Before 2018, our efforts to calculate market-specific engagement were primarily focused on building market-specific panels of influencers that were based in each of the markets we serviced. For example, if an influencer lived in France and created content in French, we assumed that their followers were primarily centralized in France.
However, we also knew certain influencers were unique, mainly those that:
lived and worked in multiple countries (Victoria Beckham),
were popular around the world (Cristiano Ronaldo), or
lived in one country but claimed most of their followers from another (Nikkie de Jager, aka NikkieTutorials).
To account for these cases, our Data Science and Operations teams developed a more accurate and nuanced method to better determine where an influencer actually has influence, using a combination of home location and Instagram audience data from a third party vendor.
Our Approach: Modeling by Audience Location
We started by determining a home location for all of the Instagram profiles in our system, looking at the cities or countries noted in their description as well as their geo-tagged posts. We also used basic web searching and our partner's designated home location as a cross-reference. Using home location served as both an initial and final check, allowing our models to be accurate while preventing any anomalies in the audience data from affecting our estimates.
Once we had a set of influencers we believed to be based in each market, we reviewed the Instagram audience data for that set. Each influencer had a percentage breakdown of their followers and likes from each market, similar to the sample data below:
Using this data, we created a model that determined what the “average influencer” looked like in each market, based on their Follower and Like locations. We could then evaluate not only the influencers that we knew lived in certain markets, but also those with unknown or multiple home locations.
Applying the Model
Once completed, the model assigned every profile in our system a score for each market, representing how confident we were that the influencer has influence in that market. If their score qualified them for a market, they were added to the pool of profiles available for that market’s panel. Most profiles only qualified for one market, but there were a set of more "global" influencers that qualified for two markets.
Once each market’s pool was identified, we further assessed them on a set of six separate metrics we use to determine a profile’s quality. Find out more about the metrics and methodology here.
While we apply this methodology as is for most markets, we allow for certain location edge cases on the US panel, given the unique diversity of the US market and population.
US-based influencer who have 100K+ US followers on either Instagram or Youtube are automatically included in the US evaluation pool, even if the US is not their main market. They are then subject to the same metrics assessment as described above.
What this means for you
In addition to increased confidence in each influencer’s location and relative influence, this technique also allows us to account for cultural and size differences across markets. Now, we can ensure that the rules applied to a smaller, more insular market like Japan are not the same as a larger, English-speaking market like the UK.