Search-Driven and Natural Language Analytics: What It Is and Why Every Data-Driven Organization Needs It

Search engines have revolutionized the way people get answers in their everyday lives. You simply type in what you’re looking for and have answers served up to you. The search paradigm is so powerful that you find it inside every app and website. The search experience is even more useful when it is coupled with suggested searches and recommendations to things you may be interested in. And these days, search has incorporated natural language voice input and response, so that you can have a conversation with Alexa, Siri, or Google. In this blog, we will explore how search-driven and natural language analytics incorporates these experiences that we all know and love to make business intelligence easier and more accessible for everyone in the data-driven enterprise.

What is Search-Driven Analytics?

Search-driven analytics is a capability in a business intelligence and analytics system that allows the user to enter a question into a search bar and get answers in the form of visualizations. In this way, users do not have write SQL (Structured Query Language) or drag-and-drop from a list of column names (measures and dimensions) in order to query their data. It is an effective form of ad hoc exploration and analysis, where users ask new questions that are not addressed in an existing dashboard or report that has been pre-made for them. Search-driven analytics, sometimes referred to as BI Search, makes it simple for anyone to ask questions from their business data and is particularly helpful to non-technical users who find it too difficult to utilize traditional means of getting analytical insights.

What is Natural Language Analytics?

Natural language analytics is a form of search-driven analytics that makes it even easier for users to ask questions, by understanding input in natural language. BI Search traditionally only accepted “keyword search” input, where the entry has to be very structured, match specific syntax, and match exactly to column names and data values. Natural language query systems, by contrast, accept more free form input and can understand a wider range of user input to produce a result. Natural language search usually implements machine learning in order improve results over time by learning from the broad range of user input and feedback.

What is Conversational Analytics?

Conversational analytics incorporates a few additional important capabilities outside of search to allow users to have a dialogue with the BI system. The first element is voice input so users can literally talk to their data. The second is voice response where the system reads information to the user like a voice assistant. Finally, true conversational analytics maintains context of the dialogue so someone can ask for follow-on or more detailed information without having to repeat the entire question. This can sometimes take the form of a data analytics assistant where you are conversing in a chatbot-like experience.

What are Examples of Search-Driven Analytics?

Let’s say you are an executive, sales manager, or someone in a sales function who wants some data points on performance. Here are few examples of searches you can make and how you would visualize them.

• Show me revenue by region and department last year would show a grouped column chart for the revenue metrics you asked for
• Show me the five states with the lowest sneaker revenue last week displays a column bar chart for the 5 states
• What was monthly revenue for Nike apparel displays a line chart showing sales trend by month
• How did sales for Nike compare to Under Armour last quarter gives an analysis of segments that showed the differences between sales of two brands

Isn’t Search Just a Simpler Version of SQL?

Not necessarily. Let’s go back to the Google analogy. Sure, Google helps you find relevant search pages and YouTube videos, but it can do countless other things too, such as be a personal time-keeper, perform mathematical calculations, convert Fahrenheit to Celsius, and find cheap flights. So while the typical search BI system just translates your keyword input to a SQL query that runs against your database, the most sophisticated systems with built-in machine learning can automate advanced analysis to uncover the reasons why your metrics are underperforming and even segment customers who are most likely to purchase.

Why Does Search-Driven and Natural Language Analytics Matter?

For decades, industry analysts have estimated that the adoption of business intelligence tools has hovered in the range of 20-30% of users in an organization. This ceiling has existed in spite of continued advancements in data analytics technologies – and even in the face of the popularity of visual analytics and data discovery tools. Analytics software has often been criticized as being too difficult to use, but with the increasing adoption of search, that should all change.

Search makes data analytics accessible to all users. There is no much potential to finally get insights into the hands of the underserved – business users and those with little to no technical skills – and satisfy those in the front lines of the business who have never had quick access to data to improve their decision making. Regarded analyst and advisory firm Gartner expects adoption of analytics to jump from 35% to 50% of organizations in the next few years with the deployment of augmented analytics that incorporates natural language search and machine learning automation that will make data insights accessible to new classes of users and applications.

What are the Critical Capabilities for Search & Natural Language Analytics?

