The machine learning-based tool also suggests the best audience segments for engagement with that content.
The good news about user-generated content (UGC) is that there’s so much to choose from.
And therein lies the rub.
This week, UGC platform Stackla has expanded its machine learning into a product to help deal with this flood of riches. CMO and co-founder Peter Cassidy told me the new Co-Pilot is the first UGC tool that predictively recommends whether specific content should be published or not, and which audience segments would best respond.
Previously, he said, the Stackla platform offered Content Intelligence, with visual recognition of images, plus sentiment and language analysis for accompanying metadata, captions and comments.
There was also Performance Insights, where user engagement — such as clicks, shares or steps toward a sale — was tracked. Plus, the platform tracked which audience segments responded best to that kind of content.
But now, he said, Co-Pilot adds the ability to watch the curating by the brand’s human’s reviewers, so it can learn what content types the brand likes. And these three components — Content Intelligence, Performance Insights and watching the monitors — are now integrated into a single algorithm.
While the Stackla platform before Co-Pilot was reliant on human reviewers to curate content, he said, now the platform intelligently pre-curates the content so there can be fewer human reviewers. Stackla’s 450 corporate customers include Subaru, McDonald’s, NVIDIA and Expedia.
Co-Pilot recommends if a piece of content should be published or disabled, as well as the most appropriate audience segment for that content. It also recommends which kinds of content should be run through rights management so they can be physically stored for use in a paid ad, compared to a completely manual rights management process previously. Here’s a screen shot of Co-Pilot’s recommendations:
Cassidy said there are no stats yet about how much of a difference Co-Pilot makes, but anecdotal feedback indicates fewer human reviewers are needed, and the process is quicker.