The Benefits of Labeling
By Emily Chin /
5 Jun 2015
Innovator. Leader. Influencer. Awesome-sauce. Yes, these are some of the labels that Sift Science gets, but not quite what I’m talking about. At Sift, labeling has pretty magical powers.
Pop quiz: how does machine learning learn? Good news, we have a study guide for you in the form of our new ebook on machine learning. Have a look and come on back when you’re ready.

Are you back from studying up? So you should now know that machine learning models learn by the process of users providing feedback (the act stage) to once again train the model. Every time you act, you’re labeling the data and further training our system to detect fraud for you.
Perhaps you’ve heard of labeling? At Sift Science, we employ a simple system of Bad or Not Bad in order to train the data to recognize fraudulent behavior.
But why does labeling matter? How does it really help?
- It improves accuracy
- It improves accuracy
- It lets us build a custom fraud prediction model for you
- Did we mention that it improves accuracy?
Yes, labeling is what drives Sift Science’s incredible accuracy. Because the global machine learning model reflects all customer data, tailoring the model to your specific website, app, or business’ needs is essential. Providing and labeling data is what customizes the experience, allowing for tailor-made predictions. When you label a user as bad, this training helps your fraud prediction model figure out the subtle and not-so-subtle traits and behaviors that distinguish a fraudster.
Labeling doesn’t just help you catch fraudsters, but can also reduce friction for good customers. By labeling known good customers as good, you can then whitelist trusted users. If you know for certain that a particular set of users are repeat, good users, you can reduce information fields or identify verification steps, simplifying their interaction with your product and encouraging future engagement!
Labeling is the name of the game in Sift Science’s machine learning system. If you’re not yet labeling regularly, your fraud detection solution has the potential to become even more powerful. To learn more about labeling, feel free to check out our website and send us questions at scientists@siftscience.com.
Related
Emily Chin
Emily Chin was a manager at Sift.