• Natural Language Query – The system understands natural language input, whether typed or spoken, and translates that into a data query or analysis.
• Search Suggestions – The system suggests search terms and phrases as the user makes their entry, guiding them to complete their question faster.
• Voice Search – Users ask a question by speaking into the system using voice, and can even have the response spoken back, like Alexa, Siri, or Google.
• Data visualization – The system returns a best-fit visualization to the user with the answer, and also allows the users to change and customize the visualization.
• Natural Language Generation – The system presents a narrative to the user (written and/or spoken) alongside the visualization that describes the key insight.
• Synonyms – The system understands words that have similar meaning, such as “revenue” and “sales,” and correctly give an answer regardless of which specific terminology is input.
• Learned vocabulary – The system can be taught new words and learn how to properly translate them as part of a query.
• Query transparency – The user can clearly see the underlying query and data columns used to generate an answer.
• Auto-indexing – The system automatically organizes the data to support the analyses that are run, even when new data is brought into the system.
• Performance and scalability – The system generates answers within the expectations of its users (usually within a few seconds) regardless of how much data is needed for the query or analysis (even if it is more than billions of records).
• Dashboards – Users can combine visualizations into a dashboard and share them with other users.
• Automated discovery of insights – The system automatically presents important and meaningful insights to the user related to what they have searched on, without the user asking for a specific type of analysis; this uses machine learning algorithms to uncover such discoveries.

What Types of Analysis Can Be Performed by Search?

Here are common types of questions and analysis that can be conducted with search-driven and conversational analytics.

• Aggregation and Group by – The user can get a calculated summary of the data they have asked for, such as a sum, average, and count, which can be grouped by another value, e.g. show total sales by region and department.
• Filtering – The user can choose to a constrain the set of data in the analysis, usually based a specific value they have chosen, e.g. what were sales in California
• Top/Bottom – The user can see the top or bottom X number of values for a metric; e.g. what were the top 10 selling brands last year?
• Time periods – The user can ask for metrics from the most recent time period, such as last 2 weeks or last year.
• Comparisons – The user can compare metrics for two or more groups, such as comparing sales of Samsung versus LG.
• Trends – The user can see trends over time for specific metrics, e.g. show me weekly revenue In Florida.
• “Why” analysis – The user can request in-depth advanced analysis that explains performance changes, particularly between time periods, e.g. why did sales of electronics increase in Texas last quarter?

What Can Users Ask in Search-Driven and Natural Language Analytics?

Below are some more examples of questions users can ask in search.

Sales

• Tell me more about revenue and cost by brand and city
• What drove customer churn last week?
• What is my monthly revenue by state for Nike?
• Compare revenue between Samsung and Sony in New York
• Why did sales increase last year?

Marketing

• Tell me more about leads and cost by campaign and channel
• What are my top five states with the lowest subscriber growth last year?
• What is my monthly spend by channel on Tier 1 customers?
• Compare leads generated between Facebook and Instagram in New York
• Why did leads increase last year?

Finance

• Tell me more about credit balances and savings balances by city and income
• What are my top three campaign offers being accepted by the most customers?
• Show me monthly customers broken down by campaign.
• Show me the comparison of personal loan customers between Dallas vs. Houston.
• Why did assets increase in Virginia last month?

Healthcare

• How many patients were admitted? Over the last week? Last month(s)? Last year?
• Show me all patients that were admitted with a case of influenza in the last month.
• Why did the use of CAT scans increase year over year?
• Show me all patients that were admitted multiple times this month.
• What drove patient readmissions last year?

How Does Tellius Fit In?

Tellius is an AI-driven Decision Intelligence platform that supports Search and Natural Language analytics to make data accessible to everyone in the enterprise, so people can instantly access relevant insights in the most intuitive way. What makes Tellius Search different:

• Natural language – Tellius interprets free form input, typed or spoken, into a data query without the need for a rigid structure when asking questions
• Conversational – You can have to a dialogue with your data supporting both text and voice interactions, maintaining the context of the conversation, creating the experience of talking to a virtual AI data analytics assistant
• Intelligent AI automation – You can ask “Why” questions that utilize machine learning algorithms that quickly analyze your data to surface underlying trends and business drivers; the system also uncovers important findings and patterns in the data based on your interests that you did not explicitly ask for.
• Elastic Scalability – You get instant response from billions of records or more, even if data is spread across multiple data sources; the system dynamically adds and removes resources to match current workload, with zero maintenance required.
• Data Preparation at Scale – Smart data preparation capabilities are available to enable your team to shape and transform the data for analysis.

To learn more about about search-driven and conversational analytics watch the Conversational AI-Driven Analytics webinar.