A robot will check an influencer

More and more brands are collaborating with influencers on social media to reach the widest possible audience of potential recipients. How do you assess the effectiveness of such a collaboration from the perspective of a company's expectations? Artificial intelligence, machine learning and data analytics can help. ITMAGINATION created and implemented, as a prototype at a client's site, an image recognition system for social media. Its main goal is verifying the quality of influencers' work, for which companies pay to position their brand and products.

What is the system used for?

The system created by ITMAGINATION enables effective monitoring of brand positioning in a selected social media channel, thanks to the automation and digitization of a process that until now required human intervention. It also makes it possible to analyze and compare campaign results in collaboration with one or several influencers, as well as to track similar activities of the company's competitors. Thanks to advanced technologies, the system helps verify the effectiveness of advertising campaigns, enabling comparison of the results of individual activities over time. That is why it may work best in long-term collaboration with an influencer.

The power of neural networks

– Advanced data analytics can work wonders. Thanks to the use of convolutional neural networks, our system based on Python and Tensor Flow technology can check hundreds of thousands of photos, which allows for a reliable examination of the market. Sometimes it is enough to provide a hashtag, the name of a product or the profile we want to analyze. This solution can be used across various social networks. However, because its primary purpose is image recognition, Instagram seems to be the best infrastructure for such analysis – notes Łukasz Dylewski, Data Science Team Manager at ITMAGINATION.

The project carried out by ITMAGINATION at a client's site involved identifying the space occupied by a specific product of that company in photos by more than a dozen selected influencers on Instagram. The use of deep machine learning makes it possible to train the model to recognize the client's brand, as well as the type and shape of the analyzed product. This requires analyzing a large number of various photos with products so that, as a result, the system can independently distinguish it from others and also identify the circumstances and condition in which it appears in influencers' photos.

The potential of advanced analytics

How does it work in practice?

How does the system work in practice? It is enough for the person responsible for monitoring the campaign to enter the influencer's nickname, and the name and type of the product to be analyzed. The system identifies all photos published by the influencer in which the given product appears. The automated system is able to measure the area of the photo that the product occupies, analyze what reaction the photo triggers among users and how, as a result, they perceive the advertised product. The system counts the post's likes and the comments under it, and can also be extended with sentiment classification, e.g. into positive, neutral and negative comments. All of this fully automatically and precisely, thanks to the use of advanced, cutting-edge technologies.

Robots that check influencers — how do they work in practice?

Based on the analysis carried out, the system prepares an aggregated report for the client. It presents KPI statistics, detailed data on the size and placement of the product in a given photo, as well as information about the circumstances and context in which it appeared – for example, whether the photo was taken outdoors or indoors, in bright or dark light. It also describes other objects present in the photo and identifies whether products from other brands are in it. Thus, the solution makes it easy to assess how the influencer handled a given assignment and whether they fulfilled the terms of the contract.

Many functions, one goal

– The image and text recognition system has many useful applications in all areas of marketing. It can work well, for example, in sports broadcasts, where there is always a lot of sponsorship activity. Thanks to the technologies used, it can identify logos and individual brand products during coverage and analyze whether they were placed in line with the sponsor's assumptions. Instagram is open to integration and the use of new technologies, which is why carrying out this type of analysis is quite easy to implement for business purposes – says Łukasz Dylewski of ITMAGINATION.

Support at the planning stage

The system supports the marketer not only at the post-campaign analysis stage. Its analytical capabilities will also be useful at the campaign design stage. Based on data from the selected platform, this solution is able to propose a group of influencers for collaboration who match the business goals and a specific company product, for example in terms of reach, target group or the photos that users "like" and publish. The client can choose the metric themselves and decide on what basis to assess the positioning of their brand in social media – whether it will be, for example, the number of "likes," the reach of a given post, the interests of users who interacted with the post, or the tone of positive or negative comments.

Summary

ITMAGINATION developed a system based on artificial intelligence and deep machine learning that automatically verifies influencers' work. The tool analyzes Instagram photos for the presence, size and context of a brand's product, and also measures user reactions, including the number of likes and the sentiment of comments. As a result, companies can assess whether the influencer fulfilled the contract and what real effects the campaign delivered.

The system can also analyze competitors' activities and recommend creators for future collaboration based on data about reach and target group. The technology can also be used outside social media — for example, in sports broadcasts, where it is used to verify the exposure of sponsors' logos. The solution works best for long-term campaigns, where comparing results over time is important.