Last year at Firebase Summit, we introduced you to Predictions, a machine learning product that helps you smartly segment your users based on their predicted future behavior. Without requiring anyone on your app team to have ML expertise, Predictions gives you insight into which segments of users are likely to churn or spend (or complete another conversion event) so you can make informed product decisions and grow your app.
As of today, Predictions makes more than 6 billion predictions per day for our developers and allows them to take meaningful actions by making predictive segments available for targeting in Remote Config, Cloud Messaging, In-App Messaging, and A/B Testing.
This year at Firebase Summit, we announced that Predictions has graduated out of beta and into general availability with a host of new features that we added based on your feedback.
Since Predictions continuously update based on actual user behavior inside your app, we heard from many of you that you wanted to know how stable a prediction was before you integrate it into your app.
To help answer this question, we created a health indicator at the bottom of each predictive segment card that gives you a snapshot of how a certain prediction is performing:
Image 1: Green means it has been performing consistently well over the last two weeks
Image 2: Yellow means it is performing well today but did not meet the quality threshold some time in the past two weeks
Image 3: Red means it is not performing well today and had other performance issues over the last two weeks
It is worth mentioning that actions targeted with Predictions have a fail-safe mechanism, so if a predictive segment is performing poorly, it simply turns inactive. That means, if you are using Remote Config to deliver a set of values to users in that predicted group, Remote Config will gracefully fall back to your default values if the predictive segment decreases in reliability. Any notifications or in-app messages directed at that predictive segment will also not trigger until the predictive segment increases in accuracy.
To help you understand how we assess the quality of a prediction, we are now exposing our evaluation criteria. For every predictive segment, we use a portion of your historical data from the last 28 days that we hold out during the model training phase.
We then compare the results of the prediction to what actually happened. This gives us two ways to score the prediction: how many of the users in the predictive segment actually behaved in the predicted way (we call that true positive rate) and how many users in the predictive segment were incorrectly classified (or in more technical terms, the false positive rate).
You can access this data from the bottom of the prediction card
Tapping on the health indicator exposes these values.
By exposing these two scores to you, you can now make a better determination about which risk profile to choose for your action.
Another common question we received during our beta phase is what went into creating a predictive segment. We now offer a details page that gives you the ingredient list! You can click through and see what data our model makes use of. This includes event frequency, volume, and parameters as well as other data like device language, freshness of app install and more.
The last thing we are excited to announce is that now, you can export your raw predictions data into BigQuery. This will give you access to the raw prediction score, the thresholds we used for each risk profile, as well as the final result. You can use this data to create your own risk profiles or if you supply your own user_id property in analytics, to do sophisticated analysis with your analytics data. For example, you can find out which countries exhibit the highest potential to churn or spend!
We are humbled to have gained your trust over the past year and hope these improvements make it easier for you to make the most out of Predictions in your mobile apps and games. As always, if you have any questions, you can find us on Twitter (@firebase) and on Stack Overflow.
For more information on these updates, check out our docs below!
Predictions risk tolerance and performance
Predictions model inputs and details page
Predictions data export to BigQuery
When Test Lab was originally launched with Firebase in 2016, it supported only Android devices. At Google I/O 2018 in May, Test Lab launched closed beta support for iOS. This included a limited set of iOS devices and a basic UI.
Building comprehensive tests for Android involves writing code, using Espresso and UIAutomator, that acts as a sort of "remote control" for your app. Similarly, on iOS, testing is performed using XCTests. In both cases, Test Lab can run your tests against actual devices in a cloud-hosted device farm.
At the Firebase Summit in Prague at the end of October, the Test Lab team announced general availability of support for iOS, including ten models of iPhones and iPads running seven different versions of iOS, including iOS 12. We have also improved the iOS documentation and console experience for you.
Test Lab also launched a number of improvements to Robo, a tool which runs fully automated tests against your app running on Android devices. Here's what's new with Robo.
Games are difficult to crawl because they often have a highly customized UI, rather than using system widgets. This makes it difficult for Robo to crawl the game's experience. Now, if Robo detects that the app under test is actually a game, it will perform random taps and swipes in an effort to interact with the game's UI. This can yield useful crash and performance data and is an early but significant step towards more meaningful automated game crawling.
