If Google Were a Restaurant

had a fascinating conversation the other day with Jon Orwant, who was O’Reilly’s first CTO, and is now an engineer working on Google’s Book Search team.

Jon noted that one of the most fascinating things he’s learned since starting at Google is its pervasive culture of measurement. He illustrated this graphically by saying something like this: “When I go to a restaurant, and look at leftovers on my plate, I don’t see food, I see information. If the restaurant were Google, they wouldn’t just take that plate and scrape it off into the trash. There would be a camera in the kitchen, photographing every plate coming back, with analysis of what people liked and disliked, and what portions were too big, helping to optimize future servings.”

leftover food:  http://www.kimberlyblessing.com/photos/v/drive05/IMG_0744.jpg (via Google Image Search)

Now, as Steve Talbott points out, our increasing quantitative sophistication shouldn’t blind us to what we can’t quantify. (Nothing in Jon’s restaurant analysis will tell the difference between an adequate meal and a great one, though it might tell us the difference between an adequate meal and an awful one.) Nonetheless, this is a thought-provoking reminder of how much Web 2.0 (or “Live Software”, as Microsoft has so insightfully called it) depends on creating information feedback loops. This is the practical plumbing that makes possible Web 2.0 systems that get better the more people use them.

  • “pervasive culture of measurement”, describes it well. I like that.

  • Rikard Linde

    Hi Tim. I tried sending an email to you using the address tim@oreilly.com. Is that the correct address?
    If you don’t want to reveal your address publicly I understand. You can send an email to rikard@gmail.com or, of course, ignore to do so:-)

  • I’m afraid to say it, but it’s a known fact within the (restaurant) industry that decent chefs keep an eye on plates coming back into the kitchen. I can’t understand restaurants that scrape plates before they enter the kitchen!

    If I leave more than usual on a plate for reasons other than the food itself, I usually ask the waiter to inform the chef that the food was ok even though it may not look like it from my leftovers.

    By the way Tim, I must apologise for not reply to a post that you wrote in response to a comment I made. I must get around to it even for the sake of the record. And, thanks for responding too.

    It’s ironic for me to comment on a post that talks about Web and food in the same breath because I manage the Web business and my business partner manages our restaurants (in Dublin, so if you’re ever in town!).

  • Oh I forgot to mention that you probably wont get trustworthy information from a Europeans plate in the US. We are not used to monstrous portions and likely to leave a lot behind even when we have had a great meal. So, like Google, some intelligent classification is likely to help gain more meaningful results that can be trusted :)

    So perhaps we could setup Content Labels that are country or culture specific. That way, the chef can ensure they have labels to gain more trustworthy results. They can then filter them before changing the menu or recipes to suit. Ok, its late and Im starting to ramble :)

  • Alex Tolley

    Sounds increasingly like Sterling’s SPIMEs. Take the analogy further – if it was like Amazon, we would have a better idea which items on the menu were popular with other customers, and which were not, which items were new, which hardy perennials, which combinations went well, with which wine. We’d already know which restaurants to avoid due to health citations, which ones had chef changes – where your favorite chef moved to. Really makes you realize how information poor we are when we walk into a new restaurant, and how the world will change when this information is collected and made public.

  • Alex, you’re talking about the Semantic guide to restaurants.

  • That would be the eat-through rate, right? (ETR)

    I love the idea of feedback loops in systems, though I’ll note that to be really effective you want to not just have the Google feedback but also a bunch of other systems doing feedback so that you get fresh insight from other teams.

  • Wasn’t is a great industrialist who’s whole core concept was “Measure everything and anything”. This of course is much easier done online where most of the measuring can all be automated. But at the end of the day how useful can this information be? Maybe on average you see that a small portion is what people like but this can change instantly and dynamically or this can blur the picture. What if its just that particular dish that is more filling or the time of the season such as summer where more people are conscience of their weight. From the example given above those factors wouldn’t have been considered.