Engaging vs Effective Affinities

Under the "Audience" tab in your Campaign Reports, you'll see two charts labeled "Most Engaging Affinities" and "Most Effective Affinities". How are these calculated and what are the differences between the two?

Every Twitter and Instagram account we encounter is analyzed and given a list of affinities -- these are topics that the user talks about and engages with on a regular basis.

When you run a campaign, your creators submit content to Instagram and Twitter and their audiences engage with the content. Because we analyze creator audiences as well as the creators themselves, we know the affinities of most people who engage with your content.

Most Engaged, think Big Picture: % of engagements by users with a particular affinity

The first chart, "Most Engaging Affinities", tells you the percent of the engaged audience that have affinity with a given topic. In the example above, close to 14% of all engagements came from users who are interested in "Style and Fashion", and about 10% of people who engaged with the content are interested in "Upmarket Fashion".

This data is interesting, but not very actionable. Of course people who follow fashion bloggers on Instagram are interested in style and fashion! Additionally, style and fashion is a very broad topic and many people are interested in it. The "Most Engaging Affinities" chart is therefore affected by the broadness of the categories and the sizes of those audiences; if 15% of an audience is interested in style and fashion, then we would expect about 15% of engagements to come from people interested in style and fashion.

In order to make this data actionable, we run a separate analysis called "Most Effective Affinities".

Most Effective, think Vertical Engagement: % chance a user with a particular affinity engaged with the content

These topics are weighted by their relative audience size, so that chart gives you a better picture of which audiences you have standout performance in.

Rather than asking the question "what percent of my engaged audience is interested in a topic", the "Most Effective Affinities" chart asks the question: "if I could reach a user interested in a topic, what's the chance that they'll engage with my content?".

In the example above, "Nordstrom" appears as an affinity with 8% effectiveness. Put another way, 8% of users interested in Nordstrom would engage with your content. Nordstrom doesn't appear at all under the "Most Engaging Affinities" chart because only a small total percentage of engagements came from audiences interested in Nordstom (Nordstrom is a narrow topic followed by far fewer people than "Style and Fashion"), but the relative percent of the engaged audience interested in Nordstrom is high.

To put numbers to the example: imagine you got 100 total engagements on your content after being seen by 5000 people (a 2% total engagement percent). The "Most Engaging Affinities" chart shows you where those engagements came from: 14 of them are from users interested in Style and Fashion. Let's imagine that out of 5000 viewers, 1000 of them are interested in Style and Fashion. Even though 14% of all engagements came from that audience, only 1.4% of the cohort actually engaged with the content. Even though Style and Fashion is the most popular topic for your audience, you captured a less-than-average amount of that cohort's attention.

On the other hand, let's say you only received 5 engagements from users interested in Nordstrom -- that's not enough to show up on the Most Engaging Affinities chart with only a 5% share. But say that out of the total 5000 people who saw your content only 60 are interested in Nordstrom. You've therefore captured 5 out of the 60 people interested in Nordstrom, or 8% (way above the average engagement percent for this campaign). This makes Nordstrom a much more effective affinity than Style and Fashion, since you've managed to capture a higher share of that cohort even though the cohort itself is small.

The "Most Effective Affinities" chart is where you should take action: you do very well with those affinities, and if you start targeting them you'll see your campaign ROI increase as you focus on the most resonant audiences for your content.

This type of analysis is called "Bayesian analysis" and is one of the machine learning tools we use to generate interesting, actionable insights on creators, content, and audiences.