What does order by Uniqueness or Affinity mean in Influencers & Brands tab?

When you click on the Influencers & brands tab on your Insights reports you will see the option to sort the accounts your audience is interested in by affinity, and also by uniqueness.

What are Affinity and Uniqueness?


Affinity is a proprietary metric that intents to help our customers to understand the interest that an audience has towards a person, a brand or an organisation. This shows the influencers that are most followed.


How is it calculated?

Audiense infers the affinity of the audience according to the interactions (tweets, posts, follows, likes, etc.) of audience members on Twitter. One of the most important inputs to calculate the affinity is the follower graph (follows & followings), because it allows us to know more about how people are interconnected and therefore help us better understand their affinities.

When you are ordering by affinity, you will be able to see the highest interests in that audience (most followed). For example, people like Elon Musk will be usual suspects when you are ordering by this; 42.41% of this audience has an affinity for Elon Musk, meanwhile 4.57% of the global audience has an affinity for this account as well (baseline selected as global - see image below).

In order to find more relevant/popular accounts for your audience when compared to a baseline, you may want to order them by uniqueness.


Why doesn't the Affinity score add up to 100% in some cases?


Please note:  The total affinity percentages for a single account may not sum up to 100% due to the presence of protected or suspended accounts. These accounts are not included in the analysis, leading to a discrepancy in the percentage results. On average, private or suspended accounts constitute between 5% and 15% of the total accounts in the sample, which can affect the overall percentage distribution.

Furthermore, when the affinity calculation involves followers from multiple handles, and these handles boast differing follower counts, the percentages are often skewed towards the larger account. This is particularly noticeable when some handles have millions of followers or surpass our 250k sample size.

In summary, it's important to note that the affinity percentages will not always total 100%. This is primarily attributed to the sample size limitation of 250k when amalgamating followers from different handles, as mentioned, along with the exclusion of protected or suspended accounts, regardless if the audience definition includes followers of one account or multiple.


Screenshot 2022-09-01 at 16.03.51


For uniqueness, we use a proprietary index in order to determine how relevant and unique a person or a brand is for the members of the audience. This index takes into account how "mainstream" the accounts are for the public and penalizes accounts that are influential for the whole population versus those that are especially relevant for the audience we are analyzing. This index allows us to highlight brands or people that are unique for a concrete target audience or segment.


The index is scored with values ranging from 0 to 100, being 100 the maximum unique affinity.


For example, when analysing a segment of vegans, when we compare to the full audience as the baseline - example using followers of @PeanutButterCo (the demo report in your account!), we see that the audience has a great affinity with vegan snacks and brands, the baseline has a higher affinity with So Delicious than a large account such as VegNews or the Vegan Society. Our index will penalize the results that are mainstream, helping us surface the real influencers for the target audience (in this case, VegNews and Daiya Foods at the top)."


How does the Uniqueness score/formula work?

We explain uniqueness as relevance, or better said, popularity, to make it easier to understand:

The specific indexing we perform takes into account the popularity of the influencer in the audience and the popularity of the influencer in the baseline audience we are comparing against.



= 100 * (200 + SEGMENT_AFFINITY - 2*B) / 300 = (200 + SEGMENT_AFFINITY - 2*B) / 3

B = BASELINE_AFFINITY: % of users from the baseline audience that has an affinity with the influencer




The indexing is designed to pop up the more unique accounts, and this sometimes translates in a gap in the actual value of the index for accounts that surpass a certain threshold (they’re either pushed down or up by these thresholds). With this we aim to promote real influencers. Someone followed by only 4 people within a segment is not really an influencer. We also take into account the Baseline being compared (i.e. if you compare a segment with the full audience, the uniqueness scores may be lower as there is similarity between the segment and the full audience).


The reasoning behind this is that raw affinity is not the same as uniqueness, as there are influencers that tend to appear in every audience due to their sheer popularity”.


In this approach, it is especially rewarded if an influencer has a high affinity with the segment and very low affinity in the baseline. In other words, an influencer who has a strong presence in the segment, but also in the baseline, will be ranked lower than another influencer who may have less presence in the segment, but has almost no presence in the baseline.


The objective is to discover influencers that are really niche within the segment and to discard generalist influencers, who have a presence in any segment, such as Obama or Lady Gaga.


In addition, in this formula we square the affinity percentage of the segment, i.e. we square the dividend. Once again, this serves to favour the most influential influencers and to penalise the 'fake' influencers a little more.


On the Excel export, once you download the data, you will have the calculation based on the audience where the influencers were exported from at the time  - i.e. a segment - in comparison to the baseline selected at the time when exporting - i.e. Baseline could be the full audience, another audience from another report, or a country (which would be a national average = Twitter users in that country)


It takes into account how popular that influencer is to the audience when compared to how popular it is to the baseline. To get the most relevant/popular influencers in order = unique, to the audience. Whereas affinity looks at most followed within an audience, uniqueness looks at the popularity/relevance due to engagement.


Please note:

Within the Sort by (view) menu within the Influencers and Brands tab, you will see "Very Low/Low/High/Very High Uniqueness, which are all relative metrics that are calculated against the benchmark/baseline.

Each of them provide different calculations, meaning when you select very high uniqueness, it's a different calculation than when you select uniqueness, and that’s why you will see different numbers.


When you filter by Uniqueness, then it's the uniqueness score you are seeing (Not the Very Low/Low/High/Very High) which can be confusing. These are useful depending on your use case.

We are aware there is confusion around this and when they should be used, so we are planning to solve it."


Our advice is to use Affinity when you want to know the influencers that are followed the most (the audience has the most affinity with them) or Uniqueness when you want to know who are the most relevant or popular influencers to an audience, when comparing them to the baseline you have selected (i.e., influencers that are popular for a Segment when comparing to the full audience in the baseline).


Related articles:

  1. FAQ: How does Audiense categorize influencers by Affinity, and how do they fall within the categories?
  2. Influencers & Brands Tab of Insights Reports