October 31, 2019

The size of the influencer marketing industry has grown dramatically in recent years. What started out as a few savvy brands identifying influential social media users with huge followings and paying them to promote – subtly or explicitly – their products, has become a global industry. Big brands are investing big bucks to ensure that their products are present in all the right places and on all the right profiles across social media. But if you’re one of the companies paying to make sure your products are present on social media, you need to make sure that they’re perfectly placed and that they’re achieving the desired effect. Łukasz Dylewski, Head of the Data Science team at ITMAGINATION, explains how Artificial Intelligence (AI) could be the answer to improving the effectiveness of your influencer marketing.


 

Most of us have encountered influencer marketing in some form or other on social media. Whether it’s one of the Jenner-Kardashian clan, Lionel Messi or Justin Bieber or even just the guy at your local gym taking an unusually high number of selfies, the power of endorsements by ‘real people’ on social media is attracting an increasingly large portion of the budget of many brands – large and small – from all around the world. According to Business Insider, brands are expected to spend up to $15 billion on influencer marketing by 2022, and 79% of brands that spend money on influencer marketing invest in Instagram – a far higher number than any other platform. In 2018, Hubspot reported that since 2015, the topics featuring the highest percentage of sponsored posts on Instagram were: fashion, beauty, fitness and food.

Social media is seen as an opportunity for us to see what the real lives of people – friends or idols – look like. And if we see a product seemingly being endorsed in a ‘real life’ context by the people we follow, it’s only natural that we’ll recall it better in the future and be more likely to buy that product. Brands know this only too well, and so are investing money in influencer marketing. But as the industry matures and investment increases, these brands (and the media or ad agencies that they work with) are increasingly looking to measure the performance and return on investment (RoI) of their influencers. After all, if you’re paying somebody to promote your product, you naturally want to know that they’re doing it well.

Up until recently, most companies have been doing this manually, with people checking the posts of their influencers, compiling their findings (which are often subjective) and producing reports. This takes time, uses up valuable human resources and often produces fallible findings. It doesn’t have to be this way.

How can technology show you how effectively your influencers are influencing?

ITMAGINATION has pioneered a solution that empowers companies to evaluate how effectively paid influencers are influencing on social media. Using Artificial Intelligence (AI), Deep Learning and Convolutional Neural Networks (CNN), the solution developed by ITMAGINATION is able to analyze the photos posted by identified users of Instagram (e.g. a list of influencers) and evaluate them based on factors like:

  • Does the image contain a picture of the product that is supposed to be promoted?
  • How much of the photo is occupied by the product?
  • Is the lighting appropriate? (i.e. can we see the product clearly?)

Such findings can then be extrapolated and enhanced to consider performance indicators such as:

  • What is the frequency or proportion of posts that feature a specific product or brand?
  • How many followers are products being showcased to? (i.e. the reach)
  • What type of, and how many, reactions does the post generate?
  • What is the sentiment (e.g. positive, negative, neutral) of the post and the reaction that it garners? (e.g. the text that accompanies it and any emotions detected within the image and the comments that the post garners)

Equipped with this data, companies are able to better understand the impact of influencers’ work on the performance of a brand. By ‘performance’ we can consider factors such as the brand’s own following on social media channels and also on sales. Companies can also use these findings to optimally structure their relationships with influencers and compensate them in a way that is reflective of their performance and the impact it generates. For example, influencer A might have a high number of engaged followers but that reach might not be fully maximized because he or she has a habit of using images that aren’t always great (i.e. low proportion of image occupied by product, poor lighting, etc.). With this solution, the influencer can be provided with data that he or she can use to improve the quality of their posts. The brand paying the influencer can decide whether to link the influencer’s remuneration to these performance metrics, thus ensuring appropriate, tangible value for money and return on investment. Importantly for the brand, less human time and effort is spent evaluating posts, which means that valuable time can be re-invested in initiatives that have strategic value to the brand.

image recognition

What’s more, the AI-based solution pioneered by ITMAGINATION can also be ‘taught’ to detect competitor products and can be ‘plugged in’ to entire user groups or communities. This enables brands to see how their products are being promoted and how influencers are performing within a specific group of users (e.g. users using specific hashtags) and to identify and evaluate new influencers before entering into agreements with them. In this way, brands can use insights generated from data to adjust and optimize their own influencer marketing strategies.

