Every online retailer is likely familiar with the painful wake-up call of chargebacks. “There is nothing worse than coming down off of a huuuggge month just to get slapped with chargebacks,” explains Derek Thompson, Director of E-Commerce at Fontana Sports. But in addition to payment fraud, a lot of online retailers are also battling content abuse in the form of fake reviews. We talked to Derek about how he uses Sift Science to fight both types of bad behavior.
Tell us about Fontana Sports…
Fontana Sports is an outdoor specialty retailer with over 5 decades of history in Madison, WI. We specialize in gear and apparel for outdoor adventure sports, including Skiing,Snowboarding,
We have two stores in Madison, in addition to our online presence. Madison is the second largest metro area in Wisconsin, and has a strong history as a college town. It’s a community with a lot of people that value nature and outdoor recreation. In our local market, we find that our customers are looking for expert advice and personal service not offered by the national chains we compete with. Online, our customers are significantly more price-sensitive, although we also do well with long-tail search and obscure products.
Our main competitors in the Madison area are REI and a couple of the other national chains. Our largest competitors online are Backcountry and Moosejaw. There are definitely some big players in our space. One trend that we’ve seen over the past few years is an increase in brands offering direct to consumer sales. This definitely started online, but recently the allure of manufacturer-level margins has been causing several of our large vendors to open their own brick and mortar locations to compete with us.
What type of fraud do you face?
We initially decided to use Sift Science to help detect credit card fraud. We see a lot of stolen credit card usage on our site, which means chargebacks. Unfortunately, the payment industry places all of the liability for credit card fraud on merchants, and then turns a deaf ear when it comes to helping us manage that risk. Chargebacks and false declines both cost money, not to mention the time investment in trying to screen for fraudulent orders. We wanted to use Sift Science to reduce our chargeback risk and reduce the incidence of false declines.
Here are a few considerations that made the decision to use Sift Science easy:
- We’re a small business, and we have a small staff. I like things that can be automated, programmed, and save us time. Sift Science does this.
- I believe that more data begets a better algorithm, which will usually trump human bias and intuition. Sift Science seems to believe this as well.
- I’m philosophically opposed to insurance. I argue that you only insure something if the potential loss will put you out of business. We don’t insure our shipments because in the long run, it costs more than underwriting the risk ourselves. If I don’t insure a shipment against loss, why would I insure the payment for that shipment against loss. A few of Sift Science’s competitors offer chargeback insurance, which is not what I was looking for. I don’t want to pay a premium for insurance. Sift Science offered a technology product. I like technology products.
Tell us about your overall experience with Sift Science…
Getting started was easy. Between reading the documentation and doing the actual programming, it took us about 30 hours to integrate. That’s about 20 hours of integration with our website to get the data over to Sift Science, and about 10 hours to integrate it with our fulfillment work flow. Within a month, we were getting really accurate results. It helped that we integrated during our busiest season, so we were able to provide enough data to spin the system up pretty fast.
There are tons of great benefits of using Sift Science. Because it’s so accurate, 90% of our orders are automatically approved without any sort of manual review. This translates in to faster order fulfillment, decreased employee fatigue and burnout, and more attentiveness on the orders that actually require manual review. You can read more about Fontana Sports’ experience with Sift Science on my blog.
You also use Sift Science to prevent fake reviews. How did that come about?
I used to work at a very large SaaS E-Commerce reviews software vendor, so I know better than anyone the importance of product reviews as a sales tool. Fontana Sports is small enough that human curation is an effective way for us to screen reviews for relevance and spam; that being said, I’m intricately familiar with the need to leverage these types of automated systems to detect review fraud when it comes to much larger applications.
The way that we came to integrate Sift Science with our product review approval process is kind of an interesting story. One morning, I received an email notifying me that we had received “a bunch of reviews that look like gibberish or some kind of code.” Interestingly enough, there were about 250 reviews full of words like “update,” “insert,” “select”… I think you can see where this is going.
At this point, we were already using Sift Science to screen for fraudulent orders, and had automated part of the ordering process to impose or reduce friction for customers based on their risk level. One of the custom signals that we were providing to Sift Science was the customer’s history of writing product reviews. I realized that at the point the customer was writing the review, Sift Science already had enough information to provide a risk profile for us – so it only made sense to use that information to decide how we would handle the customer’s review. Using some pretty basic logic, we broke reviewers into three tiers:
- Low-risk customers are able to submit a review which is published immediately.
- Medium-risk customers are able to submit a review, but the review is put into “purgatory” and must be approved by our staff before it is published.
- Very high-risk customers reviews are not saved. This isn’t a common occurrence, but there’s no reason to chew up disk space when a spam-bot gets trapped in a loop on one of our product pages.
That way, we’re able to automate how we manage a few different types of fraud, using the same Sift Science platform.