Espresso Drinkers Tip More

What Our Coffee Orders Say

An analysis of 172,202 orders at a Cambridge, MA cafe. Best viewed on Tablet or Desktop.



Espresso Drinkers Tip More

In fact, the tip as a percent of total order value lines up wonderfully with the "stiffness" (for lack of a better descriptor) of a drink. Iced drink customers, at the way bottom, tip only slightly more than half what a double espresso consumer tips.

Other fun tidbits:

  • The total order size of Mocha drinkers is significantly higher than those of any other drink. The Mocha also has the most calories of any drink listed - so perhaps it's associated to some more indulgent orders.
  • The Latte is about as popular as all other espresso drinks combined.
  • Drip Coffee and Iced Coffee drinkers have a strong preference for Bagels, and are the most common orders. It's also my most frequent order, so apparently I'm not very unique.
  • Cappuccino drinkers love Croissants, obviously.
  • Visa cards are associated with the worst tippers.

Seasonal Popularity

It's not surprising that Iced Coffees are more popular in the summer. Overlay the calendars (click the "Mix" button) below to explore other patterns in drink consumption. Weekends, Holidays, and Seasons have unique characteristics.

Unfortunately, the Square reader that I sourced data from was installed mid-March of 2014, so the analysis is missing a few months. Use your imagination to fill in the dark Boston winter.



Let's start with the obvious:

  • The cafe is closed on holidays like Memorial Day, July 4, Labor Day, Thanksgiving and Christmas. It's a small business, not a Starbucks, and gives its employees some time off to celebrate the major holidays.
  • People don't like Iced Drinks when its cold out. The cafe offers Iced Coffee and other cold drinks year-round, but as soon as October hits, popularity plummets. The exception is Saturday 10/18, which was the first nice Saturday in a long while.

Perhaps less obvious observations:

  • Espresso-based drinks are generally more popular on the weekends, as people take the time to sip a Cappuccino instead of taking the drip-coffee to run.
  • Drip Coffee is more popular on Thursday and Friday than it is Monday through Wednesday. I like to think it's because people start the week ambitious and ready to save some money by making coffee at home, but by Thursday succumb to the convenience the cafe offers.

Wait Times (Kind Of)

We can't get actual wait times from the Square card reader, because nobody is stopwatching customers as they come through the door. All we know is how many orders are placed per minute. Since there is only a single cashier at the cafe, I'm going to assume that the number of orders per minute roughly aligns with how long the customer has to wait in line. Very imperfect, but probably enough to determine when to show up if you want to avoid a crowd.


(Hover over the Graph for Details)


We are obviously in a college town. The highest-volume times are around 9:40a, well after most of the working population is expected to show up to the office. A lunch rush brings in a crowd for the sandwiches, soups and salads (we're not just looking at drinks anymore), but soon the activity fades and the cafe remains slow through the end of the day. While the shop stays open until 7p, it's mainly for the laptop crowd leeching off of free wi-fi and doing work with the now-lukewarm coffee ordered two hours ago.

What else did we find?

  • Get lunch early. Volume is 38% higher at 12:10p than it is at 12p.
  • What's with 9:40a? With the exception of Monday, 9:30-9:40a is one of the biggest hot spots of the day.
  • The long Monday lunch. Somehow lunch drags out quite a bit on Monday. A lot of activity from 12p to 1p.

Drink and Bakery Pairings

Analyzing co-occurrence of Bakery items with Drinks against average popularity uncovers favorite pairings. There are two directions in this analysis. If you order Drink A, how likely is it you will order Bakery Item B? Conversely, if you order Bakery Item B, how likely is it you will order Drink A? We've highlighted high likelihoods against average as Pink, and low likelihoods as Blue.

With some drinks, like Iced Coffee, buyers are less likely to buy ANY bakery item than the average consumer. All corresponding dots are blue. Other drinks are more indulgent, like the Au Lait or Chai Latte, which see higher-than-average likelihood of ordering 5 of 9 bakery items.


Blue is less likely than average and Pink is more likely.

Drinks as a predictor of Bakery Items

Bakery Items as a predictor of Drinks


It's important to keep in mind that the "average" we compare against is based on all orders regardless of whether a drink or bakery item is ordered. With this context, we can see:

  • Customers that order a Whole Pie are not likely to order any drink at all - resulting in blue dots all the way down. These are ordered as takeaways for parties, generally. I haven't witnessed a customer facing a whole pie with their Americano.
  • Customers that order a Sticky Bun usually pair it with a drink. It's somehow less embarassing to wolf down a 500 calorie "snack" if you get a cappuccino to pair it with.

This data potentially opens some real-word implications. For customers that order an Iced Chai, there is a (relatively) high success rate if the server offers a Muffin or Sticky Bun. But don't try a Pie Slice...

To Summarize

We're handcuffed by the dataset we have (sound familiar?). Tips are only available for card-paying customers, we can't identify an individual customer across transactions, and we don't have staffing ratios or names. Do higher tips correlate to certain staff being at the counter? Or a higher number of staff leading to prompter service? Who are our repeat customers, and what drives them back?

Still, some interesting patterns emerge that, after discussions with the employees, ring true to the experiences on the ground.

In a lot of ways, this analysis is a bit disappointing. It confirms what I could have gleaned from surveying the great people that work at the cafe every day, but does so in a slightly more quantified and "pretty" way. There's a lot more I'd like to do with the dataset. Incorporating weather patterns, university class schedules, or city-wide events will probably shed some more light on consumer behavior.

In the meantime, I'll be a little more conscious when I show up, how much I tip, and what I order with my drip coffee. I'm done with bagels, bring on the slice of pie.