The Unofficial Google Data Science Blog

MARCH 31, 2016

both L1 and L2 penalties; see [8]) which were tuned for test set accuracy (log likelihood). On each of the ten segments the random effects model yielded higher test-set log likelihoods and AUCs, and we display the results in the figure below. Large-scale inference: empirical Bayes methods for estimation, testing, and prediction."

Modeling 40

Modeling Statistics Advertising Testing 40

Occam's Razor

MARCH 23, 2010

My problem with these mistruths and FUD is that they result in a ton of practitioners and companies making profoundly sub optimal choices, which in turn results in not just much longer slogs but also spectacular career implosions and the entire web analytics industry suffering. Usually at least a test. This is sad. Usually for free.

Analytics 118

Analytics Consulting Metrics Experimentation 118

Occam's Razor

FEBRUARY 22, 2010

In May 2010 (!). The secret to making optimal use of CI data lies in one single realization: You must ensure you understand how the data you are analyzing is collected. It will probably be the category that will grow the most because frankly in context others look rather sub optimal. The Optimal Competitive Analysis Process.

Metrics 123

Metrics Advertising Reporting Software 123

Occam's Razor

FEBRUARY 20, 2012

It will be a great reflection of how well thought out their acquisition strategy is, or how sub-optimal it is. ~ If it stinks that tells me a lot (remember the year of mobile was 2010!). Overpimping of social icons when there has never been a social post (or all posts are sub-optimal). ~ Aw, come on!

IT 138

IT Measurement Marketing Metrics 138

The Unofficial Google Data Science Blog

OCTOBER 12, 2017

Yet when we use these tools to explore data and look for anomalies or interesting features, we are implicitly formulating and testing hypotheses after we have observed the outcomes. We must correct for multiple hypothesis tests. In addition to the issues above, does the conclusion pass the smell test?

Experimentation 61

Experimentation Testing Statistics Metrics 61

CIO Business Intelligence

APRIL 12, 2022

Perforce already owns development tools such as Helix and the testing tools, including Perfecto and BlazeMeter. It will add it to its Spot by NetApp portfolio, the collection of SaaS tools built around the cloud management and cost optimization company it bought in 2022. Microsoft buys Minit to optimize process automation.

Technology 143

Technology Enterprise Consulting Software 143

O'Reilly on Data

JANUARY 25, 2021

Both SRE and DevOps emphasize similar practices: version control (62% growth for GitHub, and 48% for Git), testing (high usage, though no year-over-year growth), continuous deployment (down 20%), monitoring (up 9%), and observability (up 128%). It’s particularly difficult if testing includes issues like fairness and bias.

Machine Learning 362

Machine Learning Software Testing Technology 362

datapine

APRIL 16, 2019

It enables companies to create new business opportunities, generate more revenue, predict future trends, optimize current operational efforts, and produce actionable insights. Optimism Bias – Making decisions based on the belief that the future will be much better than the past. The proof is in the numbers.

Data-driven 287

Data-driven Dashboards Visualization Cost-Benefit 287