Qlik Associative Analytics: 4 Colors That Reveal More Facts
Traditional, relation-based analysis is usually driven by the database and the SQL-based paradigm of joins, hierarchies and result sets. In this top-down, or tree-based, model, when filtering, drilling or slicing, we narrow the data set we are working with. As my colleague Lou Bushinsky wrote in his Disrupting BI series blog, technology is rapidly reshaping the existing BI landscape, enabling the end user to engage in intuitive data exploration and visualization.
I’d like to take a closer look at how the associative model enables information consumers to reveal new facts that might take data discovery to a different level.
The advent of the in-memory, associative data models comes with the benefits of focused analysis while uncovering new or missing facts. We are no longer limited to questions that the designer had in mind. By filtering, we can reduce and narrow the data set without divorcing from it. It simply reveals more.
In a classic relational-model-based tool, we’d select filters or drill down to narrow the data across a certain path.
Data with no joins or relationship would be filtered out from a visualization, or hidden, leading to a potential loss of important facts.
One of the tools available that embraces this technology is Qlik and its indexing engine (QIX). The example below was demonstrated in Qlik Sense 2.1, a data exploration and visualization tool.
In the Qlik associative model, as the user interacts with the analytics, the visual feedback is provided through the UI, using the colors green, white and shades of grey. With a single selection (starting from anywhere), it shows us what data is selected, what data is related, what data is possible and what data is not related – while still keeping a visual perspective of the entire data set, without removing any data from view. This also helps users understand complex data relationships.
After the data is loaded into the Qlik engine, the system automatically associates and links the tables on common fields that it discovers. The system also optimizes the model to provide the best memory utilization and performance, tailored to potential analytical questions, searches and selections. The outcome also allows for any questions. From this model, the data is available either for business consumers to explore their data or for more savvy analysts and developers to build robust applications.
Marketing campaign analysis scenario
Let’s assume our marketing department would like to measure the effectiveness of campaigns during a month. In the chart below we can see the number of consumer responses over time by media type. As we narrow the selection to a specific time period, we can see on the left there were two campaigns with responses, but also two with no responses: “Friends” and “Small Data,” marked Dark Grey. Even though we applied a filter on the chart, the nonassociated values are still visible.
What’s important is that the Light Grey chart at the bottom highlights the associated opportunity value from both associated and nonassociated campaigns.
In an SQL-driven approach, nonassociated campaigns would drop out of the result set after the filter is applied. We would not be able to see the size of opportunities that were potentially lost or which campaigns got no response in the selected time frame. Qlik Sense keeps both, highlighting them appropriately. The user can further apply the filter and exclude the nonassociated data as needed.