Honesty, the first principle of good analytics

 

ph-design-evolution-eichler-rams
Photo from Braun.com

Meet Dieter Rams – Industrial Designer

Dieter Rams is a German industrial designer most closely associated with the minimalist designs of the consumer brand Braun. Dieter was head of design at Braun for over 30 years, where he became famous for creating an austere aesthetic while focusing on user-friendliness. His philosophy is summed up in his saying, “Weniger, aber besser.” which translates into “Less, but better.” He has won many awards through the years including the World Design Medal and the Ikea Prize.

Dieter's impact reaches beyond his retirement as he is now impacting design in the 21st century, with a company widely considered a leader in technology design, Apple, acknowledging a debt to Dieter as inspiration for many of their design decisions. The Head of Design at Apple wrote, “Rams's work is beyond improvement... Rams's ability to bring form to a product so that it clearly, concisely and immediately communicates its meaning is remarkable.”

Braun-Apple-Link
From Braun, "90 years of history"

 

 

Ten principles of good design

As a prolific designer, Dieter formulated ten principles of good design. In this series of articles, I will adapt several of these principles for guidance in creating good analytics. I have selected the sixth principle of good design as the one I consider most important for good analytics.

 

The sixth principle of good design

6. Is honest - It does not make a product appear more innovative, powerful or valuable than it really is. It does not attempt to manipulate the consumer with promises that cannot be kept.

 

Freakalytics_princples_good_analytics_1_003_smallThe first principle of good analytics

1. Is honest - a good analysis seeks to share the reality of the situation. A key component of honesty is sharing the level of uncertainty about the data, the questions, the reliability of past analyses, the limitations of the methods employed and the level of uncertainty surrounding any prediction of future events.

This principle does not require that you lead your presentation with the details listed above, which could leave people feeling helpless, but you should thoughtfully incorporate the limitations as relevant in the appropriate places in your presentations or dashboards. It is incumbent upon the lead analyst to understand the audience, their expectations and their knowledge about these topics. Considering these areas, your presentation should either gently or forcefully remind them of the relevant limitations as needed.

An exception is made for critical topics that should be addressed in the beginning of your presentation. Examples of critical topics includes significant changes in methodology since the last review, changes in assumptions about future business conditions or any other material change in the data or inputs to future forecasts.

 

An example of clear deception & correction with the principle

Seeking to tell the people what they want to hear

Freakalytics_princples_good_analytics_2_001
Misleading title, hidden scale, selecting an aspect ratio to amplify growth and deleting the origin ($0) all violate the first principle of good analytics, honesty.

 

 

Making reality clear

Freakalytics_princples_good_analytics_2_002
Correcting the title, adding a clear scale, including the origin ($0), adding growth rate labels and a reasonable aspect ratio honestly inform the audience.

 

A second illustration, conveying the right level of certainty & correcting models

By displaying a single line for the future values, too much certainty is conveyed

Freakalytics_princples_good_analytics_1_001
Less-experienced people often attribute high certainty to a graph like this, often created by a well-intentioned colleague.

 

 

Much better, you can clearly see that uncertainty increases the further out you plan

Freakalytics_princples_good_analytics_1_002
Being honest about the limitations of forecasting methodologies can help an audience understand the level of "risk" about over-committing resources in advance,

 

 

Even better, by understanding the dynamics of the business
and adjusting the forecast model, a more conservative forecast is created

Freakalytics_princples_good_analytics_1_003
Alas, the analyst learned that the forecasting tool automatically selected an aggressively assumption, imputing a multiplicative growth trend. Correcting this leads to a more honest forecast for a company committed to conservative planning and decision-making.
[Post-Bottom-1] [Mail-Chimp-Signup] [Post-Bottom-2]