Everyone who manages an e-commerce has some indexes at the tip of their tongue: number of invoiced orders, average ticket, unique visitors, conversion rate, and so forth.
When you do your risk analysis on the orders, there are specific metrics to know how your operation is doing. See the key indicators below:
Chargeback rate
Chargebacks are orders challenged by the card user, whether by not recognizing the purchase or by a commercial non-compliance. It is calculated by dividing the volume of contestations by the volume of sales received.
This is the most basic indicator on the performance of your risk analysis. The card logos and holders have chargeback limits to the store owners, which, in general, should not exceed 1% revenue. Above that, the store owner has the risk of being warned and, if the problem is not corrected, having their membership suspended and even being prevented from receiving orders.
Manual review rate
When an order is considered suspect, it usually is forwarded to a review table, where an analyst thoroughly inspects it. The manual analysis is important in order to identify false positives, but it is not only financially expensive but also demands client experience, since the order is retained in a queue and takes longer to be sent.
This rate is calculated by dividing the number of orders that go to the review queue by the total orders analyzed. The ideal is that this number is small, since e-commerce has the risk of suspending orders from legit clients while searching for frauds.
The opposite rate is the automatic approval rate: from all orders received, how many are instantly approved by the system.
Rejection rate
The rejection rate is the percentage of orders denied due to fraud suspicion. One must not calculate the orders whose payments were not authorized, since in most cases, there is an issue with the card, and not in the analysis.
One may call this indicator as fraud attempts: how many frauds were started on the site against how many orders were approved.
False positive/False negative
When analyzing an order, one must check if it is a fraud. There are two important definitions from this: false positive and false negative.
The false positive is an order marked as fraud, but which is actually real. The opposite is what is known as false negative, which happens when we mark an order as legit, but in fact it was a fraud.
False negatives are easily measured. When a chargeback arrives, it increases the counting of false negatives. However, the false positive is a bit more complicated. How can one know if the order was legit if it was not accepted or processed?
An alternative is to adopt a more conservative measure: when a legit order goes into manual analysis and then approved. In this case, there is an order that, theoretically, could not have gone through there.
Therefore, when marking an order as suspect, and then approving it in the manual analysis, the count of false positives increases. The indicator can be seen as the number of orders approved after manual analysis.
Cost per analysis
One can calculate the cost of manual and automatic analyses separately, as well as together. It is our recommendation to always calculate both types, so as to have a better idea of the costs involved in each stage of the order analysis.
There are tools that charge a set value per order, regardless of the type of analysis, while others have different values for each phase. The effort of an automatic analysis is much lower than the one for a manual analysis, since in the former, an algorithm or system does all the heavy work, while in the latter, one or more people are involved, manually reviewing the order.
With a single value per order analyzed, there is the risk of paying a high price for orders whose anti-fraud cost is low. Let’s make some simple “bakery” count, playing around with an automatic approval rate for a store with 5,000 orders.
Automatic approval rate: 60%
3,000 orders at $ 0.21 = $ 630 +
2,000 orders at $ 4.50 = $ 9,000 = $ 9,630
5,000 orders at $ 1.80 = $ 9,000
Automatic approval rate: 70%
3,500 orders at $ 0.21 = $ 735 +
1,500 orders at $ 4.50 = $ 6,750 = $ 7,485
5,000 orders at $ 1.80 = $ 9,000
Automatic approval rate: 80%
4,000 orders at $ 0.21 = $ 840 +
1,000 orders at $ 4.50 = $ 4.500 = $ 5,340
5,000 orders at $ 1.80 = $ 9,000
Automatic approval rate: 90%
4500 orders at $ 0.21 = $ 945 +
500 orders at $ 4.50 = $ 2,250 = $ 3,195
5,000 orders at $ 1.80 = $ 9,000
In a more advanced operation, there are other important indicators, almost all connected to the manual analysis team, such as average review time and analyst performance. This subject will be dealt with in another post, but with the indicators above you have a much better view of the health of your operation.
About Konduto
We are a startup developing an innovative technology to bar e-commerce frauds. Our intelligent anti-fraud monitors the client throughout his purchase journey in your site and evaluates the transaction in real time – our answer is given in less than 1s! We detect only the purchases that are really suspicious, approving more orders and reducing the costs with frauds. Send us an e-mail on hi@konduto.com
