How AI-Driven Analytics Changes the Question You Ask
How AI-Driven Analytics Changes the Question You Ask
The convergence of NLP and AI in analytics architectures gives everyone access to their own personal data scientist.
Join the DZone community and get the full member experience.Join For Free
Bias comes in a variety of forms, all of them potentially damaging to the efficacy of your ML algorithm. Read how Alegion's Chief Data Scientist discusses the source of most headlines about AI failures here.
If you’re paying just a little attention to what is happening in analytics and business intelligence space, you can see the next wave of evolution forming and gaining speed. This wave is powered by Big Data, Machine Learning, and Natural Language Processing. It has the potential to fundamentally change how organizations empower their people with data and change the questions they ask.
What Is Natural Language Processing?
Natural Language BI Is a Cornerstone for Digital Transformation
Digital transformation, simply put, uses technology to innovate not just improve existing processes. As part of these efforts, the demand for analytics is greater than ever. However, traditional BI tools struggle with broad organizational adoption, usually an anemic 22%. There are various reasons for this but, generally, business users don’t have the technical knowledge or time to learn a self-service BI tool. This leaves them dependant on others for analytics which is the big problem with using traditional BI tools to support Digital Transformation efforts. Your business users are your domain experts. They know the questions to ask that could lead to transformative innovation but because of the inherent complexity of traditional self-service BI, they lack direct access to analytics. Natural Language BI turns self-service analytics on its head by focusing on delivering solutions for non-technical business users, i.e., the other 78%.
With Natural Language BI, business users ask a question, in human language, and get an answer, usually in the form of a dynamically generated chart. They can ask follow-on questions to dig deeper or run what-if scenarios without needing to understand how the data is constructed or even how to build a chart. More importantly, the interactions with data analytics can be done within applications your business users are already familiar with using, like Slack, using NLP APIs. Done right, Natural Language BI has the promise to close the data gap for business users and create the data agility needed to uncover insights that can transform a business.
Changing the Question
According to Gartner, by 2019, natural-language generation will be a standard feature of 90% of modern BI and analytics platforms. That may be an ambitious prediction for some traditional BI tools but newer solutions, like Knowi, have already included natural language processing into their platforms. However, adding NLP to BI platforms is just the beginning of the next evolution.
Let's say you run a SaaS company. Like all SaaS companies, an important metric you pay close attention to is Churn Rate. With natural language queries, you could ask a question like “what was my churn rate for the last three quarters” or “what is my churn rate this quarter compared to last quarter?” A dynamically generated chart will return the answer in seconds. You’ve got your answer and know you need to figure out what that data is telling you. It is obvious that you're trying to understand if actions you’ve taken have caused any changes (good or bad) to your churn rate. However, the question you want answering is “what should I change to reduce my churn rate?”
The answer to this question is at the junction of AI and NLP. NLP enables the question to be asked, in plain English, and AI answers the question. This is also where AI and BI converge to create something new, AI-driven analytics. Instead of thinking of questions to ask and interpreting the answers, answers are automatically surfaced for questions to you hadn’t thought to ask.
Google Ads “recommendations” is a good example of AI-driven analytics in the real world. For those not familiar with Google Ads, it is Google’s online advertising program. I don’t know the inner workings here, but as a user, I can tell you what I use the Recommendations section to do. Before, my questions where “what is my click-through rate for an ad group” or “what is conversion cost for a keyword?” With Recommendations, my question has become “what can I do to improve my ad performance?” Google suggests changes to bids, ads, keywords, etc. and provides an estimated impact of the change. In the example below, Google is telling me I can reduce my bids, i.e., lower my ad spend, for two ad groups.
The impact of applying this recommendation is predicted to be 42 more clicks and $68 less spend per week. It would take me hours to correlate the data points needed and more math than I’m interested in doing uncover this insight. Instead, I would try various things and wait a few days to see if it made any difference. It might take me multiple experiments and several weeks to uncover a similar insight.
I’m not saying reports and dashboards are going away next week, but the future of BI is closer than you might think. The difference between BI and AI-driven analytics is not only speed to insights but the insights themselves. The questions your business users ask will change from informative (tell me what is happening) to directive (tell me what to do). The answers are more likely to directly impact how your business operates and potentially competes in the future. Those able to transition their analytics architectures to deliver AI-driven analytics will have a clear advantage over their BI peers.
Opinions expressed by DZone contributors are their own.