Visualising data to improve communication and understand business

October 3rd, 2019, Published in Articles: EE Publishers, Articles: PositionIT, Featured: EE Publishers, Featured: PositionIT

This article will explore some workflows for answering pertinent business questions using data and best practice visualisation principles, focusing on a last-mile, home-delivery logistics business case. While Tableau software is used, the general principles can apply to any software used.

As mapping technologies have advanced and high-powered functionality is incorporated more and more into end-user software solutions, the results of geospatial data analysis can greatly benefit its dependent industries. The tools and solutions are becoming increasingly layman-oriented, such that even the non-tech-savvy operations or executive manager can glean actionable insight from well-prepared analytics.

The process

For a complete workflow, there must be means to capture data accurately, process that data, house it in a central repository, connect the data to a visualisation-layer software and present the data in reports that effectively speak to the end-user in an interactive and engaging format. The process, however, is only as effective as the accuracy and timeliness of data that is being captured.

Data capture and warehousing

The era of clipboards and manual input into spreadsheets is ending. The leveraging of robust software solutions that emphasise correct data capturing from those most involved in the process will ensure this stage is performed optimally. In the logistics space, software such as Trackmatic (http://trackmatic.co.za) allows the driver himself to perform the crucial task in real-time using a mobile device.

Fig. 1: Saving route execution data from the device into a database.

When performing an action on a route or trip, data gets pushed from the device into a database. Fig. 1 shows an example of route execution. Data captured in such a manner for each route will accumulate and eventually the repository will be a strong foundation for effective analytics

Data visualisation

The power of data visualisation lies not necessarily in the technical expertise of the specific analysis tool, but more in the communication of the data. Visualisation aids in storytelling, and the different types of visuals contribute to answering specific questions in the most effective way.

Fig. 2: A list of the different visualisation types that can aid storytelling.

Fig. 2 shows the different visualisation types that can aid in storytelling, depending on whether the end-user wants to focus on comparison, distribution, relationships or composition of data. Geospatial data would be more an umbrella term to define a type of data, but can extract insight from any of the aforementioned categories.

Most industry-standard visualisation software can connect to many database types: Excel spreadsheets, cloud software such as Google Sheets or Dropbox and databases such as SQL Server, MySQL and the like. As a Tableau Certified Professional, I will be look at an example using Tableau, but most of the principles will be readily applicable to other software. The software’s functionality includes powerful spatial analysis as built-in functions, which can be leveraged to create formulae for the initial analysis (see Fig. 3).

Fig. 3: Formulae for the initial analysis.

Using Tableau’s drag-and-drop functionality, a simple visualisation can be built as shown in Fig. 4. A powerful dashboard will include many different, complementary views, but a map view is crucial for ease of understanding. This wagon-wheel graph, coloured by selling store, is effective for instantly identifying the furthest-reaching customers, how much each store extends itself, and when customers are not serviced by the closest stores. For example, the far-right dark blue customer could have been serviced by the light blue store. The incremental saving in kilometres and delivery costs could be substantial if patterns like these are eliminated by simple empirical insight, or exceptions could then be correctly reported, such as a store being out of stock of the requested item, so inventory management could be more tightly observed.

Fig. 4: The wagon-wheel graph, coloured by selling store, is a spatial visualisation which is effective for identifying customer reach.

Best practice is not to bombard a visualisation with many different colours, and when the store count grows, this could be a problem. Also, it is difficult to track just how far away customers are from the store. I rarely colour graphs by a dimension, such as store, customer, or driver, but rather by a KPI. In this context, a trackable KPI would be to see how many customers are within a radial distance of a store and apply a score to each store to see which store is most extended. A simple calculation for radial distance is shown in Fig. 5a, on which I then define a True/False (Boolean) check as shown in Fig. 5b.

Fig. 5: a) A calculation for radial distance, and b) a True/False (Boolean) check.

I can apply this formula to colour, in which case my visualisation tells a different story (see Fig. 6). Now I can see the majority is in the clear. I have also achieved consistent colouring.

Fig. 6: The visualisation tells a different story when applying the formula to colour.

Another way to determine an in-bound KPI is to use a spatial file that maps out an area for a store’s reach. A kml file is quite effective, and there are tools to build these geofences online, such as the Drawing Tool for Tableau (http://drawingtool.powertoolsfortableau.com). You can then use the point data and create a kml file, which is like an xml with a geospatial structure. Once you have this, you can connect it to existing data using the <name> element  as shown in Fig. 7. And you can build an elegant solution (Fig. 8).

Fig. 7: Connecting the kml file to the existing data using the element.

 

Fig. 8: The result – an elegant solution.

To complement the dashboard, you can add some KPIs. These are simple text indicators because a good blend of text and visuals is important for effective dashboard building (Fig. 9). The Data-Ink Ratio (https://infovis-wiki.net/wiki/Data-Ink_Ratio)  can help guide a dashboard developer on how to strike the correct balance.

Fig. 9: A good blend of text and visuals is important for effective dashboard building.

I like to have a percentage KPI on a colour scale from red to green, so the end-user can detect the overall status quickly. To incorporate a time element in the visualisation, I create what I coin a rollercoaster graph. It is essentially a trend that combines lines and bars, as shown in Fig. 10.

Fig. 10: A rollercoaster graph combines lines and bars to show trends which incorporate a time element.

From here I can see that the ratio is consistently high but does seem to be declining somewhat. This should be enough to alert managers to reverse the trend. Usually, I overlay this graph on the text KPI, so the context is not lost (Fig. 11).

Fig. 11: Overlaying the rollercoaster graph on the text KPI retains the context.

To compare the stores, the most conventional type of visualisation is the bar graph, because users can acutely distinguish between different lengths as opposed to angles (pie charts), size (bubble charts) or colour (highlight tables). This allows for faster, and importantly, more accurate insight. Now, let’s create a simple bar graph with two measures: average distance and in-bounds ratio (Fig. 12).

Fig. 12: A bar graph with two measures: average distance and in-bounds ratio.

Bringing it all together

Tableau allows highlight functionality, so when I hover over a store, it will highlight on the store map (Fig. 13). I can improve this dashboard by adding more interactivity, such as a parameter to choose metrics to compare on a scatter plot, or filter panes to highlight specific customers or distance bands. Best practice is not to have more than four or five graphs on a dashboard, or the context and story might be lost in the details.

Fig. 13: The result – an easy-to-understand, accurate representation of the business.

Carefully consideration and a combination of basic and complex graphs, visualisation principle and best-practice techniques can tell a story with data in a compelling way.

This is a basic workflow for a certain business case, from data capture to intelligent reporting, but analytics and visualisation are not industry specific. These are universal skills built through practice, experience and trial and error. There are no “correct” visualisation types. They are tools to aid in effective storytelling, to convince managers and executives to pursue certain business decisions, relying on empirical data and historical trends. They open the channels of inter-company communication and allow everyone to see and understand the business. A data-literacy culture can allow for collaboration, better decision making, and confidence to tackle the data-rich world effectively.

Contact Sean Hurwitz, Trackmatic, seanh@trackmatic.co.za

Related Articles

  • Medupi magic fails Eskom and South Africa – the stuff of nightmares
  • South Africa plans for lower electricity demand, more renewable energy
  • New plug and connector range launched
  • How Coriolis measurement helps keep the Scotch in whisky
  • Automation sets in as geomatics solutions mature