Use Kognitio to empower Qlik on demand app generation

Video Transcript

In this short video we show how Kognitio performance empowers Qlik Sense by performing this On-Demand App Generation

Firstly let’s take a look at the Kognitio and Qlik architecture

Raw data is held in any persistent data source…

with data used by Qlik pinned into Kognitio RAM.

How does Qlik ODAG work?

Explore your data via overview in Qlik Selection App.

Once your selection is made generate more detailed data using ODAG to query Kognitio.

Using Kognitio to empower ODAG means this detail can be on-demand analytics (such as R code) as well as SQL.

Drill further into the detail by making further selections and generating more Apps.

The Selection App is a fully functional Qlik Dashboard showing a complex overview of the sales data.

You can use all the great associative features in Qlik to explore your data.

In this selection App we can concentrate on specific time periods by filtering the year, quarter or month and compare year-on-year performance.

The App filters the data associated to our selection.

Another tab allows us to filter by Geography presented using Qlik’s built in map.

We can see a subset of stores in specific regions or pick out individual stores to focus on.

In this demo we will concentrate on product hierarchy.

Let’s imagine we are a brand manager…

…and use Qlik search to find our brand.

We can see that most sales are in Confectionery…

…and we have 85 products in the Confectionery Product Group.

We can also see sales are very seasonal.

We want to see forecasts for all products in Confectionery, not just our own products…

…so we remove our brand selection criteria giving us 368 products in total to analyze further.

We now move on to a summary of our selection…

…and select to run a sales forecast with seasonality included.

Qlik’s On Demand App Generation is used to kick-off forecasts for our 368 confectionery products.

There’s some major work going on while you wait!

There’s some major work going on while you wait!

Behind the scenes Kognitio handles all the processing based on your data selection.

Firstly Kognitio runs the SQL query to produce the data to pass into R forecasting models.

Kognitio then seamlessly invokes and manages 368 R processes based on the resources (RAM and CPU) available on your system …

… and automatically co-ordinates the flow of data into R.

The R processes do their thing executing the complex forecasting algorithm…

…and finally Kognitio stores the R modelling results directly in RAM ready for use.

These results are then returned to Qlik for you to look at using the Detail App generated.

We can see most forecasts are valid…

…with some more accurate than others.

Let’s focus on our brand.

We focus on one product…

…and use a second App to run Basket Analysis for this product.

In this second app generation Kognitio is now running basket analysis on-demand over 2 billion rows of transactional data.

This is done using a complex SQL query submitted by the Qlik App based on the product selected.

The query identifies all baskets containing the product and analyzes all the other products bought within these baskets.

This SQL is submitted on-demand, so the platform executing the SQL must be performant enough to process the results while we wait

This query over 2 billion rows of data took less than 40 seconds.

The results can be explored further in the Qlik Detail App…

… focusing on products by department and group…

… and listing the top cross selling products.

In this short video we have seen how Kognitio can be used to empower Qlik’s On Demand App Generation over big data.

When the data gets big you need to take the processing to the data and to do this you need awesome processing performance.

We saw how Qlik’s associative features can be used to allow you to explore an overview of retail sales data and make your selection.

This selection was then used in the generation of Qlik detail apps.

Kognitio’s awesome processing performance allows us to run both complex product forecasts in R and further basket analysis while we wait.