A Mixed-methods Approach:
Building behavioral Personas of PicCollage’s Premium Users


July - August, 2020


  • UX researcher

  • UX research intern (me)

  • Data scientist

  • Product manager

  • Product designer

My Role

  • Workshop: facilitated group discussions and collected feedback

  • Recruitment: designed screeners and selected participants

  • Interview: designed the interview guide and conducted interviews

  • Diary entries: prepared guidelines and followed up with participants

  • Analysis & synthesis: coded transcripts and generated insights 

  • Presentation: planned the mini-museum with designers by visualizing findings



PicCollage is a photo-editing app with more than 20 million active users worldwide. In December 2018, PicCollage launched its subscription program on the iOS platform. Since then, subscription fee from premium users (VIPs) has become the largest source of revenue for the company. Therefore, VIP retention and VIP conversion have become some of the most important business goals for the company.


Although PicCollage had created behavioral personas of our general users in the past, we hypothesized that VIPs’ attitudes and behaviors on PicCollage are different from that of general users’. However, what the team currently knows about VIPs are solely based on quick surveys and unvalidated assumptions generated from event analytics platforms. Therefore, it’s essential to gain a more holistic understanding of VIPs’ motivations, goals, and pain-points of using PicCollage to retain and acquire VIPs. 

Research Objective

  • Understand attitude: the motivation, benefit, pain-points, and barriers of using PicCollage for VIPs

  • Identify behavioral patterns of different personas to uncover needs: this is based on the assumption that there are groups of users who share similar behaviors due to certain needs, goals, or habits. 

Business Objective - What can behavior personas achieve?

  • Enable team members to build empathy for PicCollage’s VIPs to foster product ideas during quarterly planning.

  • Identify opportunity areas and guide product roadmap for retaining VIPs

    • Identify the needs of various personas to prioritize product features

    • Understand the reason why different personas show varying levels of engagement 


As we want to learn about behavioral patterns of VIPs and understand the reasons behind their behaviors, both qualitative and quantitative research need to be carried out.

  • Qualitative research: interview + diary study. (reasons for selecting these two are explained in sections below)

  • Quantitative research: clustering


  • Showcase behavioral personas and all of the research outcomes by holding a mini-museum for all members to participate.

  • Dedicated reports for individual functions. 


Guided product roadmap

  • Discovered the importance of improving an initially overlooked feature, Template's, overall experience to retain and obtain more engaged VIPs:
    Identified the most engaged behavior persona is Template Lovers. Uncovered motivations and barriers of using templates for all the personas.

  • Identified features that will affect the majority of VIPs' experiences and prioritized them for improvement.   

Fostered members' empathy for VIPs to facilitate discussions

  • ALL the members from the company voluntarily attended the mini-museum and the research findings facilitated discussions during the quarterly planning

Group 107.png
image 30.png
Group 108.png

Members were engaged in discussions and participated in planned activities when attending the mini-museum


Process Overview

This project is a first-time collaboration between UX researchers and data scientists; both parties conducted research parallelly and discussed findings regularly throughout the whole process.

First Phase: Workshop & Recruitment

Brainstorming Workshop


We ran a workshop with members from multiple functions (PM, product design, community management, visual design, and marketing) to understand:
(1) various members’ understanding and assumptions on PicCollage’s VIPs & ex-VIPs;
(2) what they want to learn from VIP to identify the most valuable questions for different functions.



I designed the screening survey, set the recruiting criteria and recruited users

  • Recruitment criteria: 

    • Located in the USA, age above 18

    • PicCollage’s VIPs or ex-VIPs

    • Submitted a screenshot showing >5 collages  

    • Using photo-editing apps > 1/week

  • Distribution channel: Surveys were distributed on Respondent (a recruiting platform), PicCollage’s iOS apps, and through PicCollage’s accounts on social media platforms. 

Second Phase: Workshop & Recruitment

Quantitative Research


  • Sample: Randomly sampled 2000 VIPs and analyzed their events within 90 days. 

  • Clustered collages based on style: clustered collages into 4 types based on it style, which is affected by the number of features used.

  • Clustered VIPs based on their distribution of collage styles, resulting in 4 personas.


  • Each VIP persona is rated by two metrics: engagement & loyalty

    1. Engagement - how often a VIP finishes a collage: an engaged VIP saves at least 1 collage per month for 3 continuous months within the 90-day data collection period.

    2. Loyalty - how long a VIP stays subscribed: a loyal VIP user has an annual subscription or 3 continuous monthly subscriptions within the 90-day data collection period.

  • Identified the personas that is most engaged and loyal - the Template Lovers (using template is one way of starting a collage in PicCollage) → But data alone doesn’t tell us WHY! 

Qualitative Research - Interview + Diary study

We selected these two as the research methods because we aimed to gain a deeper understanding of users' attitudes and remotely observe their behaviors in natural settings :

  1. Attitudes: learn about their motivations and goals of using and subscribing to photo-editing apps, what they value, their pain-points, and what they think is missing.

  2. Behavior: observe the end-to-end experience of making a creation, the features and contents they use to achieve their goals, and the barriers they experience.


  • The 12 participants (ages 25-65) are from PicCollage’s largest market, the United States. 9 of them are PicCollage’s current VIPs and 3 of them are ex-VIPs. 

First Interview

This interview consists of 2 main parts:

  1. To understand the participant’s goals of using photo-editing apps, perceived value, motivations of subscription, and reasons for continuing/canceling the subscription. 

  2. To explain the timeline and our expectations for their diary entries. 

Group 2.png

Some examples of the guidelines for participants

Diary Entries

To understand participants’ contexts of using photo-editing collages and observe their natural behaviors, we asked participants to submit at least 2 screen-recordings of their end-to-end creation process and answer 3 questions for each. I might ask follow-up questions after receiving their diary entries.

