Psst... We're working on the next generation of Chowhound! View >
HOME > Chowhound > Food Media & News >
Oct 12, 2011 03:31 PM

Felix Salmon analyses Citi card spending at New York restaurants

I think of myself as your foppish Arts/Humanities sort, but I realise I'm a quantitative geek compared to the average American. And I do love seeing data like these breakdowns of amounts charged to Citibank cards at New York restaurants:

As I remarked on the Per Se / Sifton thread: Per Se has a larger spread of payment totals than other restaurants. And 10 per cent of its revenue derives from dinners over $3000. The median (and mode) total appears to be $800... dinner for 2 + perhaps $150 of wine + tax?

Per Se
10 Columbus Circle, New York, NY 10019

  1. Click to Upload a photo (10 MB limit)
  1. Totally useless data. Like many of of the comments to the article state, the CC bills are only the total. You have no indication of the number of covers on the bill. Makes it impossible to draw any inference between cost of dining between the spots other than Per Se is much more expensive than the other 3 but did we need Bundle to tell us that? The other 3 charts pretty much look the same and are just a typical expected distribution. Almost a perfect bell curve except that you have the compression on the left because who chrages a $1 at a resto?

    And I'm not even a stat geek. But I'm gonna use my amex on dining bills when Its below $200 and Citi when its over $200 just to screw with them.

    4 Replies
    1. re: Bkeats

      I wouldn't dismiss this data as -totally- useless. I expect Citi's sample is large enough to be fairly representative, and that Citi cards are owned by consumers across a range of incomes. (I would suspect American Express might be skewed towards higher incomes? Especially considering how aggressive they were about closing accounts during the financial panic...)

      I presume the Per Se data includes diners in the Salon, else I'd have difficulty explaining the sub-$300 expenditures. I wish they had been able to separate the Salon data...

      One of the things I learnt is that even at the very high end, I spend more than the median customer on wine. My rule of thumb is that, at a proper restaurant, wine expenditure should be equal to or greater than what I spend on food, and I always assumed at establishments like Per Se, the investment banks and hedge-fund managers would be spending much more (someone must be buying the four-figure grand cru Burgundies and clarets). It seems either banksters are cheaper than I expected, or a much larger proportion of diners at Per Se than I'd guessed hail from the 99 per cent.

      Per Se
      10 Columbus Circle, New York, NY 10019

      1. re: Winterpool

        I'm still not sure how helpful the data is. Given my experience dining out, the peak of the curve will be two people at dinner. There will usually be cocktails and/or wine. My wife and I do that on a fairly regular basis and the tab is typically 150-200 so that fits pretty much with the median of the places other than per se. But then there are the times that we're in the mood for something nicer and we're closer to $500 or we take another couple out and its even more. Then there are those times when the family is away and I decide to stop for sushi and sake on the way home from work and I blow $200 on myself or I meet a buddy and we have cocktails and apps for $100 or burgers and beer for $50. If you looked at my just my dining receipts and made a similar graph, you could not draw any conclusions without knowing more than just the total. It would look like a normal distribution except for the compression on the left as would be expected.

        As to per se and the hedgies, I think I per capita expenditure figure would help you figure that out. A group of 6 can easily hit 3k or 4 k without any grand crus. Hit that with 2, and you know some pretty good wine was had.

        And don't even get me started on how much I spend on wine. The only saving grace is that my wife likes it too so she does't complain about it.

        1. re: Bkeats

          How I wish the Salon data was filtered out... If most of the near-median $800ish expenditures were couples in the dining room, it still leaves an awfully small amount spent on wine (and that's presuming no tip in addition to Per Se's 'service compris'). I find it very difficult to believe a median couple at Per Se is spending less than $150 on drinks. Consider that the recommended wine pairing for EMP's $195 tasting menu costs $145 -per diner-...

          Thankfully the girls I take to dinner do appreciate a good Riesling, which has enabled me to explore Germany, but I wish I knew a girl who loved red Burgundy.

          Per Se
          10 Columbus Circle, New York, NY 10019

          1. re: Winterpool

            Most of the 10% over $3,000 probably comes from private dining events at Per Se.