Test Lab now detects and warns if your APK makes use of internal Android APIs. On Android P and newer, using such APIs can crash your app. Whenever such an API is accessed during a Robo crawl a stack trace is recorded in the device logs. This pinpoints the location in your app's code where the violation occurs.
Test Lab now warns developers when it notices that Robo got stuck in a crawl. For example, if the user is presented with a complicated sign-up form or a login screen, it may be difficult for Robo to satisfy the requirements of the form. In situations like this, Robo will suggest an action to the developer to help it continue a full crawl, such as providing test credentials or writing a Robo Script.
If you aren't in the habit of regularly testing your app, consider giving Test Lab a try at no cost using the free daily quota of tests. No coding is necessary to run a Robo test on Android - just upload your APK to get started. And be sure to let us know what you think in the #test-lab channel of the Firebase Slack.
As the Product Manager of Firebase Remote Config, a product that helps you modify your app without deploying a new version, I spend a lot of time talking to our customers, and one of the most common requests I hear is, "Can I get rid of the Remote Config cache and fetch new values right away?"
While we can't get rid of the cache entirely -- caching ensures that the Remote Config service stays up and running, free of charge, no matter how many millions of users you have -- thanks to some new features we've added to Cloud Functions for Firebase, you can now ensure that your users always get fresh Remote Config values whenever they open your app!
Before we get into our solution, let me explain why the problem isn't as bad as you think, by looking at a typical app that has a Remote Config cache time set to 4 hours. You might think that, if this app developer publishes new Remote Config values, most of their users won't see these new values until several hours later.
But the fact is, the vast majority of their users will see the new Remote Config results as soon as they run the app! Remember, the way the cache works is that it looks at the last time Remote Config successfully fetched data. So anybody who last opened their app more than 4 hours ago will retrieve fresh data.
To put it another way, if your user opens up your app at 7:00 pm, you push out new values at midnight, and then they open up your app again at 12:15 am, Remote Config will fetch your new values from the server. Sure, it's only been 15 minutes since you published your values, but their cache is over 5 hours old, so Remote Config will fetch fresh values.
Nevertheless, I understand that it's really important for many of you to push out new changes to all of your users right away. Maybe you want feature flags to be disabled quickly. Or, maybe you have time-sensitive values, like marketing campaigns or a flash sale, in which case it would be awkward for your users to still see messaging for a sale that's no longer running.
Fortunately, we've built a new integration with Cloud Functions for Firebase that will make it easier for you to build custom behavior based on events that happen in Remote Config. Specifically, you can now use new triggers to write code that runs whenever your team publishes Remote Config values -- whether that's through the Firebase Console or the REST API.
There are a number of ways you can use these Remote Config triggers. One significant way is to use them to notify all of your apps as soon as a new set of values get published on Remote Config.
If you're interested in learning how to implement this feature, here's a high level overview of the process:
For more information, check out the solutions guide in our documentation. This guide contains all of the code samples you'll need, both in your Cloud Function, and on the client.
You might be wondering why you can't just a) set your Remote Config cache to 0 all the time, or b) have your clients go and always retrieve the new Remote Config values as soon as they receive the notification.
The answer is that Remote Config still needs to make sure the service is usable (and free) by all apps, and we've set up both client and server-side throttles to make sure that no single app accidentally abuses the service. By sticking with the default caching time when there's no local flag indicating that there's new config available, you can avoid any client-side throttles. Your app will have a faster start-up time, too, by avoiding unnecessary network calls.
By waiting until your client goes into the foreground to fetch these new values, you avoid any server-side throttles. You'll also avoid having your clients make unnecessary network calls in the background, which means your app uses less data overall, which makes your users happier, particularly in areas of the world where data usage is expensive.
By making sure your apps are still well-behaved, you can still get the freshest possible Remote Config values while avoiding any throttles that might be applied to your app. Please consult our caching & throttling documentation for further information on this topic.
One other advantage of using this system is that if your app in the foreground when it receives this notification, you can immediately act on this information - perhaps by fetching the new Remote Config data and prompting the user to refresh their screen. So those of you who were actively polling the Remote Config service while your app was in the foreground no longer need to resort to any of those network calls anymore.
Of course, this isn't the only use for Remote Config triggers and Cloud Functions - just one of the most requested, for sure! A number of developers also want to know when new Remote Config values have been pushed to the Firebase console, and you can use Cloud Functions to send off an email to your team or push a message in a Slack channel.
