BI Leadership

Why Relational Search is Winning the Battle for BI

Search has taken the world by storm. Every day two billion people use search to get instant access to massive amounts of information. This requires no training or technical expertise, even kids know how to get answers from Google. We search for directions to a store, instructions on how to bake a cake, or for tomorrow’s weather report.

This ease-of-use at scale has inspired a new wave of analytics vendors who are using search to help business people analyze their data at work. By providing the familiar experience of search, even non-technical users can easily build reports and dashboards in seconds.

Yet as we all know from our consumer lives, not all search is created equal. Consider the recent battle between Bing and Google, some people swear by their preferred search bar! Or what about travel—do you use an online agent or go directly to the airline’s site? All similar experiences, but with different results.

So, if you’re considering bringing search into your enterprise analytics stack, it’s important to understand the difference between the various forms of search-based BI solutions.

Different Search for Different Questions

One kind of search we’re all familiar with is traditional web search—tools like Google, Bing, or Amazon. This form of search is often called “object search”. It enables users to search across data and returns a list of answers. This is the most common form of search found in BI tools today. While object search makes finding pre-built dashboards and charts easier, if you want to look at a new chart or get a report updated - you’re out of luck.

A second form of search is called natural language processing (NLP), which was made popular by Jeopardy-winning IBM Watson. NLP uses complex algorithms to interpret the intent of a user’s search and then returns a probabilistic result. Unlike object search, with NLP end users can actually begin to use search to build their own reports. But depending on what you’re searching for, NLP may not be the best kind of search for you. Consider this question: “How many McDonald’s are there in San Francisco?” An NLP engine will have to guess whether you mean “McDonald’s” the restaurant chain, the street name, or the family name. As a result, answers can often be wrong and users often have to search multiple times to try and get at the correct answer.

When you need to know exactly how many units you sold last week or which products grew more than 10% last quarter, any answer that’s not 100% accurate is unacceptable. Further complicating the issue, if you misspell or partially type an extra word by accident, then your intent can be lost and with it the accuracy of your answer.

A New Paradigm for BI

A third kind of search—Relational Search—is a new breed of search-driven analytics that allows anyone to use search to build 100% accurate charts and dashboards out of their relational data, on-the-fly, in a matter of seconds.

For example, with ThoughtSpot’s Relational Search Engine you can simply type in “sales by category name monthly” and the engine will find all sales transactions for that product category, aggregate them by month in real time, and deliver a best-fit chart based on the characteristics of the data.


Unlike object search, relational search engines don’t require BI teams to pre-build dashboards or reports. The engine literally calculates the answers in real time as you type. And unlike NLP search engines, a Relational Search engine does not interpret your search query and then guess at an answer. Relational search deterministically takes each input and calculates a single, precise answer based on each letter you type. The result is an easy-to-use, penalty-free solution that lets non-technical business people truly do their own self-service analyses. And this changes everything—now anyone can be their own data analyst.

If you want to learn more about relational search engines and the architecture that makes them work, check out our latest whitepaper: Relational Search: A New Paradigm for Data Analytics.

×