March 1, 2014

QlikView Design Best Practices

edit: glad to have the support of Donald Farmer for this approach :-)

In this post two of my passions coincide: building effective dashboards efficiently and drawing. I draw quite a lot in my spare time lately, so I was thinking, why not explain my view on the design of a dashboard using a drawing?

Flat design

I like flat design, no gradients, no shadows, not even rounded corners - the drawings suggest otherwise, but hey, I did't have my ruler at hand. Recently a very nice blogpost on the Qlikcommunity went a little deeper into this topic.


I don't like the tabrow in QlikView, so I developed a template with a number of tabs and subtabs. This number can be choosen in a .csv-file that sets some variables in the script to determine how many tabs and subtabs you want to have visible (0 - 6 tabs and 0 - 4 tabs). The objects to switch between the sheets are plain text-fields that set a variabla on click. I was never a big fan of the buttons in QlikView but since this blogpost by Steve Dark I've never used them again.

More filters

I often have more fields I want to filter on, than I have space on the dashboard - here sales rep and product. These filters I place on a special tab with the very informative name 'More filters'. When you click this special tab a kind of transparant grey overlay becomes visible in the center of the dashboard, showing the extra filters - this goes for all tabs.


In this template I have also foreseen a translation file where you can set the name of the (sub)tabs in three languages. In a country like Belgium with three official languages (French, Dutch and German) multilingualism is always an issue. For the button to change the languages with I use a picture with the flags to depict which languages can be choosen: dutch and french in this case.


I always use the persistent multicolored option. I try to use only one color per graph, but different shades of this one color. You could call this a gradient between the different instances of the dimension, where the one with the largest value for the metric at hand gets the darkest color, and I go lighter where the value decreases. In some cases a conditional calculation of colors can give a beautiful result.


Using only barcharts makes the whole a bit too boring and also takes a bit more space, otherwise I would probably use only those. I try to never use piecharts where there are more than 4 different values in a dimension and I always show the values in the legend of a piechart - and on the bars of a barchart. I often use charts as the only way to make a selection in a certain dimension.


Truth - Good - Beauty

On a more philosophical note I would like to point out the close relationship in an important philosophical tradition between these three concepts. In this view you can argue that a more beautiful dashboard supports the truth of the information it is communicating and that this is leading to a better world - if I may stretch it a bit. I am thinking about writing some more about this approach to BI but I'm not quite getting my head around it for the moment. Questions, tips, remarks are welcome.