The Coffee Warehouse

I have a bit of a love hate relationship with Starbucks. It feels expensive. The lines are long. And I resent the fact that I give them an interest-free loan every time I use their mobile app. But my go-to Pike and banana nut loaf are delicious, and the baristas at my preferred location are fun. They give me a hard time if I deviate from my usual, which I appreciate. When I’m feeling uncertain about the day, making a Starbucks run is a surprisingly good way to get my head on straight.

On a recent visit, I arrived at the pickup counter and found my order incomplete. The slice of banana bread had been warmed, bagged, and labeled, but my cup of black coffee had yet to be dispensed into its paper cup. Looking behind the bar, I saw the usual blur of green-aproned baristas moving from task to task. It was frenetic, but organized too, with defined work areas and clear routines. The scene reminded me of my days working at a distribution center. The area where we packed orders had a similar manic choreography, and watching the baristas go about their jobs I found myself trying to understand their order flows and processes. Specifically, I wanted to understand how Starbucks organizes and prioritizes its work.

Customer Flexibility vs. Operational Complexity.

Starbucks is in a bit of a slump. Sales in established locations have fallen for 5 consecutive quarters, contributing to a recent change in leadership. In an attempt to win back customers, the new CEO, Brian Niccol, has made operations a focus and pledged to reduce wait times and improve the customer experience. They are investing heavily in their order sorting algorithms and store processes, with the topic getting conspicuous attention in recent earnings calls.

Some of the operational challenges stem from the increasing importance of their mobile app. Since 2015, Starbucks has allowed customers to place orders remotely, before they arrive at the store. This grants convenience and flexibility, offloads the labor associated with order entry, and as mentioned previously, encourages customers to give them free loans to earn “mobile rewards.” But this convenience and savings come at the cost of operational complexity. There are three sales channels at a typical Starbucks today: walk-ins, drive thru, and mobile. Drinks are processed in the order received, whether placed in person or through the mobile app. This first-in, first-out system creates challenges, particularly at busy times. Operational capacity is often devoted to mobile customers who have yet to arrive, while those already at the store grow impatient. The staging area gets crowded with completed drinks, leading to that awkward seek and find many of us have experienced. 

Though mobile orders create challenges, they also represent a kind of operational opportunity. They are different from orders placed through the traditional sales channels, where customers are present at the restaurant and presumably want their coffee as soon as they can get it. With mobile, customers generally place the order before they arrive and don’t care precisely when it is finished, as long as it is complete, and reasonably fresh, when they get there. This arrival delay makes it sensible to consider processing work out of sequence.

Warehouse work

Back in my warehousing days, we thought a lot about how and when we processed work. Through our website, customers could place orders at any time of the day or night, and we committed to getting them their stuff in two days. The parcel carriers needed most of this time to get the shipment to the customer, but we generally had a few hours to fill the order, pack it, and load it onto the truck. We took advantage of this window to operate more efficiently. A few principles guided our thinking:

1. Urgent Work First: We prioritized packages that had time constraints. If an order’s truck was leaving soon, we’d complete it first — even if there were other orders that had come in before it.
2. Stay In Sync: Many customers ordered more than one thing. Even if the items were stored in areas far from one another, the customer would still expect them to show up in the same box. We would begin processing all components of that order simultaneously and only when we had enough capacity for all the processes involved.  As a result, the components of an order would arrive in the packing area at a similar time, where they would be combined and put on a truck. This reduced the amount of incomplete work floating around the warehouse, which in turn reduced opportunities for error.
   
   A consequence of the above is that simple orders (with fewer work areas involved) tended to move through the system faster. They didn’t get held up behind large and complex shipments that gobbled up lots of warehouse capacity.
3. Batch Work Where Appropriate: The longer the work queue, the more likely you are to see collections of similar jobs in the queue. So we designed work areas to take advantage of this phenomenon, batching jobs that used the same skills and materials. 
   
   Imagine a work area where people pack boxes. A packer will complete two identical boxes more quickly and accurately if she does them back to back, as the packages require the same materials (container, tape, infill, and packing surface) and skills. If she needs to shift to some other task in between, say to pack a different type of package, the process will naturally go more slowly. This is pretty much the same principle that makes assembly lines work: Keep people focused on the same task, and they will perform that task more efficiently.

Starbucks as a Coffee Distribution Center

The work queues at Starbucks are measured in minutes of work rather than hours. Still, there are a number of warehousing principles they could easily adopt.

1. Urgent Work First: Starbucks could prioritize orders based on the anticipated customer arrival time. There is no need to have the coffee sit if a customer isn’t there to drink it. Mobile orders should be distributed to baristas closer to when the customer is expected to arrive. This preserves capacity for in-store visitors, and ensures that mobile customers always get hot (or cold) drinks.
   
