Audiomob: Keeping gamers playing without distraction, with in-game audio ads
About Audiomob
Audiomob has created a new way to monetize mobile games. The company offers non-intrusive audio ads that can be listened to in game or app, greatly increasing player retention and generating better reviews, all while providing developers with monetization. Built from the ground up by developers, for developers, the company strives to offer a simple alternative to standard ads.
Tell us your challenge. We're here to help.
Contact usAs Audiomob grew, it could no longer manually review all the ads that it received, so it turned to Google Cloud to build an effective AI review pipeline.
Google Cloud results
- Replaces manual review process with scalable AI pipeline built on Google Cloud
- Built entire AI pipeline in just two weeks
- AI pipeline is 100x cheaper than the cost of manual review
- Enables unlimited scalability for the future
Manual reviews down from 20,000 to 70 a day
Whether it's a mum playing on her mobile while the kids are napping, or a commuter killing time on their way to work, gaming is now a part of the everyday lives of billions of people. It's estimated that there are more than 3.3 billion gamers worldwide, with mobile game revenues accounting for almost half of the global games market at more than $90 billion. The gaming industry now brings in more money than Hollywood. And that makes it highly valuable for advertisers.
The gaming world, with its broad range of demographics, lets brands target their advertising with a greater degree of accuracy. But brands have to get their ads right. Pop-up banners or video ads are intrusive and can impede the gamer's experience. This can cause gamers to stop playing, limiting the impact of any advertising.
Getting gamers to listen up
Audiomob understands gamers don't want their gaming experience broken up with ads. The company is driven by a goal to help advertisers better reach their consumers and help game developers better monetize their games. With its platform running on Google Cloud, Audiomob allows developers and advertisers to programmatically insert audio ads into mobile games.
"Advertising is necessary to maintain a mobile game ecosystem, like it is with many other ecosystems. So advertising is driving that revenue and what makes audio different is that it doesn't hinder the game experience," says Ville Laitinen, Head of Engineering, Audiomob. "You can continue playing the game while the audio ad plays and with that the developer creates an opportunity for incremental revenue."
Right from the start, Audiomob's two founders, Christian Facey and Wilfrid Obeng were supported by the Google for Startups Black Founders Fund, gaining both financial and infrastructure support to develop their business. The company grew from two people to a team of 33 employees in London, Abu Dhabi, and New York. The company's ongoing success is built on a simple core principle: there is a better way to advertise in games and apps. The success of Audiomob has created its own unique challenges, which the company is solving using Google Cloud. Audiomob chose Google Cloud primarily for its potential to significantly speed up its time to market, a crucial factor given its ongoing rapid growth and expansion. The Google Cloud unified ecosystem simplifies many of Audiomob's more complex tasks, helping the company get its work done more quickly.
Creating an effective AI pipeline
On average, Audiomob's advertising pipeline processes over 20,000 unique audio and image creatives every day, which come from a wide range of sources. The company wanted to ensure brand safety and stop users from being exposed to any inappropriate content. But it was impossible for Audiomob to manually review this many creatives, and to build its own AI content-moderation system would have been hugely time consuming and costly. So it turned to Google Cloud AI solutions to create a new AI review pipeline in just two weeks.
"We needed something that understands speech in multiple languages and turns it into text and we didn't want to have to train our own models. So we use three Google Cloud solutions, Speech-to-Text AI, Vision AI, and Natural Language AI, to create a pipeline for the content."
—Troy Atkinson, Staff Engineer, Audiomob"We needed something that understands speech in multiple languages and turns it into text and we didn't want to have to train our own models. So we use three Google Cloud solutions, Speech-to-Text AI, Vision AI, and Natural Language AI, to create a pipeline for the content," says Troy Atkinson, Staff Engineer, Audiomob.
The company uses Speech-to-Text AI to convert the audio ads into a text transcript, which it can do for more than 100 different languages. It has been trained on millions of hours of audio data and billions of text sentences, enabling improved recognition and transcription for more spoken languages and accents.
Once the audio ad has been run through Speech-to-Text AI, it's then analyzed by Natural Language AI, which categorizes the content and flags any potentially inappropriate or offensive material. Audiomob also uses computer vision to study banner ad images and read the text using optical character recognition. The team then runs them through the same natural language AI as the audio transcript again, creating categories. The company also uses Google SafeSearch to check the content of any images accompanying the audio ads, which helps the company spot any offensive material. All the audio and image creatives that Audiomob receives are also stored in Cloud Storage.
Building a scalable solution for today and tomorrow
"The Google Cloud tools give us different confidence levels for different categories of what could be prohibited and this helps us monitor the stream of creatives, and pull some out for manual review if needed," says Laitinen. Since launching the AI review pipeline roughly 70 creatives are flagged as potentially inappropriate and pulled for manual review each day. This has made a huge difference to the company's performance, allowing it to get ads to market more quickly and help keep gamers safe from inappropriate material.
"BigQuery really is at the core of everything we do. It forms the basis of the reports that the game publishers and developers work with to see how their games perform. We use it to feed our own internal AI models for training."
—Troy Atkinson, Staff Engineer, AudiomobThe scalability of the AI pipeline is a big plus for the company, setting it up for the future. "We had 20,000 creatives a day this year, which is far more than last year," says Atkinson. "And we just know that the AI pipeline we've built on Google Cloud is going to scale up as we grow. If we have a hundred times our inventory, it's still going to work perfectly fine."
Part of this scalability relates to the pipeline's use of Pub/Sub, which allows for services within the system to communicate asynchronously. This allows the pipeline to flow smoothly, preventing bottlenecks from forming despite the high volume of ads passing through it. Audiomob also uses BigQuery extensively throughout its systems to store its data.
"BigQuery really is at the core of everything we do," says Laitinen. "It forms the basis of the reports that the game publishers and developers work with to see how their games perform. We use it to feed our own internal AI models for training."
One such AI model helps Audiomob to define audiences to target. With an emphasis on user privacy, Audiomob doesn't use personal information to target users. Instead it has built an AI model based on thousands of contextual data points generated by user activity, such as time, broad location, device family, and other non-personally identifiable information. Audiomob's AI model pulls all those data points from BigQuery and uses them to target audiences in real time to match users with the right ad. As a result, Audiomob has improved listen-through rates (how many users listen to the ad all the way through) by up to 20%. Meanwhile, the resulting click-through rate on companion banner ads, which are shown while the audio is playing, is 2%, compared to an industry average of under 0.5%.
"Our relationship with Google Cloud feels more like a partnership where they are supporting us in our growth from a technical point of view. They have enabled us to focus on growing our business."
—Ville Laitinen, Head of Engineering, AudiomobSounds like success
Audiomob is currently valued at over $100 million, and will continue to grow into other formats and mediums with the aim of an IPO. The startup has become a company with huge potential, and a number of big name clients, including Sony Music, Babbel, and Dr. Pepper. And it has plans to continue its relationship with Google Cloud, using its AI and machine-learning solutions to improve its AI pipeline, making sure it's even more adaptable and scalable as the company grows. "Our relationship with Google Cloud feels more like a partnership where they are supporting us in our growth from a technical point of view," says Laitinen. "They have enabled us to focus on growing our business."
Tell us your challenge. We're here to help.
Contact usAbout Audiomob
Audiomob has created a new way to monetize mobile games. The company offers non-intrusive audio ads that can be listened to in game or app, greatly increasing player retention and generating better reviews, all while providing developers with monetization. Built from the ground up by developers, for developers, the company strives to offer a simple alternative to standard ads.