March 22, 2018
KPIs Big Data Analysis Rabbit
AI has captured the imagination of marketers as much as the movie Star Wars captured the imagination of movie-goers when it first came out. Last year Dave Isbitski, chief evangelist for Amazon, reached for the Star Wars character C3PO when he began his Yext Onward talk about AI and “voice driven everything.”
So how would the enormous fascination marketers are developing for AI lead to changes in analytics reporting?
Corporate leaders are certain deliberate with their AI fascination, even while acknowledging challenges. Respondents to an iAB annual benchmark study, for example, noted that AI would receive “more attention” as a significant data tool, despite over half or respondent also noting “insufficient tech” as an obstacle.
As a consequence AI-influenced martech components like image recognition, chatbots, and voice recognition are being added to websites and apps. AI has thus added more dynamic, and potentially correlated, actions for an analytics solution to track.
Examples of that “more actions to track” have arrived in the form of predictive capabilities incorporated into mainstream analytics solutions. These machine learning features serve as an analysis aid. The Data Workbench in Adobe Analytics, for example, includes a number of machine learning features such as conversion propensity modeling, correlation analysis, and cluster analysis. Google Analytics also offers machine-learning features in its Smart Goals and Audience Reports.
While truly valuable to the analysis process, the answers from a machine learning-influenced feature offer precision about the details rather than the significance of the details. Let me explain why this distinction is important to understanding how analytics reporting needs to advance to keep up with AI-driven marketing tech.
When people hear the word precision, they can immediately imagine darts that hit close to a bull’s eye as an example. But in statistics parlance, precision really describes how well attributes agree with each other. So when a machine learning aid is assisting the analytics solutions, it is addressing metric attributes to see what can be grouped together. It is useful, but may not radically impact the purpose behind examining those reported metrics.
AI, however, focuses on the broader concept of devices that act smart on a series of tasks. AI encompasses a larger set of algorithmic applications – machine learning is actually a subset of AI, despite the common usage of both terms interchangeably.
So measuring AI activity slightly shifts the analytical viewpoint from understanding nuanced details to asking about the significance of those details. So – to use an analogy – if machine learning is helping you gather the play by play details of a sports event, an AI output can help clarify the significance of those details.
That perspective can help us understanding what to look for when examining AI marketing tech. Here are some examples on how this can work:
- Image recognition: AI analyzes the pixels in an image file to highlight significant differences that identify an object within the image. So an AI algorithm can detect, for example, if a dog, a car, or street sign is within an image
- Voice recognition: Similar to the image recognition process, AI examines file attributes; except in this case, the focus is on audio files. AI surfaces significant differences that identify the word being spoken
- Chatbots: In chatbots software, AI, in the form of natural language processing NLP, responds to queries with answers
In each of these scenarios, AI was applied to clarify the relevance and importance of results. So what should marketers look at in analytics, when AI is a factor? A potential starting point is mobile engagement metrics from mobile-related visitor traffic. Smartphones are a key environment for image recognition, voice assistance, and access to chatbots, so measuring mobile user engagement verifies an audience exists for adopting the AI-tech in the first place. Despite the rise of mobile-first indexing and increased smartphone usage for search, many businesses still have not adjusted their websites to be mobile friendly.
As for analytics dashboards, marketers should look for aids to clarifying their understanding of results. Marketers need to gain a sense of understanding of the dynamics behind the metrics. For example, Adobe Sensei, which uses statistical analysis to highlight important hidden data relationships, is an example of the kind of clarification marketers should seek.
Direct analytics for AI marketing tech are also available. For chatbots, marketers can use analytics platforms like Google Chatbase to understand what phrases are mentioned frequently, or what requests a chatbots has difficulty interpreting. It’s amazing how fast AI marketing tech has become of mainstream relevance to business analytics. A smart strategy lead to analytics derived from AI applications supporting business concerns very well.