            Per Se
            10 Columbus Circle, New York, NY 10019

    2. As both a stats guy and a food guy, not only is the data useless, but the restaurant comparisons are off too. Comparing Megu to Mr Chow? Comparing Babbo to Per Se?

      "Why would anyone but the most hard-core stat geeks want to look at a bunch of charts about how much people spend at restaurants?"
      Real stats geeks look at real data, not charts. The fact that this guy's a finance blogger makes a world of sense. If the majority of people in finance actually understood stats instead of just looking at pretty charts we wouldn't be in this mess.

      5 Replies
      1. re: fooder

        I've been browsing the Bundle site but have difficulty finding where Felix mined his data. The interfaces are too fancy for me! I suspect Felix got the raw data directly from Bundle / Citi.

        There is some interesting data on the public Bundle site: in ZIP 10014, most spending goes to Blue Ribbon Sushi and Starbucks. Average [I presume mean?] monthly spending on dining out is $828.

        His corresponding piece on Reuters isn't exactly the same as his GrubStreet post:

        As to Mr Salmon, I cannot speak to his financial or statistical credentials, but I've been reading his blog since he was at -Portfiolo-, and he has hardly been a cheerleader for irrational exuberance. If anything his attitude towards Wall St has been rather critical, even 'Leftish'.

        Edited: comparing what I presume are mean expenditures...

        Daniel ... $450-460
        Eleven Madison Park ... $410-420
        Le Bernardin ... $510-520
        Masa ... $360-370
        Per Se ... $890-900

        I'd rather have had median expenditures, and of course I'd most of all have preferred the raw data, but I suspect one has to pay for that?

        1. re: Winterpool

          You have to understand that I'm in the group of stats guys who don't consider what Bill James does to really be stats, so my view on what stats is is a bit narrow for general consumption.

          But I'm confused. How can mean expenditure at Masa be 360-370 when the minimum price is 450 pp? I assume Barmasa checks are counted in this?

          1. re: fooder

            There doesn't appear to be a separate vendor listing on Bundle for Bar Masa, so your assumption is probably correct.

            'You have to understand that I'm in the group of stats guys who don't consider what Bill James does to really be stats...'

            Ha! Not a fan of -Moneyball- then?

            I am hoping one of my friends in market research can get his hands on the raw Bundle data or something similar.

            1. re: Winterpool

              Actually I misspoke (errr... mistyped?) There are stats involved in what he does, but he is not a statistician by any means, which some people do see him as.

              The raw data would be a little more useful, but as mentioned above, the lack of per cover data really takes a lot of value out of it.

              I'm more interested in the direct comparison aspect of it. Whereas a layman may try to compare Per Se and Daniel just because they both have three Michelin Stars, the fact that Daniel tries to turn tables at least once in that big restaurant of theirs makes it a very different restaurant from Per Se.

              1. re: fooder

                Right, it seems that Per Se will always be operating at a huge advantage because its revenue per transaction appears to be twice the other Michelin 3-star / NYT 4-star restaurants. It's interesting to note that Keller introduced Per Se at, I believe $150, then escalated. Would diners have accepted an attempt by Daniel or EMP to increase their prices well above $200? (Daniel has a $220 menu but offers many lower price points.)

      2. Felix Salmon has continued his analysis of Bundle data and asks, 'Where the one percent eats'.

        'Perhaps even more interesting is the fact that the percentage of luxury diners at the city's super high-end restaurants don't even come close to ranking among the top. For instance, just 3 percent of Per Se's clientele comes from the luxury crowd; Eleven Madison Park is at 2 percent.'

        '...but the rich tend to be neither gourmands nor food writers'

        1 Reply
        1. re: Winterpool

          As with most economists, they get access to good data, do a cursory correlational analysis, and end up with useless conclusions because they know nothing of the real world. This article is another case in point.

          Rich people have always favored convenient, easy foods and were in general never gourmands. It's why filet mignon and jello (because gelatins could be molded into pretty things) were rich people foods while lobsters were served to prisoners.