How APG Assigns Pictures to Create Data-Driven Personas

The following is a post from the APG Team’s summer 2020 intern, Jaad Mohammed.
——–

If you have ever wondered how photos or headshots are assigned to a persona profile by the AGP system? Well, you’ve come to the right place.

First, you need to understand where exactly APG collects data and how it processes this data to create the basic structure for a set of base personas.

APG links to and accesses data from online social media platforms (e.g., Facebook, Google Analytics, YouTube) via the Application Program Interface (API) of each analytics platform, given an account holder’s permission [1]

Via the APIs, APG collects the detailed interactions of users with each of the online content pieces on the corresponding platform. This data is available to only owners of a particular social media channel (e.g., YouTube channel) and is not available to the general public

Data provided by these API’s contain variables of gender, age, and country of the set of users, audience members, or customers, provided at an aggregated group level.

To understand these aggregated group levels better, APG must dis-aggregate it. To do this, we implement a matrix representing users’ interaction with the online content, such as videos.

This matrix is implemented by non-negative matrix factorization (NMF), a process which looks something like Figure 1.

Figure 1 Matrix decomposition using NMF. Matrix V is decomposed into W and H. g denotes demographic groups in the dataset, c denotes product units, p is the number of latent behaviors of demographic groups over product units, and ε is the error term.
Figure 1 Matrix decomposition using NMF. Matrix V is decomposed into W and H. g denotes demographic groups in the dataset, c denotes product units, p is the number of latent behaviors of demographic groups over product units, and ε is the error term.

The result of NMF becomes a set of base personas, which APG turns into a rich persona by adding personality [2], such as demographic information, name, and other information, …. including the photo!

To assign a photo to this base persona, APG stores thousands of commercial stock photos of models for different ethnicities, genders, and ages, to which the copyrights were purchased.

Here, the selection of different styles of figures to represent different professions, interests, etc. can strengthen the expressive power of the persona, so we selected varied photos for each demographic group and tag each photo with the appropriate meta-data.

Then, through age group, gender, ethnicity, etc. of a representative user segment, we can assign an appropriate photo to a persona.

There! That is how a data-driven persona gets a picture!

References

[1] Jung, Soon-Gyo, Joni Salminen, and Bernard J. Jansen. “Giving Faces to Data: Creating Data-Driven Personas from Personified Big Data.” In Proceedings of the 25th International Conference on Intelligent User Interfaces Companion, 132–33. Cagliari Italy: ACM, 2020.

[2] An, Jisun, Haewoon Kwak, and Bernard J. Jansen. “Personas for content creators via decomposed aggregate audience statistics.” In Proceedings of the 2017 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining 2017, pp. 632-635. 2017.

2 thoughts on “How APG Assigns Pictures to Create Data-Driven Personas”

  1. Pingback: How APG Assigns the Job of a Data-Driven Persona – The Persona Blog

  2. Pingback: How APG Assigns the Education Level of a Data-Driven Persona – The Persona Blog

Comments are closed.

Scroll to Top