this past week, doordash launched zesty, an ai-powered app for discovering local restaurants.
their pitch: instead of reading reviews, scrolling social media, and toggling between maps and menus to find a good restaurant, you simply tell the chatbot what you’re looking for (e.g. “brunch spot good for groups” or “romantic with a vintage feel”). from there, zesty will pull personalized recommendations based on information from doordash, google, tiktok, and more. users can also post about restaurants they’ve visited, follow others, and generally treat it like a mini social network.
screen grabs of the zesty app
i think this product bet is a miss, for two reasons: no right to win, and limited tam (total addressable market).
1) no right to win
google maps already dominates this use case. it has a massive installed base of users who have woven the product into their daily lives: saving restaurants, building lists, leaving reviews. these routines generate massive amounts of invaluable + proprietary data for better recommendations than doordash can ever hope to match.
plus, anytime zesty ships an innovative feature, google can quickly copy it and ship it to their massive user base.
2) limited total addressable market
one of my favorite startup talks ever is from dalton caldwell and michael seibel (managing partners at y combinator) on “tar pit” startup ideas: big ideas that seem obvious, but really have been tried and failed a million times. restaurant discovery apps is one of the first examples they provide.
at first glance, the opportunity seems obvious: why isn’t there a better app to help me find more great restaurants? the real answer: in most towns in america, there just aren’t many new great spots. the problem isn’t non-ai search; it’s supply. in most neighborhoods, there’s only a few new restaurants a year, and you hear about them from a friend. or you drive by. or… they’re just on google maps.
imo, this is the real nail in the coffin for zesty. it might feel useful in major cities like nyc or sf, but this value prop won’t generalize to smaller cities (i.e. most of the country), where there’s just not enough restaurant turnover to make a standalone discovery app worth using. notably, doordash is testing zesty exclusively in san francisco and nyc, which feels like another miss! they might see positive data signals in these markets that will never carry over to smaller metros throughout the united states. a good reminder for product builders to choose your beta customers wisely, so as to avoid sample bias!
ultimately, i still joined the zesty waitlist, and i’d love to be proved wrong. but i’m highly skeptical, and i’d bet they wind it down over the next couple years. to me, this feels like a quintessential “we could do this ai thing, so we did!” instead of “we should do this ai thing, because we can dominate this space for x, y, z reasons”.