Stuck with 2Ys –“Latch” – Final


  • Opportunity

According to Pew Research poll, 40% of Americans use online dating(1) and 59% “think online dating can be a good way to meet people”(2). UK country manager of a dating app, eHarmony, Romain Bertrand mentioned that by 2040, 70% of couples will meet online(3). Thus, the online dating scene is a huge and ever growing market. Nevertheless, as of 2015, 50% of the US population consisted of single adults, only 20% of current committed relationships have started online, and only 5% when it comes to marriages(1). There is a clear opportunity to improve the success rate of dating apps and improve the dating scene in the US (for a start). As per Eli Finkel from Northwestern University (2012) (3), likelihood of a successful long-term relationship depends on the following three components: individual characteristics (such as hobbies, tastes, interests etc.), quality of interaction during first encounters, and finally, all other surrounding circumstances (such as ethnicity, social status etc.). As we cannot affect the latter, dating apps have been historically focusing on the first, and have recently started working with the second factor, by suggesting perfect location for the first date etc. For individual characteristics, majority of dating apps and websites focus on user-generated information (through behavioral surveys) as well as user’s social network information (likes, interests etc.) in order to provide dating matches.  Some websites, such as Tinder, eHarmony and OkCupid go as far as to analyze people’s behavior, based on their performance on the website and try to match the users to people with similar or matching behavior.

Nevertheless, current dating algorithms do not take into account vital pieces of information that are captured neither by our behavior on social media, nor by our survey answers.

  • Solution


Our solution is an application called “Latch” that would add the data collected through wearable technology (activity trackers such as Fitbit), online/offline calendars, Netflix/HBO watching history (and goodreads reviews), and user’s shopping patterns via bank accounts to the data currently used in apps (user-generated and social media) in order to significantly improve offered matches. According to John M. Grohol, Psy.D. from PsychCentral, the following are the six individual characteristics that play a key role in compatibility of people for a smooth long-term relationship (4):

    • Timeliness & Punctuality (observable via calendars)
    • Cleanliness & Orderliness (partially observable – e-mails/calendars)
    • Money & Spending (observable via bank accounts)
    • Sex & Intimacy
    • Life Priorities & Tempo (observable via calendars and wearables)
    • Spirituality & Religion (partially observable via calendar, social media, Netflix/HBO patterns, and e-mail)

Out of the six factors mentioned above, 5 are fully or partially observable and analyzable through the data already available online or offline via the sensors mentioned earlier. As all the information we would request digs deeper into privacy circle of a target user, we would be careful to request only information that adds value to our matching algorithm and will use third-parties to analyze such sensitive info as spending patterns.


  • Commercial Viability – Data Collection

The dating market size has reached $2.5 billion in 2016(6), there are over 124 million accounts registered on online platforms.As a new company entering the market Latch would have a clear advantage over the current incumbents, as it would not have to use old and commonly used interface of dating process. As per Mike Maxim, Chief Technology Officer at OkCupid, “The users have an expectation of how the site is going to work, so you can’t make big changes all the time.”Prior to the launch, we would have to collect initial information. In order to analyze only the relevant data we would have to analyze the behavioral patterns of current couples before they started dating. Thus, we would aggregate data available on their historical purchase decisions and time allocation in order to launch a pilot.

  • Pilot

The pilot version will be launched for early adopters based on human- and machine-analyzed historical data of existing couples. Step 1 – Collecting Initial DataOur initial observation pool will be University of Chicago and peer institution students. We chose this group of people, as in order to compile the initial data on compatibility we would need to have a trust of people providing us with their private information, such as email accounts. We will start by approaching existing couples across the universities, focusing on couples that were formed recently. In order to attract as many students as possible, we will offer university-wide benefits, such as cafeteria coupons, book discounts etc. As a result, we will collect historical information (before the start of a relationship) on social media, e-mail, calendar, fitbit and other activity from tools mentioned earlier, by gaining access to respective accounts etc. Step 2 – Analyzing the data

We will combine all the collected data and observe for recognizable and significant patterns. For example, how much being early birds or night owls have hustatistically significant effect on the likelihood of them matching their partners etc. Using machine learning we will analyze the data until we find all the significant variables that contributed to the likelihood of our existing couples matching. We will be focusing on the following observable characteristics in each tool used:


Tool Characteristics
Calendar Accounts
  • Actively fill in Google Calendar, iCal, or other calendar at least during last 6 months
E-mail Accounts
  • Topics discussed
  • People contacted
  • Overall organization of e-mails (how organized are incoming e-mails, how many days on average e-mails are unread etc.)
Social Media
  • Topics discussed
  • Things liked
  • People contacted
  • Locations visited
Entertainment Accounts
  • Movies, Books, Songs watched/read/listened/liked (ranked)
  • Sleeping patterns
  • Preferences in sports
  • Activity Level
  • Nutritious preferences
  • Physical parameters
Bank Accounts
  • Purchases by categories, amounts, time of the year, season, brands


Step 3 – Attracting Early Adopters / Market adoption

Based on the compatibility variables derived from the initial observation pool analysis, we will attract early adopters. In order to attract early adopters we will position ourselves as the new dating app that uses a much broader set of data in order understand the user and find him/her the perfect match.

