Company
Boosta
My responsibilities
Stakeholder Interview
Competitive Analysis
In-depth interviews
Prototyping
User Testing
Industries
HR Tools
Date
December 2022
About Coverler
Coverler is a web platform that writes a unique job application cover letter based on a person's short professional bio and job requirements. At that time, it was based on a GPT3 AI engine. Coverler is one of the products from Boosta IT Company. Target Audience is the US and other Tier-1 countries.
Task
✹ To propose design changes to improve the UX, which should increase the product's conversion rate.
The discovery phase aimed to identify ways to improve the user experience of the existing functionality for creating cover letters, which could increase the CR (purchasing the subscription).
✹ Stakeholder Brief Analysis
✹ Competitor Analysis
✹ Moderated usability testing
✹ Users behaviour patterns
✹ Wireframing
✹ Clickable hi-fi prototype
✹ Moderated usability testing
Discovery. Stakeholder brief.
A stakeholder provided detailed info about the product domain, features of the current website, description of the website's target audience, and the results of previous marketing&website usability research.
UI requirements
High-fidelity prototypes need to stick to the current website style and color scheme.
Requirements to respondents
Mandatory condition — respondents must know English well, have experience in job search in Tier-1 countries, and have experience in writing cover letters.
Target Audience Info
General: US or Tier1 countries, aged 25-45, 60% females
Industries: Information Technology and Services, Computer Software /SaaS, Internet
Positions: Managers, Marketers, Product Managers, Project Managers, Sales
Using sites: LinkedIn, Indeed, Glassdoor, Blind app
Interests: job search, job hunt, career switch
Key characteristic: are looking for a new job now
Discovery. Competitor Analysis.
As the main functionality at coverler.com was a form to generate a cover letter, there was a need to compare the same functionality on competitors' websites.
The outcome of the competitor analysis was that some websites had CV uploaders while Coverler didn't. Furthermore, Coverler didn't have a character counter to help users understand the minimum and maximum symbols needed to create a qualitative cover letter.
After analyzing the UX of the form by myself, I had a hypothesis regarding the example section in the 4th step — the desire of copy-pasting the text from the example to test functionality. It wouldn't provide the real value to purchase. Thus, I needed to conduct a series of in-depth interviews and moderated user testings with TA to check my hypotheses and to get insights for more hypotheses.
A stakeholder wanted our respondents to be people from Tier 1 who had experience a looking for a job there. Respondents were both Americans and Ukrainians. In-depth interviews were conducted via Zoom in English.
I created an interview scenario with research questions and questions to respondents. The results of the in-depth interviews I organised in an affinity map in FigJam.
Outcomes of in-depth interviews (partially)
As I was limited in the number of participants, I combined in-depth interviews with user testing (as the form for creating a cover letter was quite short, it wasn't exhausting for respondents).
User Testing method
Moderated user testing
Tool
Via Zoom
Number of participants
6
User criteria
Respondents knew English well, had experience in job search in Tier-1 countries, and had experience in writing cover letters.
Evaluation criteria
✹ Task success rate
✹ Time
✹ Misclicks
The results of user testings I organised in FigJam.
User testing showed that users easily coped with the 1st and 3rd steps (short data — Name, Surname, Email, Position, Company). But where it was necessary to write down their experience and requirements (2nd and 4th steps), they stopped and began to think, “What and how to write here? How much info to write down/copy-paste?”. They were faced with the fear of white blank and were frustrated. There were two patterns of how users interacted with this form:
These users copied the data from their LinkedIn or CV. The problem, in this case, is that users could copy-paste too much information. As the stakeholder explained, AI must be given not too much data to receive the best version of a cover letter.
Users copied the text from the examples to test how AI works (at that time AI hadn't gained enough familiarity, so users were intrigued how it worked in general). The problem is users didn't fill in info about themselves and couldn't objectively value the quality of a cover letter generated by AI. As a result, the chances of them buying a subscription could be lower.
Based on the results, I formulated hypotheses and created wireframes that should solve these problems and increase the quality of the cover letter.
If we add one more step in the form (by separating "experience" and "skills"), then the user will have a clearer understanding of what to write down, because skills and experience are different categories in CV.
If we add "Skills" as chips, then a system will show available options and the user will type the first letters and just choose (and fill in the form faster), because there won't be the need to write down the whole name of a skill.
The logic of using it is the same as on LinkedIn — when you start entering a skill, the system shows the available options. If there is no such skill, you create a new one.
If we delete the "example" section and instead add the brief and clear hint, then the user will feel the AI value and the desire to purchase the subscription, because the user won't copy-paste the example just to test, the user will write his/her own info and will get a peronal cover letter.
If we add the character restrictions, then the user will get a more qualitative cover letter, the value will be higher, because the user won't be copy-paste too much text or copy text from the example.
All these hypotheses were aimed to improve UX, so that the user can understand the real value and buy the subscription.
Results
Clickable prototype of the form was created in Figma