   The people at Starbucks are well aware of this opportunity. Internally, they have a “4-4-12” performance metric (four minutes to get a drink when walking in or driving through, and twelve minutes via mobile), which they have struggled to meet at busy times. Earlier this year, they began using a new order sorting algorithm that tries to shift mobile orders closer to the time when the customer arrives to pick them up. In the stores where this algorithm has been implemented, in-store service times have fallen by 2 minutes, with 75% of orders meeting their 4-minute standard. Importantly, the new algorithm did not cause an offsetting increase in mobile service times. The company expects to roll out the change broadly this year.
   
   This change seems to be related to the recent acquihire of the leadership and engineers at Empower Delivery, a company that specializes in helping omnichannel restaurants sort orders so that kitchens don’t get overwhelmed. In this video, a representative walks through the features of the Empower system, which sounds quite aligned with Starbucks current challenge.  
   
   Starbucks is also experimenting with allowing customers to select an expected arrival time, a behavioral tweak that might help align mobile order completion times with customer expectations. The app currently tells you when they expect your order to be completed, but the number isn’t always accurate and I generally ignore it, relying on my own internal heuristics instead. If I were to pick the time myself, I might be more likely to show up at the expected time. Another approach would be to adopt the geofencing technology of other quick-service restaurants, like Chick-fil-A and McDonald’s. These companies only start working on mobile orders when the customer is close enough to ensure their food won’t sit too long.
2. Stay In Sync:  The behind-the-bar area at Starbucks is organized into several work areas, each with the appropriate equipment for making food and drinks. There is a hot bar built around one or more espresso machines (for making lattes, cappuccinos, etc), a cold bar featuring blenders and ice (frappuccinos, refreshers), and a warming area with the necessary ovens (food). In addition, there are separate customer service areas devoted to taking drive thru and walk up orders. At full capacity, one or more baristas are staffed to each of these places. 
   
   But capacity in each area is not always matched to order demand. Similar to my old warehouse, multi-item orders often involve more than one area.
   
   I think back to the moment when I arrived to find my banana loaf ready, but my coffee nowhere to be found. My order was in progress, my banana loaf taking up counter space and making it harder for other customers to find their orders. Maybe it would have made more sense for Starbucks to delay my bread slightly, prioritizing other customers' food orders until the hot bar was ready to pour my Pike. After all, I needed to wait longer either way. This type of logic could be implemented as a part of the new order sorting algorithm (and might very well be).
3. Batch Work Where Appropriate: Starbucks has clearly put a lot of time into the design of the work areas described above. I’ve observed baristas working at peak times, and it's impressive to see how quickly they can move from task to task. This is enabled by great training and smart placement of the tools and equipment needed for each type of beverage. For their busiest stores, Starbucks is taking this a step further with a “Siren” cold bar system (video), which includes automatic ice dispensing, faster blenders, and automatic milk dispensing.
   
   Despite innovations like this, sequencing work differently could create greater efficiency and overall speed. Just as with my box packing example, it would likely be faster to make two Lavender Oat Milk Lattes in sequence, rather than adding a Caramel Macchiato in between them. This could be done by examining the work queues and choosing to process drinks with very similar attributes together, rather than sorting orders purely by the time at which they were placed.

What Business are they in?

The problem is that Starbucks doesn’t want to be a coffee distribution center. They think of themselves as the premium, experiential coffee brand of their beginnings. But the mobile experience (and perhaps drive-thrus before that) have made them something different, and they don’t seem to have fully accepted it.

In a recent podcast interview, founder and former CEO Howard Shultz referred to the mobile app as:

the biggest achilles heel for Starbucks. It created unbelievable convenience for our customers. But we are an experiential brand. As this thing was growing, it became so seductive for the company. It was beginning to deteriorate the third place experience in the sense of ‘community.’ Then it overflowed to the point where it disproportionately created an environment in our stores where the mobile app became the primary vehicle as well as the primary vehicle for dissatisfaction, because people couldn’t get their drink on time, people were confused whether that was their drink. A lot of anxiety. 

The company is obviously aware of how mobile ordering has changed its business, but doesn’t want to fully commit to a service model built around this variety of customer convenience. There are certainly emerging competitors, like Dutch Bros Coffee and Scooter’s Coffee, that have no such qualms. They have totally stripped away the in-store “experience” in favor of drive-thrus and mobility. I wonder if by trying to be both things — coffee house and coffee warehouse — Starbucks fails to do either well.

Scope Creep

• It turns out that the receptacle you drink from can have an impact on how you experience a beverage...