Redefining Yelp

The Challenge

In 2005, finding local businesses through Yelp felt like magic. Today, this magic has been captured at the top of every search result by Google.

As a part of a pre-interview assignment for Yelp, I took on the challenge of re-imagining what the platform could do for users with the goal of increase stickiness.

My Role

Over a week, I explored four types of insights to formulate a solution: user, market, industry, and technology. A UI prototype and potential data science and business implementations were also created.

Assessing the problem

Methdology

User Insight

Star ratings and dollar signs don't capture social factors important to users.

Do the user tend to prefer a cozy or energetic ambience? Are they accustomed to Michelin starred cuisine or are they comfortable with fast food? Do they feel more at home in casual or formal environments?

Takeaway: Yelp can better fulfill users' discovery goals by accounting for social factors.

Business Insight

Segment leaders provide personalized experiences.

Spotify is #1 in music streaming subscribers.
Their recommendations engine has increased the diversity of artists listened by by 37% since 2014.

Facebook is used by 50% of all internet users.
Their personalized news feed is the #1 most interacted with feature, accounting for 85% of the company's revenue.

Takeaway: Yelp should personalize recommendations and search results to solidify it’s position as a market leader.

Technology Insight

Yelp's reliance on web users leaves it vulnerable to competitors.

Mobile Growth accounts for 87% of user growth since 2015.
79% of all Yelp searches are now made on a mobile device.

Mobile App Reach only accounts for 29% of all Yelp's mobile users.
This leaves 71% of Yelp's mobile users vulnerable to web competitors.

Takeaway: Yelp needs to find a compelling proposition for its app in order to protect against dominant web competitors.

Competitive Insight

Yelp is the only platform that combines discovery and transactions.

Discovery: Reviews and Business Information
Google Maps/Places, TripAdvisor

Transactions: Deliveries and Reservations
OpenTable, UberEats, GrubHub

Takeaway: Yelp needs to find synergies between discovery and transactions to avoid being outmaneuvered by specialized competitors.

From problems to solutions

exploring a strategy

By combining our exploration with information from Yelp's financial filings, here is how we can translate Yelp's vulnerabilities into a strategy.

Business Risks Strategy
Unhelpful/irrelevant recommendations Provide recommendations that suit the person, place, and time.
Increasing Competition Leverage Yelp's advantage as the only platform for both food discovery and purchasing
Reliance on Web and Search Provide a compelling reason for users to rely on Yelp's mobile apps

Reframing the problem.

How can Yelp recapture the magic of business discovery with today's technology in a way that makes sense for the business?

Introducing Yelp Butler.

  1. Contextual, personalized recommendations to drive discovery and ad revenue
  2. Genuinely helpful prompts to drive transaction revenue
  3. Exclusivity to native mobile apps.

Discovery

Contextual, Personalized recommendations

The right place for the time, weather, and person.

Whether it's ice cream on a sunny afternoon or a cozy date spot for Friday night, Yelp Butler's recommendations reduces the user pain of having to search multiple sources to find a good fit for them.

Opportunity: By fostering reliance on Yelp's recommendations, this increases stickiness and isolates users from web search competitors.

Ad Revenue

Microtargeted ads

Combining personalized recommendations with rich visuals.

With more accurate recommendations combined with larger UI cards, microtargeted ad campaigns have an opportunity to reduce their cost per click while engaging users in a more visual way.

Transactions

Genuinely helpful prompts.

Driving trasaction revenue by reducing ambiguity.

Whether it's a restaurant reservation, your favourite takeout order on the way home, or redeeming a promotion, Yelp Butler reduces ambiguity by keeping the call to action visible. This helps users accomplish their tasks in one-tap.

Opportunity: As what would be first to market as the only concierge style trasaction platform, Yelp Butler reduces user ambiguity of next steps while increasing transaction revenue, win-win.

Implementation

Potential Tech Implementation

Classifying businesses using existing reviews.

  • Determine features/lables for classification dimensions
  • Utlize NTLK dictionaries of words for each dimension
  • Tune NLP model via supervised ML applied to reviews & additional metadata

Classifying user preferences through search behaviour and reviews.

  • Build SVM model with selected dimensions
  • Monitor click logs to predict user preferences
  • Compare user reviews to restaurant classification to determine preferences

Generating suggestions by comparing classifications and adding context.

  • Setup MySQL database in AWS RDS
  • Combine user preferences and business classification with context (time/date/weather)
  • Prompt user to evaluate recommendations

Potential Business Implementation

Soft launch to Yelp Elite to reward existing users.

  • Rewarding esteemed members causes an exclusivity effect that can drive further growth
  • Yelp Elite users have a strong history of search and review data - increases confidence in classification and prediction.

Validating traction using growth metrics.

  • Stickiness: DAU/MAU
  • Reach: Butler WAU/Yelp App WAU
  • Depth: DW Search Volume/Yelp App Search Volume
  • Retention: Butler week-over-week retention

Gearing up for general availability with data collection and ad buys.

  • Improve classification by prompting business owners to fill out "more business info" (i.e. attire, ambience , noise level)
  • Provide incentive for data collection through free Butler ad credits.
  • Invite engaged business to create Yelp Butler exclusive promptions

Fini

Takeaways

In the end, this pre-interview assignment was not resulted in an interview with Yelp but also led to development of my data and design skillsets.

Data Science

As a part of the research process for Yelp Butler, I also started DataCamp courses and am enrolled in a data mining course at UWaterloo to further develop my understanding of data science and various algorithms.

Design

For future projects, my frustrations with Photoshop during this process have prompted me to migrate to Sketch and more recently, Invision Studio for designing screens.

Home Yelp Butler UW Food Finder Photography