Learning what makes an Instagram account thrive, understanding why some posts get more attention than others and how to create the perfect post are some of the challenges that we face when managing a social media account or trying to understand how the social media ecosystem surrounding our product works.
We’ve shown you what we’re able to do in text-driven environments like Twitter, and now we will dive into mixed ones with Instagram as an example.
Providing you with automated information retrieving, labeling of the graphical content of the images and videos and text analysis of the captions and comments, we provide the perfect picture for any situation so that you base your decisions on accurate, complete facts about the problem you’re facing.
Let’s say, for example, that we want to better understand what the TechCrunch account, which has close to a million followers, has been sharing and how it has been interacting with its audience.
With just a few clicks we’re able to get the data we wish and upload it to Graphext, visualizing the following network, where each node is a piece of media (photo or video) shared by the account, and have been clustered according only to their graphical content.
Looking to the left, we get a pretty concise description of the clusters formed, with the words that describe the content of the image or video.
As this is a product driven account, we can see that we’ve automatically been able to differentiate between different kinds of products, like the vehicles (orange cluster), electronic products (teal cluster) and many others, including posts about other things other than the product itself like communications, design…
As an example, we can highlight the orange cluster and get a better view of the vehicle media.
Now we should be able to tell which type of content gets more love from their audience. To do this, we unselect the orange cluster and then select the top posts according to our second variable (you can think of it as the column of a table) with a simple click and drag.
We see by doing this that while the blue cluster gets more content, the orange cluster is the one that has more outstanding posts, and clusters like the yellow one get very few outstanding posts compared to others with similar volume.
Also we can do a similar quick analysis with any column, like the comment count, to see that the public relations (green) and the sunglasses (pink) clusters are the ones that get significantly less comments than they should.
Combining this with the ability to filter every attribute of the picture like caption, picture description or date, we get a very powerful tool to design the perfect post, and we also get the ability to filter for sensitive or violent content automatically flagged by the AI.