Replai uses computer vision and data analysis to figure out the best video ads

Replai automates analysis of video ad effectiveness and it makes recommendations for how to fix details to make ads more engaging.

The company spent the past couple of years building out its computer vision technology to capture the right data from video advertisements for mobile games and apps. Then it had to create AI to analyze the ads for what was working as the ad creators made nuanced changes to the videos. And now it is offering automated recommendations on the ideal ads for various mobile and social platforms.

To date, the company has managed $2.5 billion in ad spending on video ads that are being processed via San Franciso-based Replai. The recommendations part is why Replai was able to raise $4.7 million in seed funding in October, said João Costa, CEO of Replai, in an interview with GamesBeat.

Gaming is the biggest category of customers for the Replai platform, but a variety of app makers across industries are also using it. The company has been around since 2019, when Costa and Francisco Pacheco, chief product officer, started it.

This kind of data is essential now because companies now have to build a model that works with the “top of the funnel” metrics that are available to advertisers now. The reason is that Apple’s focus on user privacy over targeted ads has limited the amount of information that advertisers can get about the effectiveness of ads. With users opting out of giving their data via the Identifier for Advertisers (IDFA), the whole precision over the effectiveness of ads on iOS has gotten foggy. That is, the bottom of the funnel, where users spend money, isn’t as clear anymore, so the top of the funnel where you first encounter users is where the data is more precise now on iOS.

While IDFA is the big topic related to privacy in the past year, data privacy concerns were galvanized with Europe’s General Data Protection Regulation back in 2016. So it wasn’t hard to see that some of the performance data was going to disappear, Costa said.

Above: Replai uses computer vision to identify components of video ads.

Image Credit: Replai

“The loss of data and the rise of video really led us to believe that video intelligence would be the next step,” Costa said. “The ability for you to segment your users through video itself, using video intelligence, wasn’t very clear. But that’s what we do now, and it’s why we started the company with this vision.”

Tracking how the combination of user acquisition spending and small changes of the various pieces of an ad is increasingly important. And Replai has the ability to track an unlimited number of details related to an ad. For instance, if you’re Ubisoft and you’re running an Assassin’s Creed ad, you’ll probably get a quantitative analysis on how much better the results are when you use the main character Ezio in an ad, as opposed to another character.

“There’s no limit to what you can track,” Costa said. “That’s the key to Replai.”

On average, Replai tracks about 10 tags per ad, with perhaps 30 custom tags overall. Replai figures out what are the best tags, what are the best tags for a social platform and what are the best tags for certain operating systems. Then they correlate those tags to real-time performance data.

“That means that we can recognize automatically pretty much anything in a video, which was the reason for our first funding round at the end of the last year,” Costa said. “We want to be the world’s video intelligence layer.”

On the dashboard, you can see the videos and their various versions. These creative assets are pushed out as ads and then analyzed for small nuanced differences in results. You can see at a glance which creative asset is trending the highest. The system will tell you to swap one character in or another, based on the results and its own recommendations.

Above: Replai can detect nuances in what makes one video ad better than another.

Image Credit: Replai

Replai allows companies to scale the analysis of the video, opening a stream of data, including the measurement of any custom event in a video. And after the analysis, the advertisers should understand why a video performs well or doesn’t. Then Replai makes recommendations for what should work.

As a software-as-a-service, Replai charges for its platform on numbers of channels, number of apps, and numbers of custom tags. And Costa said that the return-on-investment has been clear.

As for competitors, some are focused on production of videos, but fewer are combining that with the intelligence that goes with it, he said. About 50% of customers have adopted Replai as a key tool in the post-IDFA environment because it enables them to continue with data-driven decisions, Costa said.

Costa said the company has used part of the recent funding to triple the size of the product team to 30 people in London, San Francisco, and Porto, Portugal. Costa said only a few companies use computer vision as much — Google Cloud, Microsoft, and others — while Replai makes this its main purpose and so it can parse more actions and data. Replai can be integrated with mobile marketing data partners and channels, such as Facebook, Appsflyer, Adjust, Google, and others. He said Replai has trained its computer vision tech on more than a million hours of video footage.

Above: João Costa is CEO of Replai.

Image Credit: Replai

The company has raised $6 million to date, and most of the customers so far in Europe and the U.S. The recent round’s investors include Hoxton, Sonae IM, Lux Capital, Sequoia, and Accel, as well as existing investors.

While the early focus is on games, Costa said it works with any business in any industry — ecommerce, entertainment, education, dating, sports, etc. — as long as video is used.

Costa said five public game companies are using the Replai tech today. Customers who spend as little as $15,000 can discover how effective their advertising can be. As clients tweak their ads, they get more feedback and learn more from the small changes they make in the video ads.

“That data contributes to a new loop of learning. This new functionality works for all the verticals, including gaming,” Costa said. “This is exactly where we saw the world going.”