Second Interview

There are 2 main purposes of this interview: 

  1. Follow up with the observed behaviors in participants’ screen-recordings worth exploring. For example, why they decided to choose certain tools over others and why they chose some options over others in PicCollage

  2. Understand the triggers and barriers of using template 

  • Whether participants are aware of templates 

  • What’s their experience of using templates 

  • How do they decide whether to use template or not.  

Reason of asking: Quantitative research discovered that the VIP persona that uses relatively more template has the highest engagement. However, compared to other approaches of starting a collage, using template has the lowest completion rate. Therefore, we’d like to contextualize the conflicting data and understand the experiences of using templates.


I adopted a grounded-theory approach to analyze the data from interview and diary studies: 

  • During open coding, another user researcher and I each coded 2 transcripts and took memos along the process. Afterwards, we aligned on the categories that we found most relevant to the research questions
  • Following the guidelines of selective coding, we reviewed the emerging categories and analyzed relationships between them

  • Lastly, insights were generated by integrating the themes that emerged through selective coding. 


An overview of basic information of participants


Coding of transcripts


Mini-museum - Display Behavioral Personas

  • Motivation: we chose to hold a mini-museum to let all members cultivate empathy for PicCollage’s VIPs by immersing in a physical environment and interacting with artifacts.

  • Process: I collaborated with designers to visualize research findings and create interactive activities.

  • 4 Behavioral Personas: all of them are grounded on both qualitative and quantitative research findings.

"Take a peak at PicCollage's 4 VIP personas!"


  • The least engaged & loyal persona.

  • Collage style: all of their collages are simple collages, using only text, background, and layout.

  • Goal: put multiple photos together quickly. 

  • Pain-point: difficulty of finding some of their favorite backgrounds.


Template Lover

  • The largest percentage of VIPs, and is the most engaged & loyal persona.

  • Collage style: characterized by using more templates (than other personas) to create collages.

  • Goal: make appealing collages quickly with the help of suitable templates.

  • Pain-point: elements on templates are often not customizable, forcing them to start from scratch.



  • Collage style: Almost half of the collages are decorated by stickers and photo effects

  • Goal: decorate their collages with some personal touch and make sure the collages look aesthetically coherent.

  • Painpoint: find stickers that match the collage’s theme and style is tedious work 



  • Collage style: characterized by using more doodles and animations to create their collages

  • Goal: create collages without format restrictions and experiment with various features to make satisfying creations

  • Painpoint: difficulty of matching colors of various elements on the collage

Function-specific report

  • Visual Design function: report on the types of templates that VIPs want.

  • Product Management function: report on the potential of business use cases.


1. How might we enhance the largest number of VIPs' experience?

Research Finding


Group 105.png

“I have so many background packs that sometimes when I'm looking for specific background in a specific pack I have to scroll through ALL of them.”

  • Quantitative research showed that all the VIP personas create at least 50% of simple collages. That is, all of the VIPs frequently make collages that only consist of basic features, with text, background, or layouts.  

  • We discovered that VIPs have painpoints and unfulfilled needs with basic features from diary studies  Ex, they encounter difficulty finding backgrounds and have needs of more text editing effects.

Prioritize improving PicCollage’s basic features that all of the personas frequently use.

Research Finding


2. How might we retain PicCollage’s most engaged & loyal persona, and possibly turn other personas into it?

The most engaged and loyal persona is Template Lover and they account for the largest percentage of PicCollage’s VIPs. Although VIPs found templates valuable to their creation process (e.g. serve as inspiration and speeds up the process), there are two major barriers to using templates:

  1. Template inflexibility: both template lovers and other personas have the need to customize various elements (text, sticker, photo frames, etc.) on templates, and may give up if they aren't able to do so.

  2. Template discoverability issue: other personas who have the need of using templates may not even be aware that PicCollage provides this feature. 


Sometimes templates never catches VIPs' attention, or they don't know what they are for.


Upon entering, some VIPs cannot effectively see the templates they want; therefore, considering them unuseful.

Path Copy 2.png

“Like on this template, if I want to make the thank you' smaller and I can’t do it, I’d discard it because it’s not as customizable.”

Path 2.png

Enhance VIPs experience of using templates by allowing users to customize templates and improving template’s discoverability and accessibility throughout the user journey


  • The insights and behavioral personas were used in the planning and prioritization process during the quarterly planning, leading to multiple projects. 

    • Improve basic features: Projects aiming to enhance the basic features were prioritized in the following sprints. 

    • Enhance template's overall experience: PMs decided to improve the overall experience of using templates, starting from increasing its discoverability.​

      • Improve discoverability on the homepage​

      • Design easier ways to search for templates

      • Provide more entry points to access templates

      • Improve template flexibility


  • Evaluate incentives and consider creating a participant panel: I encountered difficulty recruiting qualified premium users for the diary study, which delayed the research process. Even though creating a participant panel requires high effort, it may decrease the difficulty of recruitment for future studies.

  • Continuously generate hypotheses to complement quantitative research: although I shared diary entries with the whole team twice a week, I should have facilitated team members to continuously generate new hypotheses based on new data. By doing so, qualitative data could better complement quantitative research by helping data scientists select metrics for clustering to speed up quantitative research’s iterative process.

  • Think about stakeholders’ needs from the very beginning to generate actionable reports: even though learnings from users are insightful for researchers, some of them are not necessarily actionable for other stakeholders. Therefore, it’s important to keep stakeholders’ needs in mind since data analysis to present findings in an actionable way to make an impact.