I was reading an article regarding the former CEO of Starbucks meddling in the current affairs of the company, and that too, publicly on LinkedIn. Somewhere in the article, titled, How Howard Schultz is Back-Seat Driving at Starbucks, this caught my attention:

“Customers are fed up with mobile sales that takes too long to complete. They abandon their orders before paying.”

I am not a frequent Starbucks customer since I am a tea person. So I had to research a little to find what the mobile ordering process is at Starbucks. As per the company website:

” Mobile Order & Pay allows customers to place and pay for orders in advance of their visit and pick them up at a participating Starbucks® location. The mobile ordering experience is seamlessly integrated into Starbucks world- class mobile app and My Starbucks Rewards® loyalty program.

Store locations appear based on the GPS functionality of a customer’s iPhone® or Android™ device. Upon first use, customers will be asked to accept location services, allowing Starbucks to identify the nearest location offering Mobile Order & Pay. If customers choose not to accept location services, they will not be able to use the Mobile Order & Pay.

Proceed to the selected Starbucks® store to pick up food and beverages: Orders are immediately sent to the selected store where Starbucks partners (baristas) will begin preparing the items.”

From the process description, the bottleneck can only be in the following areas:

• Lag in mobile orders getting in the queue at the assigned store. This is a highly unlikely scenario that we will disregard for our analysis.
• Too many orders are assigned to specific stores, which overwhelms the staff and the current processes at the stores. This seems like a realistic bottleneck, so let us pursue it.

I will obviously make certain assumptions since I do not have the data. So, we are going ahead with the hypothesis that too many orders are getting assigned to specific stores, which overwhelms the store staff. Let us start with the first cue in the process description on Starbucks’ website: “the nearest location offering Mobile Order & Pay.” This indicates that not all locations are currently enabled, and not all stores offer mobile order and pay.

This leads to another assumption. There is a cost of “enabling” a store for mobile pay. The decision whether to enable or not depends on whether the volume at the store is high enough to cover the enablement cost and then some.

The first question is, what is the additional cost of “enabling” mobile ordering at all stores vs a select few? There are two ways analytics can help analyze this.

One scenario is that if you offer the products at the same price for mobile ordering, will investing in enabling specific stores will help bring additional revenue to offset the cost at the minimum? The second aspect is, if you add a small mobile ordering surcharge, will your customer be willing to pay that?

Let us start with the first scenario where there is no price change. You can perform this type of analysis, leveraging some intuition in Excel as well. There can be various rough-cut approaches.

For example, you can look at the revenues or transactions for locations in the same zip location. If the transactions or revenue for a store without mobile ordering is close (within a certain percentage) to revenue of a store that has mobile ordering enabled, these stores may be good candidates. If a customer learns through experience that traveling a bit further to another location cuts their waiting time in the drive-thru by half, they may be willing to switch to that store, and offset the load at overwhelmed stores. However, with this rough cut analysis, one challenge still remains. Will enough customers pursue this route, so as to make “enabling” the stores feasible?

The second scenario involves price optimization as well. How much “mark-up” will customers be willing to accept for mobile orders, if that cuts their waiting time significantly. Having seen glassy eyed folks waiting for their turn in their cars and lines inside the stores, for their early morning cup of kick, there is definitely a sweet spot that customers can live with. The question, that makes this analysis go beyond the realm of rough cut analytics, is how many customers will be willing to absorb the markup and what will be the drivers?

This is the part where AI comes to the rescue! For both scenarios.

We will explore the AI-enabled analysis in the second part of the article. The suggested approach will leverage deep learning in tandem with other AI algorithms to insert a significantly high-level of precision into this analysis. But the more important aspect is that the infrastructure setup for collecting data for this analysis will allow Starbucks to not only get answer to this specific question but continuously enhance services, and explore the possibilities of offering new products and services.

The second part will be published on 05/13.