Author Gary Nakanelua

Experimentation At Blueprint: Customer Experience, Data Science, & Coffee

April 29 2018 | Data Science, Customer Experience, Experimentation

It started with a basic data challenge. A group of our data scientists were working on an advanced video analytics solution and were at the point where additional video data was required. Ideally, the video content needed to include a group of employees performing a series of jobs while interacting with customers. This is when our invention and disruption team stepped in. 

The Opportunity
We took this as an opportunity to run a micro-experiment. As a company, we target our digital strategy solutions and products towards a number of key industry verticals, so we chose to focus this experiment on the Retail vertical, where it would be easy to blend company operations and customer experience into a single video feed. We’ve created some experimental concepts for clothing retail that would have been a great fit, but decided to target a retail market that is part of the very fabric of the Seattle area: coffee.

The consumer coffee drinking experience has evolved in the last ten years. Instead of a large coffee with two sugars and one creamer in 30 seconds from a fast food drive-in, we have been invited to an Italian coffee bar experience. Coffee takes longer, but we learned words like “Venti” and “Grande” while having our coffee customized specifically for our individual tastes. While local coffee bars and national coffee chains are always looking to enhance and improve the coffee drinking experience, the basics remain the same; choose your drink, select a size, personalize it, pay, and wait for pickup.

In this micro-experiment, we chose to challenge those basics and ask a simple but key question: would customers embrace a coffee drinking experience that focused on the expertise of the barista, rather than a large menu and quick delivery? The question has roots to an eating experience found at a number of Michelin star restaurants, where customers are asked to trust the chef. From the menu to how the food is prepared and delivered, everything is part of a carefully crafted culinary journey, led by a chef that is intimately attached to every plate that leaves the kitchen. Would coffee drinkers trade personalization and speed, for a similar experience?

The Approach
To answer our question and generate the video data needed, we built a low-resolution prototype, which is basically a quick and inexpensive way to communicate an idea. In the technology world, we often think of low-resolution prototypes as napkin sketches, whiteboard diagrams or simple mockups. However, a low-resolution prototype can also be an experience, as long as it’s cheap, fast to put together, testable, and built with the customer in mind.

After some preliminary planning, we started internally recruiting our coffee professionals. This was an important step on our path to answer our key question, as the bulk of the experience was dependent upon the skill of the baristas. Two folks from our marketing and accounting departments, James Snyder and Addie Hallstrom, were a perfect fit as they were both currently coffee enthusiasts and baristas during their college days. Our final addition was Suzy Dolbow, Blueprint's marketing manager, whom would play the part of store manager.

We chose to host our experiment at our headquarters (HQ) building. Normally, an experiment like this would be conducted at our makerspace/lab facility, but we wanted the foot traffic that occurs at HQ, in order to simulate “the rush” that coffee shops tend to experience in the morning. In order to generate a healthy variety of data, we decided to leverage two different operating layouts for the experience. One would put the entire operation in a linear, bar style format. The other would be in an “L” shape. We quickly ran through a mock operation of each layout, similar to the “Speedee System” layout test shown in the movie “The Founder”.

Next, we planned the menu. Aside from the baristas, this was an incredibly important component of the experience. It needed to be simple yet invoke a sense of curiosity on the part of the customer. We took advantage of the two coffee bean varieties we keep in stock at HQ, and highlighted the brew methods available. This would help bring the customer closer to the coffee making experience and provide a point of conversation during the point of sale. We chose the AeroPress and Chemex as our premium brewing methods. The nature of these two brewing methods is beyond the scope of this article, but both are popular for competitive use in the World Brewers Cup and World Aeropress Championships. As a control, we also offered drip style coffee and highlighted brew times for all three methods so customers could make an informed decision.

Our menu

In order to capture the video data that the data scientists needed, we decided to take advantage of three different cameras. All would be positioned behind the line, with one on each end. The third would be a 360 degree camera. We had success capturing 360 degree video during our 2018 Innovation Cup finals and the footage would give some interesting perspective on the operational nature of the experiment.

Finally, we put together the supporting materials that would enhance the overall experience. Because this was a low-resolution experiment, we kept our experience enhancements to a minimum, leveraging donated equipment and decorations. Using Azure Database for PostgreSQL, we quickly put together a simple point of sale system to track all the orders processed during the micro-experiment. Branding is important, so we came up with a quirky branding strategy for our coffee shop, and finalized our todo list to complete the following morning when the experiment would kick off.

The point of sale system we created

The Results
Based upon the results of our micro-experiment, it appears that customers would embrace a coffee drinking experience that focused on the expertise of the barista, rather than a large menu and quick delivery. 

The experiment started at 9am and ran for a full hour. At one point, we had invited everyone from a morning standup to come to the experience, creating a rush for the entire coffee staff. Unbeknownst to the coffee staff, we had planned a couple moments of chaos, one being the point of sale system would go down, the other was a mock robbery. The latter was handled in good fun, but it occurred during a rush, which added an additional layer of commotion.

Our first customer

Of all the orders that were handled in the hour, only 22% of customers chose the drip style brewing method. The other 78% was evenly split between the AeroPress and Chemex brewing methods. The bean selection was nearly evenly split as well, with the Vivace bean coming out ahead by only a few percentage points. Based upon time of order, the entire coffee experience averaged 2 orders per minute, which is impressive, considering one of the brew methods had a brew time of nearly 5 minutes.

For the data scientists, we were able to generate nearly 60 GB of video data, spread across three different cameras. Video includes opening shift, customer interactions, two different operational layouts, slow time upkeep, and closing shift.

With the primary objective completed and preliminary validation of a retail use case (albeit, limited in scope), we consider this micro-experiment a success. It was incredibly low-cost, quick to setup, and everyone involved had a great time. Plus, we were able to provide a unique Friday morning breakfast experience to our folks at HQ.


From webinars and hackathons to experimentation engagements and workshops hosted at our lab facility, if you are curious about how to approach ROI driven experiments within your organization, contact us. We can help you discover your next big thing.


Ready to start your next big thing?

Contact Us