time and effort to manage effective samples and reporting. Now that there are many tools on the market, once the hypothesis is clearly defined, the variations can be left to run so that the statistical analysis will lead to data-driven results and the business can focus on the outputs and decision-making to optimise their customer experience and conversion. 

Automation does not remove the role of people altogether, as done well, it should empower people to be more focused and effective. However, the bias people can bring may damage company performance, as well as the fact that people have their own motivations and ways to work. Therefore, positioned well, these will propel a business forwards, while positioned badly, you could end up with empire builders and parochial views, which slow a business down and cause duplication. Machines do as they are told, but a machine is only as good as its inputs. If we relied fully on these services, over time, they would deteriorate as they fail to take into account changing business models, consumer purchase patterns, and other innovations. Consequently, having some specialist manual intervention, at least from time-to-time, is still important. 

In recent months, we had to adapt our skills to be able to switch from marathons to sprints and back again, but it has proven that automation supports this switch and those who have mastered the collaboration between people and automation are winning. Those who have found an appropriate balance are also those who are more likely to succeed in the long-run as people can run a marathon, but not multi-marathons back-to-back, and they certainly can’t sprint for as long as a marathon. Having the right automation to support the journey will mean that eventually they will go further. Do you have the right automation?

This editorial was published in the Fraud Prevention in Ecommerce Report 2020/2021, the go-to source in securing transactions while offering a frictionless customer journey.

About Catherine Tong

Catherine is an independent fraud, risk, and payment specialist with STRATGranat. Previous projects include complex fraud investigations, Global internal audit programmes, and setting up and managing ecommerce fraud and payment teams. She has also served as a European Board member for the MRC and is currently an Ambassador for the European Women in Payments associations.


About STRATGranat

STRATGranat aims to be the Payment sector service company. It supports fast growing payment sector companies with a very horizontal portfolio of 100 packaged services, Strategic, Operational or HR related. Its credibility is based on 40 experts with more than 15 years of experience in their fields, who have worked for 100 tier-1 companies. 


Free Headlines in your E-mail

Every day we send out a free e-mail with the most important headlines of the last 24 hours.

Subscribe now

Keywords: Catherine Tong, STRATGranat, ecommerce, digital transformation, fraud, automation, machine learning, acceptance rates, checkout, data, retail
Categories: Securing Transactions | Digital Identity, Security & Online Fraud
Countries: World
This article is part of category

Securing Transactions