The Unofficial Google Data Science Blog

APRIL 23, 2024

If the relationship of $X$ to $Y$ can be approximated as quadratic (or any polynomial), the objective and constraints as linear in $Y$, then there is a way to express the optimization as a quadratically constrained quadratic program (QCQP). However, joint optimization is possible by increasing both $x_1$ and $x_2$ at the same time.

Experimentation 52

Experimentation Optimization Uncertainty Metrics 52

The Unofficial Google Data Science Blog

JULY 22, 2020

Instead, we focus on the case where an experimenter has decided to run a full traffic ramp-up experiment and wants to use the data from all of the epochs in the analysis. When there are changing assignment weights and time-based confounders, this complication must be considered either in the analysis or the experimental design.

Experimentation 95

Experimentation Statistics Testing Strategy 95

The Unofficial Google Data Science Blog

OCTOBER 7, 2015

This is essentially the same as finding a truly useful objective to optimize. accounting for effects "orthogonal" to the randomization used in experimentation. Recently, we presented some basic insights from our effort to measure and predict long-term effects at KDD 2015 [1].

Modeling 61

Modeling Experimentation Knowledge Discovery Testing 61

Occam's Razor

SEPTEMBER 15, 2014

Company UX leaders are happy to stink less by taking the sub-optimal path of responsive design, rather than create a mobile-unique experience (your customers tend to do different things on your desktop site than your mobile site!). Media-Mix Modeling/Experimentation. Media-Mix Modeling/Experimentation. Many reasons.

Metrics 141

Metrics Reporting Analytics Marketing 141