Super-fast Machine Learning to Production with BigQuery ML
How to use Bigquery ML to deploy your models in no time, and focus on what really matters.
A few months ago at Agorapulse, we kickstarted two new Machine Learning based projects. Since we’d been using BigQuery for almost 2 years, we had to give BQML a shot. Here is what we learned, and what you could learn from it!
The biggest challenges might not be what you think
As a Machine Learning engineer, you usually think about How to create the best performing model, What algorithm to use, or How to optimize the model to go from a 94% accuracy to 96%.
But if your company is at the beginning of its Machine Learning journey, it’s probably not the most important part of the process. At least not yet.
— It surely is if you’re training a self-driving car and dealing with human lives, but not for the rest of us.
First, you’ll have to face 2 under-mentioned challenges in Machine Learning.
1. Adoption & Culture Shift
In this article published by Harvard Business Review in 2019, authors say “Technology isn’t the biggest challenge. Culture is.”
When your Marketing or Product team asks for a specific insight or analysis, you can freely focus on the technical part (eg: data analysis). Since you’re replying to a question they already have, adoption shouldn’t be a problem.
But Machine Learning, for people not used to it, is perceived as a “black box”: What’s under the hood? And what can it do for me?
When launching a machine learning project, you will probably have back and forth with the users, to help on the adoption, improve interpretability, or iterate on the algorithm itself to improve the result. Adopting ML takes time.
In one of our surveys nearly 90% of the companies that had engaged in successful scaling practices had spent more than half of their analytics budgets on activities that drove adoption, such as workflow redesign, communication, and training.
— Harvard Business Review — Building the AI-Powered Organization
