Hard to classify user intents (3 keywords are not enough, Google/Hummingbird, NLP, semantic searches are ways to remedy). Internet is very noisy (spam and Javascript), algorithmic search around the tail is heavily manipulated (SEO). Each query is unique (Power law). Google tricks: autocompletion, excluding words (to return more generic results), people also searched etc. Document/Page retrieval has limitations. Search is more conversational and local.
Typical queries/searches can take hours and many page clicks before getting a satisfactory result. But more often than not leave a frustrated user with poor results.
Power laws. Few results to esoteric queries/searches.
iOS app is live (Beta being tested actively in Tokyo)
Users were vocal about lazy on-boarding and privacy.
Main feed screens: local communities are better than any algorithmic engine when it comes to cleaning/curating content. Captcha, DuoLingo (crowd translation), Amazon Mechanical Turk, Google Image Annotation/visual Captcha just to name a few examples! Accuracy and scale especially if humans are incentivized enough (e.g. healthy rewards: see local kindness/reputation coins below). ML to the rescue: every post is summarized automatically using few of the 500+ popular/common keywords and classified into one of the main 27 topics — facilitating topics browsing, answer suggestions and search.
Marketplace screens
Every Time a local freelancer gets hired, we charge a market commission based on the total charge for the service provided and facilitated through the platform.
Users accumulate coins every time they rank posts/comments (voting), accept answers (curation) and award bounties (speed and quality answers). Coins can be also “minted” by inviting connections and are capped at 100 per day (to avoid abuse). Coins can be used as bounties or “spent” to obtain coupons/discounts offered by Local business partners
