Welcome to part two of a three-part series in which we look at the impacts of COVID-19 on digital commerce and fraud patterns, and insights our data provides on identifying fraudulent behavior.
In our last blog, Where have all the Fraudsters Gone? we looked at data that showed overall ecommerce shopping trends are up, reflecting increased shopping activity that is more than typical seasonality. It’s even more than a typical holiday season from Black Friday to Christmas. But what does this mean for fraud and your business?
For this analysis, the data we examined included the trends for our global Confidence Score. Our score provides a risk assessment of a transaction by leveraging the millions of patterns across our network and the five billion global links connecting names, phones, addresses, email and IP in our Identity Graph. The Confidence Score ranges from 0-500 – a higher number indicates a riskier transaction, and a lower number indicates a less risky transaction. See the following examples:
Sample of a low score – indicating low risk
Sample of a high score – indicating a higher risk
Risk Scores are Trending Down
The Product team analyzed digital identity information and shopper activity patterns from January through the end of April to discern trends. The analysis showed a distinct increase in new online shoppers with very low risk scores. Overall, Confidence Scores were going down – indicating that risk scores are dropping.
Global lockdowns and stay-at-home orders have forced more people to shop online. While many ‘new’ shoppers are appearing across the globe – data insights and machine learning tell us they are low-risk customers. Because Ekata has an established data set of over 20 years, we can connect the new email to an address and/or phone that is long established, as well as an IP that is low risk. By linking these attributes and insights, we can see that many of these increased users are in fact good customers whose behaviors are changing.
When you look across the data, the trends, and the volume, it has very similar patterns to the holiday shopping season – the surge of good customers dwarf the fraud attempts. Similar patterns appear on Black Friday, and the weeks leading up to Christmas.
Additional Data Insights
Let’s look a little deeper into a few more data elements that play into Confidence Score and assessing risk, such as email data trends. Fraudsters usually burn through email addresses, and recently created email addresses tend to have a higher risk than established emails. Data elements with longevity and history tend to be less risky. In the chart below, we will look at how long ago an email address was seen as well as when the email address was first seen.
First, let’s look at the dark purple bars – email last seen data. This metric tells us how many days it’s been since the last time we saw a particular email within our customer network. As you can see, this metric jumps up more than in December. This means we have seen an email before, but it has been a while. You can see it jumps up by around 2.5 times from February to March.
Next, the lighter purple bars show us email first seen data, meaning how long ago we first saw an email address for the first time. Here the delta from December is well over 100 days longer, meaning we see shoppers that don’t often shop online, sometimes not even during the holidays.
These scores, signals, and data insights tell you about the online shopper and how to assess risk, but how exactly can you protect your business and still capture the good shoppers? In the next blog post, we will look at options to protect your business and good customers.