Dating App Analysis

Frank Lapinski, Carly Siegel,

Alice Heyeh, and Chase Seklar

Objectives

Goal: Build a new dating app for college students

Objectives

1. Understand how customers rate features
1. Key Features (i.e. Superlikes, Boosts, etc.)
2. Our Planned Feature: Date Recommendations
2. Market Share (current; w/ new product)

Process

1. Research apps to find out features
2. Develop a set of different products to test and create a survey using Ratings-based Conjoint Analysis (RBC)
3. Analyze data to understand feature importance and market

Preliminary Research

10 Apps Researched

Features & Levels Chosen

Research Process

• Installed apps, looked at different subscriptions, & selected features that were key to dating apps
• Then we boiled down the key features into different levels

Data Collection

Defining the 6 key features across apps:

Data Collection

Posted Survey On Features on College Reddits (117 Respondents)

Majority Demographics:

College Students, Ages 18-30, East Coast Universities

Minority Demographics:

Ages 30-45 Survey Takers

Data Analysis

What is RBC?

• Ratings Based Conjoint (RBC) is a quantitative marketing research technique that analyzes the trade offs customers make when selecting products.
• Customers are surveyed based on profiles created from the different levels a certain factor may have. They rate each profile based on how much they like or dislike it. This data is analyzed in aggregate.

​

Factor

Levels

• We used RBC due to its capabilities in comparing the relative importance of dating app attributes and calculating potential market share of a new dating app along with its accessibility for analysis in Excel (unlike Choice-Based Conjoint).

Participant

Prefers

Lower Prices

Data Analysis

Profile Creation

• Created 18 dating app profiles using fractional factorial design. Essentially, we chose the minimal number of profiles to understand the individual-level effects each potential option of each factor provided.
• We did not use full factorial design = Profiles for all combinations of every level for each factor =

3*3*3*2*2*2 = 216 possible options for survey takers to choose from = WAY TOO MANY!

• Each profile was rated on a sliding scale from -100 to 100 (-100 being very unlikely to use and 100 being very likely to use)

Ratings-Based Conjoint Analysis

Analysis Process

• Once we had our products, we performed a regression. Each (Fx, Ly) pair in the equation represents a categorical variable telling us if the factor at a certain level was present in the product the customer considered where Az = the individual level effect or part with a that feature at that level had:

​

• We performed a matrix calculation to determine the regression coefficients for all 117 respondents. Each individual had their own regression. We take the simple average for each coefficient A for each respondent’s regression to develop our aggregate regression.

​

​

​

Key Insights - Importance and current market

Factor Importance

• Price most important by a landslide
• Superlikes per week and Boosts per month second-most important factors
• Our new proposed feature was not highly valued by users

Current Market Share

• Calculated market share by plugging each of our dating apps into all of the 117 respondent equations and to see which option each customer would choose
• Most popular apps were free; Match Basic was most popular due to being free and having an additional feature
• Badoo Premium was least popular despite having a lower price than the other two paid options

Key Insights - Importance and current market

Our Product

• PennDates is able to provide the most features at lowest price
• Has a huge demand by doing so = potential to undermine other market players

New Market Share

• By offering one of the features of Badoo Premium plus, PennDates is able to steal a large portion of its market
• By offering more features it also takes a large portion of the Basic Free App and Match Basic

Improvements and Future Research

Areas for Improvement

• Model didn’t reflect reality (i.e. Tinder/Basic Free App not Badoo is most popular)
• More demographic data collected for trends among ages, location, etc.
• More mundane realism would’ve helped customers better rate applications (i.e. UIs for the dating app experience)
• Survey had monetary incentive which may have influenced user responses

Future Research

• Seeing how the number of active users in addition to testing more features to see the impact it had on user preferences.
• Check out this video on the Cold Start problem
• Much of this data could’ve been used for cluster analysis, allowing us to see if certain features may have been preferred together

Resources

• Much of the information was from Zhenling Jiang’s class. Unfortunately I can’t share it here but you can reach out to her to ask. You can find her CV and contact here.
• Our analysis and data can be found here.