You could also use Cloud Functions to keep different projects in sync. For instance, you could use a cloud function to copy a set of Remote Config values from your production project to your development or testing project.
If you're interested in giving this new approach a try, we encourage you to read over the full documentation in the solutions guide. As always, if you have other questions about Remote Config or other feature requests, we're happy to hear your feedback. Please reach out to us in the Firebase Talk group, or on Stack Overflow.
If you're building or looking to build a visual app, you'll love ML Kit's new face contour detection. With ML Kit, you can take advantage of many common Machine Learning (ML) use-cases, such as detecting faces using computer vision. Need to know where to put a hat on a head in a photo? Want to place a pair of glasses over the eyes? Or maybe just a monocle over the left eye. It's all possible with ML Kit's face detection. In this post we'll cover the new face contour feature that allows you to build better visual apps on both Android or iOS.
With just a few configuration options you can now detect detailed contours of a face. Contours are a set of over 100 points that outline the face and common features such as the eyes, nose and mouth. You can see them in the image below. Note that as the subject raises his eyebrows, the contour dots move to match it. These points are how advanced camera apps set creative filters and artistic lenses over a user's face.
Setting up the face detector to detect these points only takes a few lines of code.
lazy var vision = Vision.vision() let options = VisionFaceDetectorOptions() options.contourMode = .all let faceDetector = vision.faceDetector(options: options)
The contour points can update in realtime as well. To achieve an ideal frame rate the face detector is configured with the fast mode by default.
fast
When you're ready to detect points in a face, send an image or a buffer to ML Kit for processing.
faceDetector.process(visionImage) { faces, error in guard error == nil, let faces = faces, !faces.isEmpty else { return } for face in faces { if let faceContour = face.contour(ofType: .face) { for point in faceContour.points { print(point.x) // the x coordinate print(point.y) // the y coordinate } } }
ML Kit will then give you an array of points that are the x and y coordinates of the contours in the same scale as the image.
The face detector can also detect landmarks within faces. A landmark is just an umbrella term for facial features like your nose, eyes, ears, and mouth. We've dramatically improved its performance since launching ML Kit at I/O!
To detect landmarks configure the face detector with the landmarkMode option:
landmarkMode
lazy var vision = Vision.vision() let options = VisionFaceDetectorOptions() options.landmarkMode = .all let faceDetector = vision.faceDetector(options: options)
Then pass an image into the detector to receive and process the coordinates of the detected landmarks.
faceDetector.process(visionImage) { faces, error in guard error == nil, let faces = faces, !faces.isEmpty else { return } for face in faces { // check for the presence of a left eye if let leftEye = face.landmark(ofType: .leftEye) { // TODO: put a monocle over the eye [monocle emoji] print(leftEye.position.x) // the x coordinate print(leftEye.position.y) // the y coordinate } } }
Hopefully these new features can empower you to easily build smarter features into your visual apps. Check out our docs for iOS or Android to learn all about face detection with ML Kit. Happy building!
Here at Firebase, we believe that apps improve the way people live, work, learn, and socialize. That's why our mission is to make app development as easy as possible by giving you a platform that solves key challenges across the app lifecycle. Whether you're an up-and-coming startup or a well-established enterprise, Firebase can help you build your app, improve its quality, and grow your business.
It's exciting to evolve Firebase alongside our passionate community. Right now, over 1.5 million apps are actively using Firebase every month. We love hearing your stories; they inspire us to keep making Firebase even better so you can continue to succeed. One story we heard recently is from Hotstar, India's largest entertainment app with over 150 million monthly active users around the world.
A few months ago, the Hotstar team safely rolled out new features to their video watch screen during a major live-streaming sports event. These changes, along with updates to their onboarding flow, increased user engagement by 38%! Impressively, by using a combination of Firebase products, Hotstar was able to do this without disrupting users, sacrificing stability or releasing a new build.
Learn more about their story here:
Today, we're hosting the third annual Firebase Summit in Prague to meet many more members of our developer community and learn about the great things they're building. All sessions will be posted to our YouTube channel or you can read on to learn about all the exciting new updates we're announcing today!
We've been working hard to make it easier for sophisticated app development teams to use Firebase. Today, we're excited to share that we'll be adding support for Firebase to our Google Cloud Platform (GCP) support packages, available in beta by the end of this year.
