Diverging bars with integrated neutrals are commonly used in market research, in particular in product testing, to identify the quantifiable product features that need changing in order to increase product acceptance, and the respective direction of change.
Responses in the 3 categories are easily comparable between attributes, and no mental math is required for the +/- balance (there are some indices, not included here, that show exactly that).
My personal preference, when dealing with Likert scales, is to complement the presentation of the detailed responses with the so-called Dominant Opinion Index, a sort of Net Promoter Score weighted with proportion of non-neutral responses (mentioned in Sandu & de Jong, 1996) :
My personal preference, when dealing with Likert scales, is to complement the presentation of the detailed responses with the so-called Dominant Opinion Index, a sort of Net Promoter Score weighted with proportion of non-neutral responses (mentioned in Sandu & de Jong, 1996) :
DOI = (% positive – % negative) x (% positive + % negative) = (% positive – % negative) x (100% – % neutral)
The formula becomes slightly more complicated if the intensity of opinion (e.g., agree vs strongly agree) is taken into account and the percentages are weighted based on that.
The index ranges from -100 (strongly negative) to +100 (strongly positive), with 0 midpoint as neutral.
The first part (% positive – % negative) gives the direction of the group-level sentiment, while the multiplication gives the intensity : the less neutral responses, and the larger the gap between positive and negative responses, the higher the intensity. Large absolute values come from both a large +/- gap and a large proportion of NON-neutral responses. In contrast, small absolute values indicate either that positive and negative opinions even out, or that there is a majority of neutral responses.
DOI indicates the dominant direction and intensity of the sentiment, and allows easy comparisons between cases like these two, which generate the same DOI 😄 :
The first part (% positive – % negative) gives the direction of the group-level sentiment, while the multiplication gives the intensity : the less neutral responses, and the larger the gap between positive and negative responses, the higher the intensity. Large absolute values come from both a large +/- gap and a large proportion of NON-neutral responses. In contrast, small absolute values indicate either that positive and negative opinions even out, or that there is a majority of neutral responses.
DOI indicates the dominant direction and intensity of the sentiment, and allows easy comparisons between cases like these two, which generate the same DOI 😄 :
- 55% positive & 35% negative, vs
- 45% positive & 15% negative
In this particular example, neutral responses are "Just Right", positives "Too Much", and negatives "Too Little". The recommended direction of action on a particular attribute is in the opposite direction as the sign of the respective DOI; i.e., a positive DOI indicates the need to decrease the intensity of the attribute (↘), and the other way round.
It is important to note that, in this case, recommendation for action (signaled by the presence of an upward or downward arrow) is not determined based on DOI, but on a different index, one that takes into account the loss in overall evaluation when moving away from the "Just Right" category (not described in this post).
Perhaps diverging bars with extra neutrals (below) are better suited to illustrate the DOI, compared to those with integrated neutrals (above).
