Data Science Drives Better-Performing Influencer Campaigns
It was less than 10 years ago, marketers had to rely on self-reported numbers for influencer
marketing campaigns. Those days are long gone with the advent of advanced analytics offerings.
The days when influencers would work for hire based on free product are done. Influencers have come to understand the value they provide for brands. Because of that shift, Earned Media Value (EMV) should no longer be the metric used to diagnose influencer campaign success.
There is no doubt Influencer Marketing is evolving at lightning speed and will be dramatically transformed by new technology. There are already many similar, rapid changes in the world around us: the potential of self-driving cars, robotics in the home (think, Alexa and Roomba), to fitness and healthcare applications. Artificial Intelligence (AI) and deep learning with neural networks are already bringing about big changes – and are now being deployed by influencer marketing companies.
Through data, we will be able to uncover key insights and use that data to strategically measure the effectiveness and true impact of influencers.
Artificial Intelligence (AI)
The emergence of AI as a buzzword is relatively new, but the core technology and idea have been around for decades. AI is the application of systematic computer frameworks capable of performing complex operations modeled from human intelligence. It allows us to design learning algorithms and work with large data sets. AI is widely being applied across industries due to advances in computing power and more open source platforms for data scientists to efficiently launch their algorithms to production environments.
For example, Collective Bias’ Social Fabric platform is powered by AI technology that collects millions of data points in a real-time feed to discover and match influencers with CPG brands using Natural Language Processing (NLP), Affinity Analytics, and Machine Learning (ML) Classifier Models based on influencer performance metrics.
Why Natural Language Processing (NLP)?
Think about Natural Language Processing as you would your iPhone and Siri. Siri is a form of NLP. The computer must first understand words spoken into the microphone to convert individual sounds and audio frequencies into text. The power of Siri and voice-to-text is in the NLP engine that parses words to extract the context and semantics of spoken words and sentences. The accuracy of speech recognition technology has improved dramatically with the advent of deep learning and machine learning techniques, reducing error rates to less than 10 percent. Siri learns your patterns of writing or your speaking voice and corrects your writing based on your historical patterns.
Collective Bias utilizes similar technology and methods pioneered in academia furthered by Apple and other software companies. It starts with tokenizing and vectorizing words to find the most impactful keywords in context. For example, geolocating your iPhone when you ask about the “weather.” Alternatively, in the case of CB, an influencer’s tendency to apply for more campaigns containing the word “recipe” coupled with given affinities for favorite products. The latter kicks off a stream of cosine similarity metrics linked to those keywords, and the result shows which CPG brands best match an influencer’s natural preferences and propensities. In short, it allows us to pair influencers and brands with the highest probability of success. The model learns patterns in speech/context paired with additional algorithms that are designed to determine historical performance benchmarks. The quality of predictions is continually enhanced as the number of influencer campaigns grows within each brand, category and subcategory.
Why Deep Learning?
Deep learning is the future horizon of Influencer Marketing and its evolutions. It is a form of machine learning that layers representations of variables – often called neural networks. This can be used to uncover patterns in data. Deep Learning builds upon a principle of assembling advanced resources that learn to represent the world with accuracy and speed across petabytes of data.
Facebook’s AI Director Yann LeCun explains deep learning with an illustration: “Imagine a box with 500 million knobs, 1,000 light bulbs, and 10 million images to train it with. That’s what a typical Deep Learning system is.”
Massive distributed datasets with complicated idiosyncrasies have pressed the need for data scientists to evolve their techniques into advanced deep learning. Algorithms that have consistently been included in the data scientist's’ toolbox, such as linear regression, tree-based methods, and gradient boosting machines, no longer perform the computation required to achieve suitable performance on large datasets. The performance is not optimal, as training time can take hours and the predictions are suboptimal.
Our complex web of data sources provides the framework in which deep learning exercises can take place, and the influencer marketing landscape of tomorrow (much like all aspects of social and digital) will rely on our ability to properly map and interpret these datasets with the help of our skilled data scientists.
Why Neural Networks?
Neural networks are applied to many different industries, including retail and sales. Forecasting sales, marketing with consumer segmentation and targeting, and banking and finance with derivative securities pricing decisions are all examples of how they can be applied in these areas.
In the influencer marketing space, neural networks are built for many applications, but the newest and most relevant case is image processing and object recognition. Influencers rely on visual imagery to capture the consumer’s attention. Neural networks are commonly adept at analyzing the color, contrast, and the objects pixel by pixel to determine the optimal images to include in a post. More important, object and face recognition allow us to scan first for brands and, subsequently for sentiment, reading the emotions and facial expression in the photos. We couple that with search criteria, allowing us to find influencers able to stay on point with brand messaging that plays out in the grander social strategy of manufacturers.
Each of these technologies can be tailored for different tasks and requires thoughtful application of data to help create solutions at scale.
This article was provided by Collective Bias, an Inmar Social Engagement Company.