If you already have a paid GCP support package, our beta will let you get your Firebase questions answered through the GCP support channel - at no additional charge. When this new support graduates to general availability, it will include target response times, technical account management (for enterprise tier), and more. You can learn more about GCP support here.
If you're planning to stick with Firebase's free support, don't worry - we don't plan to change anything about our existing support model. Please continue to reach out to our friendly support team for help as needed.
In addition to Cloud support, we've made improvements to Firebase across the board. More below!
Manage projects with ease via Firebase Management API
We've worked hard to open access to our server-side APIs, so that you can easily integrate Firebase services with your existing systems. Today we're releasing the Firebase Management API, a REST API that allows you to create and manage projects and apps programmatically. Now, you can create and destroy Firebase environments as part of your existing developer workflow.
The Management API is also enabling partners to build awesome new experiences. We're thrilled to share that you can now deploy to Firebase Hosting directly from within StackBlitz and Glitch, two web-based IDEs. Their platforms will now automatically detect when you are creating a Firebase app and allow you to deploy to Firebase Hosting with the click of a button, without ever leaving their platforms.
What's especially cool is that this isn't just for partners. This is a new, extensible API and we're very excited to see what you build with it. You can learn more and get started here.
Enhanced face detection with ML Kit
Launched at Google I/O in May, ML Kit makes machine learning easy and accessible for all app developers, regardless of your experience with ML. If you're new to the space you can use ML Kit's out-of-the-box APIs, like text recognition or face detection, or if you're more experienced you can bring your own custom TensorFlow Lite models and serve them through Firebase.
Today, we're expanding on the face detection API with the beta launch of face contours, allowing you to detect 100+ detailed points in and around a user's face. The face contours functionality empowers apps to easily overlay masks or accessories on facial features with high fidelity and accurate positioning, or add beautification elements, like skin smoothing or coloration. See our docs to learn more!
Increased confidence in deployment with Cloud Firestore
In the past, we've heard feedback that testing can be tricky on Firebase. For example, it's sometimes hard to set the right rules to ensure your apps are secure. To help with this problem, we're releasing local emulators for Cloud Firestore and the Realtime Database. These emulators let you develop and test locally, and can be built into your continuous integration workflow so you can deploy with more confidence and peace of mind. Learn more about the emulator here.
Propagate Remote Config updates in near real time + integration with Cloud Functions
Developers love using Remote Config because it gives them the ability to modify their app, customize the UI, or release a new feature without deploying a new version that could disrupt users. But, there was no easy way of knowing when an app's Remote Config was updated! You had to fetch updates from Remote Config every few hours to ensure your users always saw the latest changes in their app.
Today, we're happy to announce that Remote Config now integrates with Cloud Functions & Firebase Cloud Messaging so you can notify your apps in near real time when you publish (or rollback) a new config. This reduces the set-up complexity of Remote Config and uses less bandwidth on devices because apps only need to fetch when a new config is available.
Additionally, Remote Config can now trigger developer-defined functions when you publish or rollback your config. This way, you can keep different Remote Config projects in sync (for development/staging/production environment workflows), as well as send Slack messages to your team when a new config is published. To learn more, visit our docs!
From our early access partner eBay:
"The combination of the Cloud Functions with the Firebase Remote Config REST API has allowed my distributed team at eBay to be instantly notified of any changes to our application's configuration. Using these tools to create a function which pushes changes to Slack ensures everybody who needs to know about a configuration change has that information immediately."
- Jake Hall, eBay Classifieds Group Mobile Architect
Test Lab for iOS graduates into general availability
At Google I/O, we also launched the beta availability of Firebase Test Lab for iOS. Over the last several months, we've expanded the iOS device farm, added support for iOS 12 as well as for older iOS versions, and integrated the UI for iOS into the Firebase console. With these updates, we are graduating Test Lab for iOS out of beta and into general availability. Learn more and get started with Test Lab today!
Performance Monitoring: Sessions insights and issues management
Even if you run thorough tests throughout your development lifecycle, bugs and performance issues will pop up in your production app from time to time. Performance Monitoring gives you insight into these issues and automatically surfaces the most critical issues in a given trace instance (i.e. a particular app start or checkout flow). Now, you can dive into an individual trace session to see exactly what was going on when a performance issue occurred.
