|
Warning: potentially biased opinion. Speaking only for myself, but informed by my job. There are lots of problems with scraping-based approaches. One, yes, you need some really good tech to scrape data from menus, which, even though they are “structured”, next time you’re at a sit-down restaurant, pay attention to all the subtle discrepancies in formatting between different sections/categories on the menu. Two, if the menu isn’t html, but is an image or a pdf upload, now you need some strong OCR on top. Three, the website is generally not likely to be current with what’s actually on offer in the establishment itself. Specials, seasonal dishes, or items that are out of ingredients (“86’d”) will still appear on the menu. That’s going to lead to complaints, refunds, or generally bad customer experience from whoever’s consuming your data / using it to buy food. Four, you’re going to want to to be paid for all this tech and customer support you’re electing to intermediate between the end purchaser and the restaurant, as a service, and so you’re going to tack on some fees and either jack the price up on the consumer or try getting the restaurant to pay you a finder’s fee, cutting in to their already narrow margins. Five, if you’re trying to provide ordering service and not just menu data, you still need to submit the order into the store itself, somehow. Which either means calling it in, robo-submitting an online order (if you’re lucky), or sending a courier to place the order and wait. And then, on the other side, whoever’s taking orders for the restaurant has to punch in the request to the register to actually complete the transaction. Which means the system you really want to talk to isn’t the website, it’s the point-of-sale. Good luck with all that. Source of bias: I work for a company that helps restaurants enable online ordering and POS integration so they can pay much less in fees and focus on making exceptional food. |