Create awareness:

In order to attract early adopters we plan to promote the app during events that attract large numbers of people, such as musical concerts, sport events and various fairs.  We would also offer promos around important holidays, such as Valentine’s day or New Year Eve.

From Awareness to Download to Use:

U.S. Millennials have 3-4 dating apps installed on their phones on average  due to the low quality of matches, all of them require attention (e.g. swipes, coffee beans etc.) and can be quite frustrating, hence we will promote usage by directly addressing the low return-on swipe utility.

In order to lower the churn rate that is very high for any given app (50% per year)(7), we would make the initial registration very easy: no long surveys to be filled out, only linking your social media account (facebook). Once the user starts using we will show improvement opportunities for higher quality matches by adding additional sources of data such as e-mail accounts, calendar accounts, other social media accounts (incl. professional), entertainment accounts (Netflix, Audible, Amazon Prime etc.). Once the user starts using the app we will offer more time consuming opportunities such as answering survey questions developed by behavioral psychologists.

Step 5 – Growth

Second Phase Expansion 1. Future expansion opportunities exist in increasing the number of information sources once the user base is large enough. On top of existing data sources (e-mail accounts etc.) we will expand into integrating DNA ancestry information (such as provided by MyHeritage DNA) as well as medical history information.

Second Phase Expansion 2.

Using the collected user preference data we will expand into offering perfect first date setups, such as movies that would be liked by both users. This will create monetization opportunities for referencing users to events and restaurants.

  • Team:

Alexander Aksakov

Roman Cherepakha

Nargiz Sadigzade

Yegor Samusenko

Manuk Shirinyan

  • Sources:

Stuck with 2Y’s: Rewarding Safe Driving (Profile)


The Opportunity

Road safety continues to dominate as a non-natural cause of death. At the same time, car insurers continue to base their pricing decisions mainly on the history of accidents, car age, and tickets, as well as self-reported mileage. In the face of increasing average age of cars, as well as increasing repair costs, that strategy decreases insurance companies margins.

Yet before the self-driving cars populate our streets, augmented intelligence is capable of assisting the society by detecting bad and good driving behavior. Safe driving could be rewarded.


Driveway, a telematics software start-up founded in 2008, developed an application that allows measuring driving behavior1. A GPS-tracker allows to measures speed, acceleration, braking habits2, with all these data put in the location (speed limits, road quality, altitude) and time (weather, brightness) context. Additionally, the mobile application also analyzes distracted driving data such as texting (that is so proven to be a cause of many accidents). The software app analyzes each trip and provides a ranking of the driving behavior as well as suggestions on what could be improved.

There are two major uses of the data: individual car insurance pricing and selling of the insights generated from analyzing the data (such as identifying bottlenecks in the road structure, understanding the aggregated driver behavior in response to various factors, learning the patterns that lead to collisions etc.)

Effectiveness and Commercial Promise

Driveway shoots for the stars by creating a personal driver’s profile and offering insurance programs at a discount if the profile outperforms peers – a positive reinforcement. They also plan to sell aggregate driver’s behavioral information to insurance companies, so that they could offer better programs to different categories of drivers.

The company competes with the legacy insurers that offer customers to install GPS devices in their cars and provide behavior-based pricing based on tracked GPS data (23% insurers launched usage-based insurance programs, but only 8.5% customers opted in as of 20143), as well as independent telematics firms such as Root, Drivefactor, and others4.

Other parts of the ecosystems are the marketplaces where different companies can exchange their data to make all risk scoring models more robust: Verisk Telematics Data Exchange5, DriveAbility Marketplace.

Driveway uniqueness comes from its integration of the distracted driving sensors, proprietary algorithms, as well as making it cheaper and faster for customers to enroll (as no device has to be physically installed in the car). Essentially, Driveway uses smartphones as a source of sensors and transforms the data into unique insights through the proprietary algorithms.

Motor Vehicle Accidents are the leading cause of unnatural deaths. Hence, besides commercial potential, Driveway offers a great social benefit. a survey of drivers has shown that a turned on telematics device changes behavior: 56% of surveyed drivers who installed a telematics device reported changes towards safety in the way they drive3.

Unnatural causes of death:

Alterations and the untapped potential

More available data and combinations of sensors might increase the number of driving metrics. Having enough history and combining it with human-based evaluations, Driveway could evaluate how tired a driver is, whether he/she is intoxicated and/or overall just inexperienced. Furthermore, Driveway could expand a use of its software behind pricing car insurance. Potential applications could be cross-marketing, finding high-quality candidates for Uber/Lyft peak hours and routes.

It would also be useful to bundle Driveway with other products in order to minimize positive selection bias (i.e. only good drivers opting in for user-based insurance). While such bundling poses certain ethical and privacy challenges, one example that would be beneficial to both society and businesses is for people to provide their Driveway data in order to get a credit score.

The data can also be used for behavioral/social studies and offer unique insights into non-driving related fields.


Team members:

  • Alexander Aksakov
  • Roman Cherepakha
  • Nargiz Sadigzade
  • Yegor Samusenko
  • Manuk Shirinyan