Visual analytics tools allow business analysts and other users to query and combine data sets using point-and-click gestures in a visual interface, instead of actually writing out queries in a programming language like SQL.
These tools represent a significant advancement in the modern, “self-service” model of BI. In this model, business analysts access and query data themselves, instead of accessing and querying it through technologies controlled by the IT department.
Visualization is key to self-service BI, since it’s a way for users who don’t know how to write queries themselves to retrieve the data they need. Users can perform analytical operations merely by clicking on pie charts, adding new dimensions to maps etc., instead of expressing such operations in SQL or another language.
Since visual analytics is still an evolving technology, we’ll describe the major capabilities these tools offer. We’ll also explain how the market breaks down, since visual analytics capabilities are found in various types of BI solutions.
We’ll guide you through the following topics:
Visual Analytics != Dashboards
Capabilities of Visual Analytics Software
Visual Data Discovery vs. Visual Analytics in Traditional BI Systems
Choosing: Dedicated Visual Analytics Platform, or Traditional BI?
Many readers will know enough SQL to recognize “!=” as the “does not equal” operator rather than a typo, but if not, that’s precisely why you need a visual analytics system. SQL syntax becomes even more complex once you go beyond the basic operators.
Visual analytics tools are frequently confused with dashboards. Let’s take a look at why.
Exhibit A is an actual dashboard:
Exhibit A: Sales manager dashboard in Board
Exhibit B is the interface of a visual analytics solution during analysis:
Exhibit B: Visual analysis of accident reports in Qlik Sense
At first glance, it can be very difficult to tell the difference between these two visual interfaces for presenting trends in data. There are, however, a few, including:
Dashboards are templated visualizations of KPIs that integrate data from a variety of operational sources: CRM systems, e-commerce/order processing platforms, inventory management systems, accounting systems, supply chain management systems etc. They either update in real-time or are regularly refreshed with new data. Most dashboards aim to help end users (sales managers, call center agents etc.) understand their individual performance or the business’s performance.
Interactivity is highly limited in a dashboard, because analysts in conjunction with business leaders determine how performance is calculated—not the end users. Users may be able to click on a chart element to get more details on a KPI, but they can’t decide, for instance, to swap out all of the line graphs in the dashboard with scatter plots, or to blend the data in the dashboard with a spreadsheet on their desktop.
Visual analytics graphical user interfaces (GUIs), on the other hand, are blank slates for accessing and manipulating data sets with point-and-click, drag-and-drop gestures on visual data displays (pie charts, tree graphs, heat maps, scatter plots etc.). Whereas the business “freezes” KPI calculations into dashboards, visual analytics tools are designed for free-form visual analysis of any old data set: a spreadsheet, a SQL database, a NoSQL database etc. Moreover, users can blend data from multiple sources during analysis, instead of having to rely on the blends that have been built into a dashboard.
Users thus choose the visualization types they want to use in visual analytics software. If one chart type doesn’t work, another can be used in its place. Users also choose the dimensions (data categories such as customer, product etc.) and measures (numerical values like the number of items sold in a given transaction) that they want to combine in these visualizations. Generally, analysis is a process in these tools—once a pattern has been spotted, the user explores it with further visualizations.
Visual analytics tools are thus specifically designed for business analysts who spend all day spotting new patterns in business data to explain problems and highlight opportunities.
The following table summarizes the differences:
|Dashboard||Visual analytics interface|
|User base||End users throughout organization||Business analysts and other data explorers|
|Purpose||Present role-specific KPIs||Facilitate free-form analysis|
|Level of interactivity||Minimal||High|
|Data connections||Prebuilt||Ad hoc|
Visual analytics tools generally offer the following capabilities:
|Visual GUI||A visual interface supports data manipulation via drag-and-drop gestures rather than SQL clauses.|
|Library of templated chart types||Users can pick from bar charts, heat-maps, treemaps, scatter plots, bubble charts and a range of other visualization operations. Many tools will even recommend an appropriate visualization based on the data.|
|Ability to promote visualizations to dashboards||Analysts can template KPI analyses as dashboards and share them across the organization (generally requires a server license in addition to user licenses).|
|Ad hoc data connections||These tools can connect directly to a wide range of data sources, including spreadsheets, relational databases, NoSQL databases, cloud data sources etc.|
|Data blending||Users can combine data from different sources on the fly to discover new insights.|
|Linked visualizations||If a user alters one element of a visualization (say by adding a new dimension), the other elements will update automatically.|
|Data cleaning/preparation||Since data access in visual analytics software is frequently ad hoc, data typically needs to be prepared for analysis with features for normalizing fields, removing trailing spaces etc.|
|Back-end SQL engine||Visual analytics software includes an engine that translates users’ gestures into SQL queries.|
|In-memory data cache||These tools also process data in random access memory (RAM) instead of writing it to disk, which allows for rapid processing of huge data sets.|
Visual analytics tools—also known as data discovery tools—evolved as a response to two problems with traditional BI systems:
Dashboards and static reports are a strength of traditional BI systems, since in these systems the IT department works alongside analysts to extract data from operational databases, calculate metrics and push KPIs out to end users via PDF reports, dashboards or some other medium.
In this use case, free-form analysis by the end user isn’t necessary or even encouraged. Instead, the organization standardizes on a single data model (a schematization of the relationships between data types, data sources etc.), which is then built into the BI system.
Visual data discovery tools thus evolved for those end users who do need to perform free-form analysis, i.e. business analysts, since dashboards and scheduled KPI reports aren’t enough for these users.
It may seem that visual data discovery tools have a clear edge over traditional BI. However, data discovery tools suffer precisely because of the freedom they enable. One analyst may use a different process to visualize data than another, which makes it possible for the analysts to wind up with two significantly different interpretations of the same data set.
Traditional BI systems were designed to control access to data such that companies had a “single source of truth” about business performance metrics. Data discovery tools are catching up in this regard by introducing data governance features (role-specific access to certain data sources, data modeling languages etc.). However, they’re not as robust in this area as traditional systems.
Data modeling in Looker BI
Moreover, traditional systems have now incorporated many of the visual analysis features originally found only in data discovery tools. Both visual data discovery tools and traditional systems can be used to create dashboards.
The following table presents the most important selection criteria for deciding between these options:
|Visual data discovery tool||Traditional BI system|
|Free-form visual analysis of data|
|Regularly scheduled batch extractions of data from operational databases (extract, transform and load)|
|Standardized and centrally governed data model serving as a “single source of truth”|
|Collaborative data modeling among workgroups|
|Ad hoc connections to new data sources and recombinations of data sources|
|Organization-wide deployment for end users|
|Workgroup deployments for analysts|
These criteria unfortunately aren’t always as clear as they’d ideally be, since, as we’ve seen, the distinction between these categories is gradually eroding away.
We're able to offer this service to buyers for free, because software vendors pay us on a "pay-per-lead" basis. Buyers get great advice. Sellers get great referrals.