We have developed a sophisticated Content Discovery Engine that is able to ascertain the Demand that is expressed on social platforms.

However, you may be wondering what happens in the following scenario: "What if people aren’t using the official hashtags or the official pages on social media? Are their comments and posts not included in demand measurement?"

The short answer is that comments and posts are included for this scenario. To explore this topic further, let us now outline some of the sophisticated processes that are employed.

Social media data ingestion

Our verified and stable channel properties are used as a baseline towards our models to ingest the initial pull of Demand Expressions for a show. They are also used for identification of surrounding topics, such as hashtags, pages and other topics of interest, which is then used throughout our platform to enhance our listening across all the demand expression platforms we monitor against. This also helps us widen the net and allows for us to monitor the "silent" or "out-of-band" fandom conversations which are happening outside of the standard channels.

Outside of the "fan-declared" properties of interest, we also employ a mixture of other classifications for the raw data we capture to ensure a verified funnel of non-standard declared interactions. For instance, we continuously update our Content Genome which provides us with an incredible amount of metadata about each show, and this is utilized in our process for show identification.

We use a variety of features outside official page handles and hashtags to pick out conversations towards a specific title, such as character names, (while comparing with the real actors name), title-specific themes – both from official plot material and fan-generated content – ranging from key plot themes to episode-specific themes, combined with context-setting metadata to filter wider conversations about the show mood, all the way to producer, distributor and platform carrier. We leverage our genome to feed all of these tags towards the identification process at the raw fire hose level feeds we capture.

We also apply additional levels of classification by ensuring the data captured can be mapped against topics of interest, ensuring that the interactions captured have a high confidence level of belonging to the TV, Movie, or Talent space, instead of an unrelated topic: We start by "casting the net as wide as possible", and subsequently verify against all known metadata about a specific title.


We do not weigh social media comments as highly as other much stronger expressions of demand such as sustained video consumption (streaming video). Contrary to other measurement systems, which attempt to analyze sentiment (to varying degrees of severity) across different languages, and then try to quantify this into a metric, Parrot Analytics instead utilizes a more robust way of avoiding error in demand assessment by ingesting and ranking a multitude of demand input sources.

Nevertheless, we do account for both positive and negative sentiment in the source data from social platforms. Taking a non-discriminatory approach to sentiment allows us to capture demand for content where audiences are expressing negative emotions towards plot lines and characters: Despite the very negative sentiment expressed (-100), these audiences are exceptionally engaged with the show (+100).

Considering sentiment during collecting enables us to create the most robust and sustainable way of accurately representing both the "volume" and "context" for the demand that is expressed. We do this in order to:

  • Weight every interaction: Naturally, consumption and creative participation trumps sentiment expression, which enables Demand Expressions to stand on their own.

  • Provide the necessary data for our upcoming Sentiment module with which our customers will be able to contextualize the Demand that is being expressed by title, in every country, each day. The module will provide insight beyond the current level of Demand that is expressed, and will also include what audiences are actually thinking.

  • Ultimately evaluate the possibility of context-aware sentiment and emotion recognition, beyond positive and negative sentiment, providing in-depth context for our Demand Expressions metric.

To our knowledge, there exists no system currently which captures every single conversation about a title on the planet, inclusive sentiment, and the affinity and demand of every silent fan. We believe the processes outlined above come closest to defining the industry standard.

Did this answer your question?