For example, in the following dashboard, you could see that CPU usage spiked after the app fetches and renders a product image, which tells you the specific part of code to investigate.
With all the data and issues that Performance Monitoring surfaces, it can be hard to prioritize your efforts. That's why we're also launching the ability to "mute", "close", and "reopen" issues in your console. Muting temporarily silences the issue, so you can concentrate on other work, until you're ready to tackle it. Marking an issue as closed indicates that it's been solved, but Firebase will notify you if it occurs again.
Learn more about session insights and issues management here.
Crashlytics is now integrated with PagerDuty
App performance and stability issues can occur anytime. To help you stay on top of stability, even when you're away from your desk, we're introducing a Firebase Crashlytics stability digest email and a new integration with PagerDuty. The stability digest highlights emerging issues that could become problematic in the future, while the PagerDuty integration allows you to alert your team about a high impact crash, any time of day. To connect Crashlytics with PagerDuty, follow the steps here.
Do more with your data with BigQuery + Data Studio
Earlier this year, we integrated Crashlytics with BigQuery so you can run deeper analysis on your crash data. To help you get started in BigQuery, we've put together a Data Studio template, so you can quickly produce a shareable report. You can preview the template with mock data, and then customize the report to suit your needs. Learn more here.
Predictions graduates out of beta into general availability
Last year at the Firebase Summit, we introduced you to Firebase Predictions. Predictions applies Google's machine learning to your app analytics data to create user segments based on predicted behavior. Without requiring anyone on your app team to have ML expertise, Predictions give you insight into which segments of users are likely to churn or spend (or complete another conversion event) so you can make informed product decisions. This year, we're excited to announce that Predictions is graduating out of beta and into general availability with a host of new features designed to make Predictions more useful.
Wondering what goes go into any given prediction? We added a new details page that shows you what factors the ML model considered (like events, device, user data, etc.) to make that prediction. We also now expose performance metrics for each prediction, letting you see how the prediction has performed historically against actual user behavior, so you can better calibrate your risk tolerance level. And, if you want to do a deeper analysis of prediction data or use it in third party services, you can export your complete prediction dataset to BigQuery.
Check out our docs to learn more!
Reach users more effectively with dynamic audiences in Google Analytics
Google Analytics for Firebase has always given you the ability to segment your users into audiences based on events, device type, and other dimensions. Now, we're enhancing the audience builder with a few major updates: dynamic audience evaluation, audience exclusion, and membership duration.
First, audiences are now dynamic by default, meaning that Firebase will pull in new users that meet your criteria and automatically remove users that no longer meet your criteria. If, for example, you set up an audience of users who are on level 5 in your game, as users beat the level and move on to level 6, they will automatically be removed from the audience. Conversely, once users advance to level 5 they will automatically be added to the level 5 audience you've established.
Secondly, you can refine your audience by adding exclusion criteria using and/or statements which enable you to create audiences like "users that have added to their shopping cart but not made a purchase".
Finally, audiences can now include a membership duration, allowing you to ensure that your audiences stay fresh. This enables you to target users who have completed an action within a specific time period, e.g. "made a purchase within the past two weeks".
Dynamic audiences allow you to reach your users more effectively with relevant messaging and a more personalized app experience. Learn more and get started with dynamic audiences here.
Run automatically recurring campaigns with Cloud Messaging
Once you've defined your user segments using Analytics or Predictions, you can use Firebase Cloud Messaging (FCM) to send notifications to latent users to bring them back into your app. We've redesigned the notifications console to support more sophisticated campaigns. This new UI gives you the power to set up recurring notification campaigns that automatically send messages to new users as they meet the targeting criteria. Previously, you could only schedule one-time sends.
Additionally, the new notifications UI allows you to easily target users based on the date of their first session or the number of days since they last opened an app. And last but not least, we've updated the campaign results view so you can track the effectiveness of recurring notification campaigns day-by-day.
Check out the new UI on your console!
We're excited about all the updates to Firebase that we've announced today. As we continue to grow and enhance the platform, we'd love to have your feedback. Join our Alpha program to get a sneak peek of what we're building next, share your thoughts with us, and shape the future of Firebase.
If you weren't able to join us in person in Prague, all of our sessions are recorded and posted to our YouTube channel. Thanks for being a part of our community and happy building!
