Swedish SaaS company Flowbox now invests in AI
The Swedish startup Flowbox, which has developed a platform for collecting user-generated content from social media, is growing at a furious pace and recently raised SEK 80 million in venture capital. Now, the company is soon launching its first AI-based solution.
What’s the first thing that comes to your mind when you hear the term artificial intelligence? Maybe self-driving cars or talking robots? Artificial intelligence, or AI, is no longer something that belongs to science fiction movies in scenes that take place in a distant future – it’s present today in several different applications. An increasing number of companies are now choosing to use AI in new contexts and in new areas of application, although not always in the form of human-like robots but also, or perhaps more often than not, as more subtle solutions.
The SaaS company Flowbox has developed a platform that allows companies to collect user-generated content (UGC) from social media, in the form of images and videos, for example. The content can then be distributed and used in marketing in owned channels like a website or online shop. Now, Flowbox is taking the next step into the AI world by focusing on developing smarter solutions that will help companies use the platform more efficiently.
Erik Lundberg is a data scientist at Flowbox and was recruited to the company at the beginning of the year. He has a broad experience in mathematical statistics, programming and machine learning and is currently working on solutions that will make Flowbox’s platform more intelligent and create value for its customers.
“We’re currently building a model, using advanced statistical algorithms, that will sort the image flows in our platform in a way that ensures that the most engaging and relevant images are the ones that users see first. We’re hoping that this will improve the user experience for our customers and ultimately lead to a higher click-through rate,” says Erik Lundberg.
The algorithm that will help users to sort the images goes under the name of Flowscore. Simply put, it consists of two main parts: click-through rate, and how recent the image is. The amount of data available is also included in the calculations: the more data in the form of, for example, clicks and impressions, the higher the score. An image with many clicks and impressions is thus ranked higher than an image with fewer clicks and impressions. New images receive extra bonus points that affect the score, to ensure that the content displayed at the top of the flow is fresh and up-to-date.
Erik Lundberg emphasizes that the functions now being developed aren’t really based on machine learning but rather just advanced statistical methods, but he also points out that this is just the beginning and that there are plans on more advanced projects in the future.
“One future project is to create filters that allow customers to filter the images, to be able to more easily find the content that is relevant to them. One company might for example want pictures with only people on, while another company might just want outdoor pictures, and so on. With machine learning we can teach the computer to find patterns and recognize such images.
Machine learning – a competitive advantage
The terms machine learning, deep learning and artificial intelligence are often used synonymously but in reality they have different meanings. What really sets these terms apart? For those who aren’t working professionally in the field, it might not be obvious.
AI as a concept was coined already when the first computers emerged in the 1950s and had a rather broad meaning, which then referred to methods for creating intelligent machines capable of solving problems and performing tasks that usually require human intelligence. A lot has happened since the fifties. AI as a concept has gotten several branches where new research areas have gradually emerged, and today, professionals distinguish between AI and machine learning, although the concepts are related. Erik Lundberg explains:
“Machine learning is a subgroup of AI and has to do with using advanced mathematical models to find patterns in data. Oftentimes, very large amounts of data, so called big data, is required for these models. Simply put, you could say that if we have a certain amount of historical data, we try to find a pattern and then apply this pattern to newly received data. Artificial Neural Networks is one of many algorithms that can be used in machine learning. Deep learning is a special case of Artificial Neural Networks.”
Christian Kaunissaar, CTO at Flowbox, believes that the possibility of soon being able to use machine learning to create smarter functions in Flowbox’s platform will make a big difference for the customers and be a major competitive advantage.
“We will be able to simplify customer decisions to a great extent as well as automate a lot of things. Today there are really no technical limitations, it’s more an ethical question where you have to decide how much you want to limit the customer's own influence and decision making,” says Christian Kaunissaar.
A win-win for both customer and supplier
The new Flowscore feature, which will make it possible to sort images by click-through rate, is already finished in a trial version and will be tested by the first customer in the summer of 2020. The new feature will help Flowbox’s customers identify the best images in the flow, much faster.
“Right now, we’re working with a few selected customers, looking at their specific needs. But as we find solutions to their problems, we will also be able to solve similar problems for other customers who are in the same industry or segment. For a company that sells for example glasses, it can be to identify whether there are people and glasses in the pictures. Such a solution would probably benefit other companies in the same industry as well. When it comes to identifying whether there are people in the pictures or not, it’s something that could most likely benefit most of our customers. In this way, we will be able to find general solutions based on specific challenges,” says Christian Kaunissaar.
But even though smart algorithms and machine learning can simplify a lot of work for both customers and suppliers, not everything will be automated. According to Christian, transparency is important and every customer should always be able to choose whether they want to use the computer's calculations or if they want to show all available images.
“This will be a way to make our product even smarter and more efficient. As a user, you get a way to select images that will happen automatically and is much better than using your gut feeling. With the help of AI, the newest and most engaging images will appear at the top of your feeds. Our aim with this project is to give our customers a better user experience of the flows and by doing so, also help them increase their sales. If our customers can do that and also spend less time using the platform, then it is a win-win for both them and us,” says Erik Lundberg.
Flowbox was founded in 2016 and since then, the team has constantly developed new features in the platform. The reason why the company hasn’t invested in machine learning until recently is simply that the timing hasn’t been right before, Christian explains.
“When we started the company, our first priority was to create a good workflow and build functions that allow companies to collect the content. The next step has been to link the images with the customers' products. After that, we’ve added features such as our insight part for statistics, our publishing tool and much more. Now, we have a solid foundation and the resources needed to invest in data science and expand the platform with additional features that add that little extra to the experience,” says Christian Kaunissaar.
Christian Kaunissaar, CTO, Flowbox
Erik Lundberg, data scientist, Flowbox
For more information, contact:
Helena Nordh Myhrman
Head of Marketing & Communications, Flowbox