Get the essential data observability guide
Download this guide to learn:
What is data observability?
4 pillars of data observability
How to evaluate platforms
Common mistakes to avoid
The ROI of data observability
Unlock now
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Sign up for a free data observability workshop today.
Assess your company's data health and learn how to start monitoring your entire data stack.
Book free workshop
Sign up for news, updates, and events
Subscribe for free
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Getting started with Data Observability Guide

Make a plan to implement data observability across your company’s entire data stack

Download for free
Book a data observability workshop with an expert.

Assess your company's data health and learn how to start monitoring your entire data stack.

Book free workshop

Percentage Model Enhancements (March 2024)

and
March 28, 2024

Founding Data Scientist @ Metaplane

March 28, 2024
Percentage Model Enhancements (March 2024)

We’re improving the ranges of our monitors that calculate percentages. All nullness, uniqueness, and percent zero/negative monitors will see these improvements roll out over the coming weeks. The expected ranges of these monitors may change.

Handling 0% and 100%

Previously, monitors with high volatility would sometimes not alert when a metric went to 0% or 100% because they were within the expected range, even if those values had never been observed in the past. Many cases should alert at these extreme values - imagine a NULLNESS check on a column going from 97% (some data) to 100% (no data). Ranges will no longer include 0% or 100% unless they were observed during training or marked as normal at some point in the monitor’s history.

Smaller Ranges

Ranges will now tend to be smaller so you get alerts on changes that are relatively large - such as a jump from 1% to 2% - even if they’re not large on an absolute scale. The ranges will also shrink as more new values are observed and model confidence increases.

Table of contents
    Tags

    We’re hard at work helping you improve trust in your data in less time than ever. We promise to send a maximum of 1 update email per week.

    Your email
    No items found.
    Ensure trust in data

    Start monitoring your data in minutes.

    Connect your warehouse and start generating a baseline in less than 10 minutes. Start for free, no credit-card required.