Have you ever accidentally labeled a user as “Bad” or “Not Bad”? Or perhaps further investigation on a user left you wishing you could undo the label you initially selected? If this describes you – you aren’t alone! In response to popular demand, we’re excited to introduce the new and amazing “unlabeling” feature. Starting today, you can fix your labels quickly and easily.  

Why does accurate labeling matter?  

When you mark a user as “Bad” or “Not Bad”, you are training Sift’s advanced machine learning system to better find and predict fraud for your business. For example, if you’ve mistakenly labeled a legitimate and good user as “Bad”, then Sift will learn incorrectly and may mistakenly identify other good users as potentially fraudulent.

We understand that labeling errors do happen and that people change their minds. After all, we’re all only human.

This new feature lets you remove the “Bad” or “Not Bad” labels for a user – think of it as the undo button for your fraud team! With unlabeling, you don’t have to worry because it’s easier than ever to focus on fighting fraud.

Best of all? Unlabeling is available today! You can either label directly in the console (as shown in the GIF below) or use our Labels API.

New to labeling? No worries – read about it here and start labeling today!

Happy Labeling (and Unlabeling)!

  1. nice feature!

    it would also be nice to have a third "suspicious" label, for when we’ve seen a trxn or user and are doing further research on our own.

    For purposes of your machine learning, it could behave like an unlabeled visitor, but leaving suspicious users as unlabeled makes them hard to distinguish from items we’ve never looked at…