A lot of energy (not to say blood, sweat and tears) has been has been spent for all the preceding steps to create an analytical pipeline, but last and of course not least the final steps delivers the result to the audience / the user of the pipeline.


Sometimes the final output of a tremendous effort is just a single number, let's say 0.84, or just a bar chart, or a more fancy Sankey chart.

No matter, if it's "just" a number that represents the probability of the occurrence of a certain event, or "just" a chart that compares different categories

of customers to support a decision about the future direction of a large enterprise, this step is very special, special in two ways:

  • It's literally the most obvious one
    often the audience is not interested, if the result e.g. the bar chart, was produced by using a small text file as data source, a spark cluster, or a large Azure Analysis Services database, as long as the result supports the decision making process and the result is created in a timely manner
  • Sometimes the audience does not understand, and of course does not need to understand the subtle differences between a "Multiclass neural network" or a "Multiclass decision jungle" algorithm that was used to create the above mentioned number (hopefully someone made the decision for one the algorithms wisely)

Both points can lead to one of the following extremes

  • A poorly chosen visual that is not able to communicate the findings in a proper way, creating the odd feeling inside the audience "if this is the result of our big data / data science / business intelligence initiative, than we better stick to our lovely spreadsheet"
  • Neglecting the preceding steps and focusing on just the data visualization part, because "color sells"

 Here I will focus on the three aspects of this step Data Visualization, Collaboration, and the Architecture of Information Delivery.