Is it ever right to use the “smoothed lines” option?


About 18 months ago, a large business in the FTSE 250 asked us to give constructive feedback on an executive dashboard that their IT project staff had created using tools from the Cognos suite.

One of the graphs included in their dashboard, a trend of historical and future revenues by division, looked like this.

Smoothed line chart


This is a screenshot of part of the actual dashboard, but I have removed the descriptions of the data to protect the anonymity of the organisation involved.

This graph suffers from so many problems that it is not really possible to describe all of them in one blog post.  The particular issue I want to focus on is the use of smoothed lines, which seems to be an ever increasing trend in the presentation of data.

Smoothed lines are favoured by some designers of MI dashboards because they make volatile data series easier on the eye by softening jagged peaks and troughs.  Unfortunately, they can substantially detract from the integrity of the underlying data, as is the case in this graph.

Let’s focus particularly on the data series in our exhibit that contains the largest swings. The dashboard also included a table of numbers, and these were as follows for the final 12 months shown in the chart, which is the period that exhibits the highest volatility:


nos chart


The values that hitherto were monthly up to December 2012 change to quarterly values for 2013 and are stored against the final month of each quarter, with the first two months of each quarter showing revenues of zero.  Graphing quarterly data in monthly time slots in this way is clearly not at all helpful, but let’s focus on the effect that the smoothed lines have had on the integrity of the data.

A closer look at the scaling of the chart shows that zero revenue is denoted by a horizontal orange line.  The smoothing algorithm deployed in this chart has caused negative revenues to be plotted between January – February, April – May and August – September.  This is clearly absurd, and is not reflected in the data itself.

My strong advice is never to use smoothed lines in displaying management information because of the risk of compromising the integrity of the underlying numbers.  If you need to smooth the data, do so using a meaningful transformation of the numbers, such as a moving annual average.

There is an excellent blog post on this subject at


  1. Gerald Davies

    You need to know If a data series is volatile – it may tell you something about the characteristic of the business. Best avoid smoothing.


Leave a Reply