Best Restaurants Finder (with Google Maps API)

The app is live here: https://best-restaurants-finder.streamlit.app/

Background

• I had a personal problem and decided to solve it.
• In Yelp, you can only search for Top Rated or Most Reviewed restaurants. The problem with this is that the Most Reviewed restaurant may only be a 3.3 star place, and the Top Reviewed restaurant is usually a new or obscure place with 1 review by the owner.
• In Google Maps, we have the same issue. You are only allowed to search for a minimum Rating. No options to search by number of reviews.
• Therefore, since it appears that no one actually uses the product as a customer at both Yelp and Google Maps, I decided to try to implement this myself. A search engine for Restaurants (and other) based on 2 main criteria:
  1. A minimum number of reviews (eg. 500+ reviews), and
  2. a minimum rating (eg. 4.3+ stars)

Stack

• Python
• Google Maps API
• Streamlit
• Gumroad/Stripe (TBD)

Competitors

Yelp gives you top rated restaurants with 1 review:

Or gives you thousands of reviews but 3.9 stars:

Google Maps gives you results in a random order:

Major Blockers

• Google Maps API can only return 60 results max. Eg. If I want to search Asian restaurants in San Jose, it will only return a max of 60 results. There is no way to return more unless you hypernarrow the constraints, but that is not what the user (me) ideally wants. No one by default when in a mood to try a new restaurant trying to search in a 2 mile radius.
• Yelp API can only also return 50 results max.
• Until these can be resolved, maybe by scraping the internet, theres no way to accurately rank all restaurants by city.

To Do Next

• Figure out how to limit results to only a certain city. Eg. Korean food in Milpitas only
• Change filter to apply before outputting results so you get 60 filtered results, as opposed to 60 total results and then filter it down to meet the criteria.
• Redo Gmaps extract from places() function
• Create master "score" that is some weighted combination of n_rating * n_reviews. Find some way to translate 4.1, 4.5, 4.7 into S, A, B, C, D, F tier.
  • Based on market research, seems like theres an exponential decay for n reviews.
  • For example, 1000 reviews = 3000 reviews. However, 500 reviews > 100 reviews.
• Remove extraneous columns
• Make the name column be hyperlinked with the permalink.
  • https://docs.streamlit.io/library/api-reference/data/st.column_config/st.column_config.linkcolumn
  • Blocked: streamlit/streamlit#6787
• Drop down autofiller multiselect for city
• Add "Is Open Now" filter
• Add "Is Operational" filter (not perma or temporarily closed)
• Hard Add "Is Chain" filter. Eg. Gyu-Kaku Japanese BBQ has 4.4 stars, 2k+ reviews in every major US metropolitan city. Not what you're trying to go for when you're finding good new spots.
• Plot the top 10 results on a Gmaps Embed
• Rearrange dataframe column order depending on if its a mobile device (bring Gmaps link to the 3rd col) or not
  • Blocked: streamlit/streamlit#708
• Figure out how to extract/display the photos in the photos column?
• More exact error handling - what in the filter criteria caused the API to not be able to retrieve any data?
• Fully understand what radius parameter does.
  • Option to expand/reduce radius within City. After you understand it, then add it in.
  • Eg. Pleasant Hill query returns Hayward results.
• Option to specify state so that the city is not confused with another state (eg. Paris, France vs. Paris, Texas)
• Column for n miles away from current location.
• Given current location, permalink that gets directions to that place.
• Column for n minutes away including current traffic data from Google Maps.
• Any unstructured data (image of restaurant, embedded Google Maps snapshot)
• Column for current wait time at the restaurant.
• Column for make a reservation now (eg. with Resy)
• Non-urgent - current limitation is only 60 results from Google Maps API. Find a way to expand this past 60 results.