How does this ‘product detection on social media’ solution work?

The solution utilizes a variety of different technologies from within the domain of Artificial Intelligence (AI) and Deep Learning, such as Convolutional Neural Networks (CNN) and Computer Vison. The solution is taught to recognize specific objects within images. This is done by analyzing a large amount of different images of the product with variances between them, such as exposure, lighting, focus and many more.

image recognition

In this way, the system ‘learns’ to detect and identify the product in a variety of environments and conditions (such as those found in images posted by users to Instagram). It’s not as simple as simply teaching the system to recognize just one product. First, classification takes place, so let’s say that we’re talking about a beauty product that comes in a bottle. The system is taught to first detect any bottle present in an image, and then conduct a series of classifications (e.g. shape, transparency, label) to determine if it’s the exact product that we’re seeking to identify. This ‘classification’ process is important, as it enables transfer learning to take place. This means that, if the brand using the solution starts with one product in a bottle and wants to shift its attention to (or extend its area of focus to include) another product that is in a bottle, a portion of the learning can be transferred, which will save the brand time and money. Once the process of detecting and classifying an object has been completed, the system can be taught to identify the surrounding context of the image, such as lighting conditions, surrounding elements, presence of people, proportion of image occupied by object, etc.

In terms of how the end user or beneficiary of the solution (let’s say a Brand Manager or Marketer) uses this technology, all they need to do is provide details of the product and a list of influencers (or other Instagram users, if being used for competitive analysis) and the system will be able to identify Instagram posts by those users and then evaluate them based on the criteria set within the system by the Brand Manager or Marketer. The output isn’t just raw data – the system can be configured to show performance against KPIs. In this way, the solution developed by ITMAGINATION empowers companies to manage their relationships with influencers more effectively and optimize their outlay on this increasingly important marketing and advertising function. 

How else can AI be used to optimize marketing and advertising spend?

The AI-in-social-media solution developed by ITMAGINATION is fully operational and ready for deployment with Instagram as per the use cases described above. But this is only the start of the possibilities. ITMAGINATION has been preparing the solution for deployment in many other scenarios, including:

  • Connected to YouTube channels or specific users to detect specific products, measure their prominence on screen, the environment in which they are being shown and how long they are on screen for. Furthermore, the AI-based solution can be ‘taught’ to detect positive, negative and neutral sentiment in accompanying audio. In this way, brands can see how their products are appearing on YouTube and understand how they are being spoken about.
  • Plugged into online television TV) or Video-on-Demand (VOD) platforms to measure effectiveness of product placement. The solution can be configured to monitor the content of online TV or VOD platforms to detect if, when and for how long specific products are on screen. This allows brands to measure how much exposure their products get and how well product placement (e.g. within televisions shows) initiatives are performing. The findings can be used to compare effectiveness versus the competition, to analyze the effectiveness of marketing spend and make adjustments, to ensure appropriate exposure at key times, and much more. Furthermore, this data can be configured across systems (e.g. Sales and CRM) to better understand the relationship between product exposure on different media and sales.
  • Tuned into live broadcasts (e.g. sports) to identify logos. ITMAGINATION’s solution can be tuned to detect specific logos in much the same way it would with a product. For example, if a brand wants to evaluate the viability of sponsoring a sporting event, team or athlete, it can plug the solution into online TV broadcasts and learn how often the brand logo appears on screen, for how long, and the quality of the appearance (e.g. size and clarity on screen). In this way, brands can make better-informed decisions about whether such sponsorships bring about the desired results.

Embrace AI to optimize the effectiveness of your investment on marketing and advertising with ITMAGINATION.

The solution developed by ITMAGINATION can empower you to better understand the effectiveness of your marketing and advertising investment, and optimize it based on data and insights. With this, you can manage the performance of your various channels – influencers, advertising, sponsorship and product placement – in a data-driven way and optimize your future strategy. If increasing the effectiveness of your marketing and advertising investment sounds like something that can help your business, talk to ITMAGINATION.

 

Learn it. Know it. Done.