Partner: Serenading AI algorithms to find the best singer of the “Bella Ciao” song

About Partner

One in three people in Israel subscribes to Partner, one of the country’s leading telecoms companies, providing mobile networks, internet Wi-Fi, telephone services and IPTV.

Industries: Media & Entertainment
Location: Israel

Tell us your challenge. We're here to help.

Contact us

About Glickman Shamir Samsonov

Founded in 1987, Glickman Shamir Samsonov (GSS) is one of the leading creative agencies in Israel. It’s part of French advertising and public relations multinational, Publicis.

To promote a new Netflix offer, Partner and creative agency GSS used Google Cloud to rate the singing skills of more than 70,000 Israelis grappling with the song “Bella Ciao.”

Google Cloud results

  • Scales from zero traffic to handling one upload every two seconds using Cloud Functions
  • Delivers accurate singing scores on multiple metrics in less than five seconds
  • Helps deliver 20% increase over six months in the market perception of Partner as “innovative”

Supports viral campaign for 70,000 singers in one week

“Bella Ciao” is an infectiously upbeat, much-loved Italian folk tune, sung across Europe for the best part of a century. It features prominently in the hit Spanish TV show “Casa de Papel,” (“Money Heist” in English) and thanks to the show’s international runaway success on the Netflix streaming platform, “Bella Ciao” is now known to millions of fans in countries worldwide.

Israel is no exception, so when creatives at Glickman Shamir Samsonov (GSS) were commissioned to promote Netflix for Partner, a leading Israeli telecom provider, they immediately thought of the song. “The song is on the show, the show is on Netflix, and Netflix is on Partner,” says Roy Gefen, Partner and Head of Digital and CX at GSS. “We wanted to make those connections, and do so in an innovative way.”

“Innovation is about going the extra mile, and that’s what we tried to do with this project. We decided to use Google Cloud Speech-To-Text and other cloud tools to do something truly innovative that engages customers and captures the imagination.”

Roy Gefen, Partner and Head of Digital & CX, Glickman Shamir Samsonov

In a highly competitive market, Partner defines itself by its approach to groundbreaking technology. “Every company is trying to get an edge, and Partner’s edge is innovation,” says Terry Yaskil, CMO at Partner. “We wanted to give people an experience that emphasized our innovative approach.” Soon, the idea of a national singing competition emerged. “Everybody is already singing the ‘Bella Ciao’ song,” says Gefen. “So we said, why not find out who sings it the best, and closest to the original version?”

To test the public’s skills, engage them in a first-of-its-kind way, and promote Partner as the place to see Netflix content (especially the popular “Casa del Papel”) GSS created a cutting-edge web app for a week-long national singing competition. Budding singers uploaded their best rendition of a 15-second clip of “Bella Ciao,” received a score within seconds, and the top performances were rewarded with a trip to Madrid.

“Innovation is about going the extra mile, and that’s what we tried to do with this project,” says Gefen. “We decided to use Speech-to-Text and other cloud tools to do something truly innovative that engages customers and captures the imagination. By using Google Cloud, we managed to pull this off in a smart, fast, and scalable way. But the true beauty is that the user’s engagement remained simple and playful. They didn't have a clue that the process behind the scenes for checking word precision was so complicated. We were paying attention to rhythm, melody, and pitch, and using machine learning to improve score accuracy as more users engaged with the activity.”

Creating new campaign ideas with cloud-based tools

Using cloud-based tools, marketing creatives can generate viral campaigns from simple ideas. Google Cloud helped GSS give the singing competition the speed it needed to catch attention and make it easy for people to share content with others. That’s why one of the key goals for GSS was to deliver results for participants in the competition just moments after they uploaded an entry.

For Amit Bendor, who put the app together for GSS, that meant using AI, algorithms, and APIs to score all aspects of the musical performance at speed. “I used to teach guitar, and as a musician, I really enjoyed bringing programming and music together,” says Bendor. “I used my knowledge of scales, tempo, and phrasing to give people a real score based on an ensemble of algorithms.”

Bendor first used TensorFlow to remove the background music from the original song to have an isolated vocal track to use as a standard. He then devised multiple tests to run concurrently, so that entries could be analyzed and scored in under five seconds. “I used signal processing libraries for Python to analyze the tempo and the tone, to check if people were singing in time and in tune,” says Bendor. “To see how close people could come to the original song, I incorporated a ‘fingerprinting algorithm,’ which accurately compared cover versions to the original recording, and also checked that no one cheated the system by uploading the original song itself. And to see if Israelis could sing Italian lyrics with no mistakes, we relied on Speech-to-Text.”

“This is an algorithmic project, so we wanted to spend our time making our algorithms as accurate as possible. With Cloud Functions, we didn’t have to worry about scaling infrastructure at all. We simply set the precise RAM required for each function, then left it to autoscale.”

Amit Bendor, Practical AI/ML Expert, Glickman Shamir Samsonov

For Bendor, choosing the Google Cloud API was a simple decision. “This is the best speech-to-text API available today,” he says. “You can even identify specific words that you want the API to be more conscious about. When I took the isolated vocal track from the original song, I recognized that one particular word could sound unclear. I tweaked the API on just that one word and the results were much more accurate.”

When a viral campaign is launched, nobody knows if it will be even a minor success, let alone a hit, so the team used Cloud Functions to deploy the project with ease and autoscale its components in line with demand.

“This is an algorithmic project, so we wanted to spend our time making our algorithms as accurate as possible,” says Bendor. “With Cloud Functions, we didn’t have to worry about scaling infrastructure at all. We simply set the precise RAM required for each function, then left it to autoscale.”

“The perception of Partner as an innovator has grown around this campaign. Market analysis showed that over a six-month period, including the campaign period, there was a 20% increase in people who see our brand as innovative.”

Terry Yaskil, CMO, Partner

Building an innovative image

The game was a breakout success, becoming a national news story. In the first 24 hours, more than 2,000 people took the challenge, and within a week, more than 70,000 had played the game. Singers waited just five seconds to be scored, and typically left multiple renditions, with each player spending an average of four minutes in the app. At its peak, one new entry was uploaded every two seconds.

“The perception of Partner as an innovator has grown around this campaign,” says Yaskil. “Market analysis showed that over a six-month period, including the campaign period, there was a 20% increase in people who see our brand as innovative. The project created more than 6,000 leads for Partner marketers, as well as an increased appetite for future projects leveraging cloud technology.”

“By using cloud technology, we could make something that was fun, fair, and fast for people to use,” says Bendor. “Google Cloud infrastructure gives us the tools to do that, and as far as we’re concerned, it’s the best infrastructure you can get for AI or algorithmic projects.”

Tell us your challenge. We're here to help.

Contact us

About Partner

One in three people in Israel subscribes to Partner, one of the country’s leading telecoms companies, providing mobile networks, internet Wi-Fi, telephone services and IPTV.

Industries: Media & Entertainment
Location: Israel

About Glickman Shamir Samsonov

Founded in 1987, Glickman Shamir Samsonov (GSS) is one of the leading creative agencies in Israel. It’s part of French advertising and public relations multinational, Publicis.