On alert fatigue 

Illustration: A wolf https://unsplash.com/photos/9rloii_qmmw

I developed an anomaly detection system for Automattic internal dashboard. When presenting this system (“When good enough is just good enough“), I used to tell that in our particular case, the cost of false alerts was almost zero. I used to explain this claim by the fact that no automatic decisions were made based on the alerts, and that the only subscribers of the alert messages were a limited group of colleagues of mine. Automattic CFO, Stu West, who was the biggest stakeholder in this project, asked me not to stop claiming the “zero cost” claim. When the CFO of the company you work for asks you to do something, you comply. So, I stopped saying “zero cost” but I still listed the error costs as a problem I can safely ignore for the time being. I didn’t fully believe Stu, which is evident from the speaker notes of my presentation deck:

 

Screenshot of the presentation speaker notes.
My speaker notes. Note how “error costs” was the first problem I dismissed.

 

I recalled about Stu’s request to stop talking about “zero cost” of false alerts today. Today, I noticed more than 10 unread messages in the Slack channel that receives my anomaly alerts. The oldest unread message was two weeks old. The only reason this could happen is that I stopped caring about the alerts because there were too many of them. I witnessed the classical case of “alert fatigue”, described in “The Boy Who Cried Wolf”, many centuries ago.

The lesson of this story is that there is no such a thing as zero-cost false alarms. Lack of specificity is a serious problem.

Screenshot: me texting Stu that he was right

Feature image by Ray Hennessy

Time Series Analysis: When “Good Enough” is Good Enough

My today’s talk at PyCon Israel in a post format.

Data for Breakfast

Being highly professional, many data scientists strive toward the best results possible from a practical perspective. However, let’s face it, in many cases, nobody cares about the neat and elegant models you’ve built. In these cases, fast deployment is pivotal for the adoption of your work — especially if you’re the only one who’s aware of the problem you’re trying to solve.

This is exactly the situation in which I recently found myself. I had the opportunity to touch an unutilized source of complex data, but I knew that I only had a limited time to demonstrate the utility of this data source. While working, I realized it’s not enough that people KNOW about the solution, I had to make sure that people would NEED it. That is why I sacrificed modeling accuracy to create the simplest solution possible. I also had to create a RESTful API server, a visualization…

View original post 1,412 more words

Come to PyData at the Bar Ilan University to hear me talking about anomaly detection

On June 12th, I’ll be talking about anomaly detection and future forecasting when “good enough” is good enough. This lecture is a part of PyCon Israel that takes place between June 11 and 14 in the Bar Ilan University. The conference agenda is very impressive. If “python” or “data” are parts of your professional life, come to this conference!