Must read!

Articles
Read on the Web
When Will COVID-19 End? Data-driven estimations

This site provides continuous predictive monitoring of COVID-19 developments as a complement to monitoring confirmed cases.

Read on the Web
#GTMTips: Why Does The Google Analytics Tag Show 'Failed'

Problem you might face with Google Analytics: tags that show status Failed, and which refuse to send any data to GA.

Read on the Web
#GTMTips: Use gtag.js Parameters In Google Tag Manager

If you have gtag.js implementation on the site, you can use the events and parameters sent to gtag.js in GTM as well.

Read on the Web
What is BigQuery? - BigQuery Spotlight

Nick Orlove will be showing you all the ins and outs of BigQuery, Google’s fully-managed data warehouse.

Read on the Web
Google Analytics eCommerce Reports: Improve CVR

We're sharing 7 ways you can up your game and utilize Google Analytics eCommerce reports to increase your site's conversion rate:

Read on the Web
Using Google Analytics to Respond to Covid-19

Google Analytics can identify pandemic-related problems and opportunities for your business. I’ll explain in this post.

Read on the Web
Search insights to help you understand consumer needs in uncertain times – 20 April 2020 edition

Lucy Sinclair is the EMEA director of the insights team at Google, where she and her team analyse changes in consumer behaviour.

Read on the Web
In major shift, Google Shopping opens up to free product listings

After 8 years, Google will bring free listings into its Shopping results as part of an evolution to better compete against Amazon in product search.

Read on the Web
Microsoft Advertising adds cookie-based experiment option

Cookie-based A/B tests will ensure that customers are only shown ads from either your experiment or original campaign.

Tutorials
Read on the Web
Exporting Google Analytics Data to Google Sheets (Part 1)

Exporting data from Google Analytics to a Google sheet can be done in several ways. Check the simplest of them

Read on the Web
Getting Started: Cloud Functions quickstart

In this video, you'll see a quick tutorial of Cloud Functions, and show you how to deploy a Cloud Function from a Google Cloud project.

Release notes

AI Platform
Read on the Web
April 29, 2020
AI Platform Prediction

Feature

AI Platform Prediction now supports several regional endpoints for online prediction. Regional endpoints provide additional protection against outages in other regions by isolating your model and version resources from other regions. The following regional endpoints are available in beta:

  • us-central1-ml.googleapis.com
  • europe-west4-ml.googleapis.com
  • asia-east1-ml.googleapis.com

You can use these endpoints instead of the global endpoint, ml.googleapis.com, when you use AI Platform Prediction for online prediction. Learn how to use regional endpoints for online prediction, and read about their benefits and limitations.

Feature

You can now deploy scikit-learn and XGBoost models for online prediction using Compute Engine (N1) machine types. Previously, you could only deploy TensorFlow models when you used these machine types. Learn more about ML framework support for Compute Engine (N1) machine types.

You cannot use GPUs with scikit-learn or XGBoost models, and you can only use scikit-learn and XGBoost models with Compute Engine (N1) machine types when you deploy your models and versions to a regional endpoint.

Compute Engine (N1) machine types for online prediction remain available in the beta launch stage.

Feature

The europe-west4 (Netherlands) and asia-east1 (Taiwan) regions are now available for online prediction. These regions are only available for online prediction on their respective regional endpoints, and you can only use Compute Engine (N1) machine types for online prediction in these regions.

When you deploy model versions in the europe-west4 region, you can optionally use NVIDIA Tesla P4, NVIDIA Tesla T4, or NVIDIA Tesla V100 GPUs to accelerate prediction.

When you deploy model versions in the asia-east1 region, you can optionally use NVIDIA Tesla K80 or NVIDIA Tesla P100 GPUs to accelerate prediction.

Learn more about using GPUs for online prediction, and see which GPUs are available in which regions.

Learn about the pricing for the newly available regions and GPU resources.

Changed

We recommend against using Compute Engine (N1) machine types on the AI Platform Prediction global endpoint. Instead, only use Compute Engine (N1) machine types when you deploy models and versions to a regional endpoint.

Model versions that use Compute Engine (N1) machine types and were previously deployed to the us-central1 region on the global endpoint will continue to function.

Read on the Web
April 24, 2020
AI Platform Prediction

Feature

Visualization settings for AI Explanations are now available. You can customize how feature attributions are displayed for image data.

Learn more about visualizing explanations.


Cloud Composer
Read on the Web
April 27, 2020

Changed

  • New versions of Cloud Composer images: composer-1.10.2-airflow-1.10.2, composer-1.10.2-airflow-1.10.3 and composer-1.10.2-airflow-1.10.6. The default is composer-1.10.2-airflow-1.10.3. Upgrade your Cloud SDK to use features in this release.

Fixed

  • Fixed an issue with the CloudSQL Proxy HealthCheck that caused the Proxy Pod to restart repeatedly.
  • The fluentd spec for in-cluster build log exporting now correctly points to the production fluentd image from cloud-airflow-releaser. This fix is required for Composer to correctly perform in-cluster builds for VPC SC configuration.
  • Adjusted ImageBuilder to fix PyPI package installation issues when using VPC SC.
  • Fixed intermittent issues with airflow-monitoring during the initialization phase.
  • Fixed an issue that caused the Airflow scheduler and worker pods to take ~10 minutes to terminate.
  • Fixed an issue with upgrading the image version and improved error handling during Composer environment upgrades.

Deprecated

  • The oldest supported version of Composer is now composer-1.6.0-airflow-x.x.x

BigQuery Data Transfer Service
Read on the Web
April 27, 2020

Changed

BigQuery Data Transfer Service is now available in the Las Vegas (us-west4) region.


Cloud Dataproc
Read on the Web
April 27, 2020

Feature

Dataproc on GKE version 1.4.27-beta is available with minor fixes.


BigQuery ML
Read on the Web
April 27, 2020

Changed

BigQuery ML is now available in the Las Vegas (us-west4) region.


Dialogflow
Read on the Web
April 25, 2020

Changed

In May 2020, the Facebook Messenger integration will be updated, and you may notice slight changes related to fulfillment.

To make sure that your Facebook Messenger bot keeps functioning normally, observe the following recommendations:

  1. To get the Facebook sender.id value, use the originalDetectIntentRequest.payload.data.sender field from the Dialogflow WebhookRequest message.
  2. To get the source field value, use the originalDetectIntentRequest.source field from the Dialogflow WebhookRequest message.
  3. To send rich response messages from your webhook to the Facebook Messenger integration, use the WebhookResponse.fulfillment_mesages[].payload field.
  4. In your webhook logic, don’t rely on the fields that are not documented in the official Facebook Messenger API.

If you have any questions, reach out to your primary support channel.


BigQuery
Read on the Web
April 27, 2020

Changed

BigQuery is now available in the Las Vegas (us-west4) region.


BigQuery BI Engine
Read on the Web
April 27, 2020

Changed

BigQuery BI Engine is now available in the Las Vegas (us-west4) region.

WitBee