Shares of Riskified (NYSE:RSKD) are down 45% over the past month and a whopping 72% from their all-time high, despite being on the market for just four months. This price drop was partially due to the IPO hype wearing off as other exciting IPOs came to the market, but the chief source of this pressure was a poor third-quarter performance. 

For investors who don’t know Riskified well, it looked like a solid quarter. And to some extent, it was. However, one crucial metric closely tied to Riskified’s investment thesis dropped significantly, which understandably spooked investors. After this immense drop, is now the right time to buy Riskified on the dip? Let’s find out. 

Why I love Riskified

Since its IPO, Riskified has been an attractive company to me. It uses artificial intelligence (AI) to detect fraudulent transactions for e-commerce orders. Unfortunately, fraud can frequently look like real orders, while authentic orders can look suspicious. In this case, many e-commerce companies have to make a tough decision: potentially losing their product to a fake order or maybe turning down a real customer, making that customer angry. Naturally, E-commerce companies do not want to handle decisions like this, especially when shipping out millions of orders a year. So they hand this difficult investigation off to Riskified instead. 

Companies that do this in-house might have to ask the customer questions or take other time-consuming measures to determine fraud, but Riskified’s AI can make that decision in seconds. This speedy resolution not only gives the company more time to handle business-critical tasks, but also improves customer satisfaction for the e-commerce company. As a result, Riskified brings a better customer experience to its platform, along with saving customers money. Riskified’s top 10 customers save on average 39% in operating expenses and make 8% more sales when using Riskified. 

With benefits like that, it’s not surprising that Riskified is being adopted by major e-commerce companies such as Louis Vuitton and Wayfair. This broad adoption has led to impressive growth: Q3 revenue grew 26% year over year to $52 million, and gross merchandise value that Riskified operates grew 28% to $21 billion. The company’s net loss did increase substantially to $86 million, but this was primarily related to $65 million in IPO expenses. 

My main concern

Q3 growth slightly slowed compared to Q2, but that was not the main concern. What really worried investors was that the gross margin dipped from 52% in Q3 2020 to 45% in Q3 2021. The drop was even greater sequentially: In Q2 2021, the company had gross margin of 59%. In fact, Q3 2021 was the only time that Riskified’s gross margin has dipped below 50% since Q1 2020, and before this quarter, gross margin had been steadily improving. 

What does all of this mean? Riskified’s gross margin primarily consists of a “chargeback guarantee” expense. If its AI ever makes an error in which it approves a fraudulent order, Riskified’s policy is to pay for the lost goods so its customer does not lose any money. While this makes Riskified’s platform extremely appealing to customers, the guarantee can also shoot Riskified in the foot if its AI makes many incorrect decisions. A decreasing gross margin means that the company paid out more in chargebacks this quarter, meaning that its AI was relatively inaccurate.

There are two possible reasons for this increase in chargebacks. First, Riskified’s AI could simply be highly inaccurate, which would be thesis-breaking. Nobody wants to own an AI-based fraud detection company whose software comes to the wrong conclusion far too often.

The second potential reason is not thesis-breaking and would be much more agreeable. Management noted that during Q3, it onboarded new merchants from new geographies and industries, namely cryptocurrency brokerages. Management claims that the AI is currently inaccurate because these geographies and products are new to its system. Riskified says that as more data is gathered about these new industries, its AI will become more accurate over time, and the chargebacks will decline, improving margins again. 

How I am moving forward

I am a fan of Riskified, and management’s reasons do make sense. However, I am not one to simply take a management team’s word as truth. Consider that almost every company nowadays is tossing out the word “AI” in hopes of getting hype; it has become hard to decipher which companies use AI effectively and which are using it to gain attention. I believe that Riskified is one of the former, but its third-quarter report shook my conviction somewhat.

The only way for me to find the truth about Riskified’s AI is to hold my shares and watch the next few quarters like a hawk. If its gross margin remains low over time, I would assume that its AI is not as high-quality as I thought. If its margin improves, however, I can believe that the company was simply getting adjusted to serving new industries. If that happens, I would be more than willing to buy more shares. But until then, I am doing nothing while watching the company closely.