Overview
Follow this article to learn how Audiense Insights classifies audience data.
Throughout Audiense Insights, you'll encounter information that is either explicit or inferred (modeled). Understanding the difference between these two data types is important when interpreting your reports, as each represents a different method of collecting or generating audience insights.
Some information comes directly from publicly available X (formerly Twitter) profile data and user activity, while other insights are generated using machine learning models and analytical techniques to provide additional context about an audience.
This article explains these two types of data and provides a reference for the enrichment fields available within Audiense Insights.
Glossary of terms:
- Explicit - clearly stated, so there is no room for confusion or questions.
- Implicit – implied or suggested, but not clearly stated.
- Inference – a conclusion made based on both information/evidence and reasoning.
- Extrapolation - In broad terms, extrapolation can be likened to deduction or inference, which involves drawing conclusions about unknown factors from known information.
Enrichment Table: Explicit vs Inferred
|
Enrichment name (All based on Twitter Data) |
Description |
Explicit/Inferred |
|
Activity |
balance of activity of the account in terms of how much original content (Tweets), conversation (mentions) and sharing (Retweets) they have published |
Explicit |
|
Automation |
balance of activity of the account in terms of how many tweets are published manually or automatically |
Explicit |
|
Client App |
most used Twitter client app |
Explicit |
|
Days |
what days of the week the user is most active |
Explicit |
|
Devices |
most used devices |
Explicit |
|
Engagement |
amplification ratio (Retweets received/total Tweets posted) and applause (favorites received/total Tweets posted) |
Explicit |
|
Hours |
at what time the user is most active |
Explicit |
|
Languages |
most used languages |
Explicit |
|
Operating systems |
most used operating systems |
Explicit |
|
Platforms |
most used platforms |
Explicit |
|
admin3_code |
other administration levels, sometimes the postcode |
Inferred/Modeled |
|
admin4_code |
other administration levels |
Inferred/Modeled |
|
city |
Inferred/Modeled |
|
|
country |
Country name |
Inferred/Modeled |
|
country_code |
ISO-3166 2-letter country code, 2 characters |
Inferred/Modeled |
|
precision |
precision level, i.e. country, state, province or city |
Inferred/Modeled |
|
province |
Province name. 2nd level in the country, administrative level |
Inferred/Modeled |
|
province_code |
Province code. 2nd level in the country, administrative level |
Inferred/Modeled |
|
state |
State name. 1st level in the country administrative level |
Inferred/Modeled |
|
state_code |
State code. 1st level in the country administrative level |
Inferred/Modeled |
|
Interests |
IAB interests taxonomy. 5 levels of categorization |
Inferred/Modeled |
|
Gender |
Inferred/Modeled |
|
|
Entity type |
Inferred/Modeled |
|
|
Affinities Influencer/Brands |
Following % |
Explicit |
|
User ID |
Twitter user ID |
Explicit |
|
Twitter Handle |
to be translated and consistency check |
Explicit |
|
Name |
Explicit |
|
|
Location |
to be updated in-app / twitter bio location |
Explicit |
|
User Bio |
Explicit |
|
|
Language |
Explicit |
|
|
Time zone |
Explicit |
|
|
Has avatar |
Explicit |
|
|
URL |
Explicit |
|
|
Has URL |
Explicit |
|
|
Is protected |
Explicit |
|
|
Is verified |
Explicit |
|
|
Time since last tweet |
Explicit |
|
|
Followers |
Explicit |
|
|
Following |
Explicit |
|
|
Tweets count |
Explicit |
|
|
Listed count |
Explicit |
|
|
Followers ratio |
Explicit |
|
|
Tweets per day |
Explicit |
|
|
Affinity |
Explicit |
|
|
Age |
IBM Watson age recognition + Internal Heuristics |
Inferred/Modeled |
|
Personality Insights (Below) |
IBM Watson personality insights |
Inferred/Modeled |
|
Consumer Preferences (Below) |
IBM Watson personality insights |
Inferred/Modeled |