+3 votes

Post and review answers and feedback to answers in the comments section of this post.

asked in Product Improvement by (567 points) | 220 views
Clarification/ Context:
- I will ask if there is any specific issue or challenge that warrants a correction
- Is this on Yelp Mobile? (I was given this option to evaluate and I picked mobile due to mobile growth and reducing dependency on Google SEO (business challenge they have))
- I will also confirm if this is for non-special user needs (normal user) and in USA

- Improve the search experience for the normal mobile user in US (Yelp mission alignment)
- Consumer goal is to find the best place to eat and enjoy without too much effort
- Yelp business goal is to drive more traffic to restaurants (one side of the market) and also transaction growth (making money).

Key Metrics: Active usage (retention), Time to first transaction (conversion), Clicks growth on search results

Yelp Mobile Customer Segments:
I decided to focus on the existing users (captive leverage) and analyze data to find the following segments. CAC is high and user retention is cheaper and more sustainable.
- High freq user of rest search and transactions
- Current users of rest search and transactions
- Resurrected users who were dormant during one or multiple periods in past
- Dormant users (not churned but not using)

I prioritized the resurrected and dormant users to focus on as the target segments. The persona of a user is mobile tech-friendly, trust in the Yelp review system, and use 2-4 times per month.

Use Cases:
-(A) User will find restaurants based on location and find closest one based on filters to narrow down.
(B) The user will use to search for places to eat at the destination using the similar method above.
(C) The user will search for a place (destination and location) for a group such as family or co-workers or friends.
(D) A user will like recommendations to kick off their decision process
(E) A user will have a list that they trust and keep picking from it
(F) a user will try random locations and explore.

I picked use cases C, D to focus primarily.

Solutions Brainstorm:
1- To enable a user to filter choices as a group, have sent as text feature to the group so that everyone can view, engage in the selection process
2- Gamify decision making by allowing a user to select friends from Yelp friend list to drive consensus as a group
3- Create. a favorite list to share and recommend privately to one or several folks
4- Voice enable search based on stated preferences
5- Affinity-based search recommendations (machine learning cluster based)
6- Stated user preferences in profile to drive upcoming search results with standard filters for flexibility

I prioritized solution based on the following criteria with an H, L, M weight- Customer Impact, Implementation Ease, Success confidence.

Selected 5 and 6 to test as they do not add complexity or friction to user flow.

# of Clicks on search results growth
# of clicks to landing page call to action for transactions
Weekly and Monthly active usage growth
% of bounces from search results

The initial test will be through a selection of a cohort of users split among A,B, C tests. This will be run in a specific geo like SF which is dense usage for about 8 weeks.

Test A- Status quo tracked
Test B- Solution 5
Test C- Solution 6

I then summarized.

2 Answers

0 votes

I will first ask to clarity what the interviewer means by ‘improving restaurant search on Yelp’. If they give me the opportunity to define it myself, I would suggest that increase in satisfaction with restaurant selection after using Yelp would be my definition of improving Yelp.

When I think about the use cases of Yelp, restaurant search is one of the most popular ones. There are many different user groups that are looking for restaurants in Yelp and they each have different needs. Here are a few user groups I can think of:
– Couples
– Friends groups
– Foodie’s
– business professional

I will ask the interviewer if it’s ok for me to focus on “friends’ groups” as I find it difficult today to pick a place collaboratively with a group of friends. Each person has a few suggestions. After each suggestion is brought up, we each have to go and search for details of them individually, and then come back and share our feedback. It’s a painful experience to pick a restaurant with a group of friends.

Here are a few ideas I can have to improve the friends restaurant search experience
– Enable comparison table between a few restaurants on key metrics (E.g. yelp rate, cuisine type, neighbourhood, price range) and allow the table to be shared / modified easily by multiple people
– Provide information around max size of the tables for each restaurant. Some restaurants cannot accommodate large groups.
– Make recommendations based on historical Yelp reviews given by the people that are planning the dinner

And here is how I would rate the three features listed above based on customer impact and cost of development.

Feature 1- High impact on customer satisfaction. Makes it a lot easier to find good places. Development effort is low (the data is already available and there is only some front end development that needs to be done).
Feature 2- Low impact on customer satisfaction. Most restaurants can accommodate for majority of groups (given most groups of people are 4 or less). Cost is Medium to high given Yelp has to find a way to start collecting this data from either the restaurant or the users
Feature 3 – Medium impact on customer satisfaction because many recommendations might not be useful (e.g. what if a user goes to restaurants their spouse likes and now that their spouse is not coming, they like to try a different type of place). Cost is high because Yelp has to find a way to first detect which accounts are getting together and then develop an algorithm for recommending places.

I suggest developing feature 1. It’s easy to implement and makes the search process easier by bringing immediate visibility to various options at a high level.

answered by (13 points)
I like the answer. It’s following a logical process to come up with an answer.
0 votes

Define the goal using the metrics, Engagement/Revenue. Let’s say it is engagement. I will further define this to identify how success would like: #Time spent searching, Reduce number of clicks by providing appropriate recommendations.

Customer segment: Tech savvy casual diner with interest in trying different cuisines (local and while travelling) who is in the age range of 20-40 years

Customer journey:
1. Search a restaurant based on my mood
2. Filter Distance by which I have to travel from my current location: Is the travel worth it?
3. Filter How much to spend?
4. Filter What other people are saying about the restaurants I have short listed
5. Choose
6. Reserve seats
7. Leave a review

Due to time constraint, I will choose 2 stages in journey, search based on mood and what other people are saying.

– Searching restaurants based on mood- celebrating, long day at work, weekend, quite time etc.
– Too many results in reviews sometime so I do not get the crux of the review – except the rating
or many reviews with or without visuals
– Not sure how many people of similar interests have been here – validation

1. Introduce user to search by mood with pre-select options and using AI suggest restaurants based on ambience and reviews. High impact: High LOE, Med. Risk
2. Use data analytics on the review to provide labels on the most used description – Ex. Fast delivery, courteous staff etc. – High impact, Medium LOE, Low Risk
3. Picture speaks thousand words, and videos even more: Instead of reading, video reviews of 10-20 seconds along with text. – Medium impact, High LOE, Low risk
4. More information about the reviewers and giving badges as incentives. So a person with Badge review (equivalent of verified purchaser review) would have more weightage for the person searching. – Med. impact, Med LOE, low Risk.
5. Voice based search – High Impact. Med LOE, Med. risk (visual)

As a short term, I would go with 2. and long term with consider mood based voice based search.

To improve engagement = time spent searching for the right restaurant on yelp I would build a feature to get a overall sentiment of the restaurants without having to check every single review and MVP it with user segment in age group of 21-35.

answered by (24 points)
Hi Swetha,
Well done! I think it’s well structured and covers the key elements of a good answer. You started with a clear goal, then moved on to the journey of a specific segment (you could have listed other segments but I’m guessing you skipped it due to constraint of time) to highlight the pain points. Finally, you listed the solutions and evaluated them each with meaningful metrics. Nice work!

Please log in or register to answer this question.

Privacy - Cookies - Terms - Contact