Harmoni Fair Share analysis is a technique for evaluating the performance of brands against a set of attributes.
In this article
- What is Fair Share
- Create a Fair Share analysis
- Standard Image vs. Fair Share analysis
1. What is Fair Share
The fair share is an expected value for a brand on a particular attribute. It considers the relative size of both (the brand and the attribute). By comparing the actual value to the fair share value, one can see whether a brand is performing above or below expectations.
The analysis calculates both Fair Share Percentages and Fair Share Differences.
Mathematically the Fair Share is calculated by applying a logit transformation to the brand/attribute proportion, averaging over brands and attributes (rows and columns), and re-applying a sigmoid transformation (inverse-logit) to restore the values to a proportion (or percentage).
In Harmoni, Fair Share has no restrictions on the number of time periods or even the requirement for tracking data. The algorithm also handles 'missing' values. For Fair Share values, significant difference testing is not applicable and is therefore not available in Harmoni.
Fair Share doesn’t do any statistical testing so there’s no threshold or significance level above which the differences become meaningful. However, the usual guidelines on sample sizes and percentage levels apply as in standard tables. With small samples, the differences need to be larger to be meaningful, and also differences between percentages around 50% need to be larger than differences between percentages near 0% or 100%.
Usually, Fair Share is used to compare brands and attributes and it is the relative difference scores between brands and attributes that are important rather than the absolute differences. So brands with a higher difference score can be considered to be performing better on a particular attribute as compared to other brands with a lower score.
Fair share on average
Fair share can also be calculated on attribute averages. A composite attribute can be created by averaging over a set of common or associated attributes for a brand. Because an average of a proportion (or percentage) is also a proportion (or percentage) the fair share algorithm can also be applied to these average measures.
Note fair share is only relevant to an average of proportions and not to averages in general. This is because the logit transformation can only be applied to probability values between 0 and 1.
2. Create a Fair Share analysis
After creating an analysis with two dimensions (an across and a down), under the modify menu you can find the options to show fair share percentages and fair share differences.
a) Fair Share percentages
Fair share percentages are calculated by applying a logit transformation algorithm. The algorithm considers the number of responses in each cell and adjusts the proportions to reflect the size of the contribution within a column or row.
b) Fair Share differences
Fair share differences are the difference between an item's share of an attribute (original share percentage) and the corresponding fair share percentage i.e. people rating that item as having that attribute (gave an answer).
Fair share differences reveal the Fair Share patterns.
Brands (or attributes) that are over-performing are indicated by a difference value greater than 0 and the under-performing ones are indicated by a difference value below 0. In other words, if a particular brand/attribute has a positive fair share difference then the brand is performing better on that attribute than its fair share. Similarly, a negative fair share difference indicates a brand/attribute that is performing below its fair share.
Generally, you’d want to see values above 0 (as this shows that your brand has clear brand imagery), as long as the attributes are fitting with your brand image.
c) Visualise Fair Share
Graphing the differences table helps the under and over differences become more obvious. A bar graph is most useful for this.
c) Fair Share analysis example
Let's look at an example. In the chart below, Tom Tom shows higher levels of absolute association for "For someone like me" and "Unhealthy".
However, if we consider the relative size of the brands, it turns out that "For someone like me" is better characterized by Just Natural, Deli Good and Cream Rich while "Unhealthy" is something that distinguishes Jo & Ben and Moo Moo.
3. Standard Image vs. Fair Share analysis
Looking at the absolute levels of image endorsements provides an indication of:
- Who is class-leading;
- Which attribute has the largest impact; generally highest endorsements overall.
Fair share analysis
We can profile the data to show the relative strengths and weaknesses of each brand:
- Relative to the average of all statements for that brand;
- And also taking into account how other brands score on a particular dimension.
Brands (or attributes) that are over-performing are indicated by a difference value greater than 0 and the under-performing ones are indicated by a difference value below 0.
- Positive = higher score versus other brands and higher compared to endorsements for other statements;
- Negative = lower score versus other brands and lower compared to endorsements for other statements.
Where to from here?
Learn more about Smart Analysis.