Driving Sales Revenue using Basket Analysis in Quick Service Restaurants (QSR)

How many times have you rushed to grab a cup of fresh brew after you just got out of a morning flight in an airport, or on your way to your morning flight for that matter? I mean, think about it, there are options all around whether you look left, right, or straight ahead. Agreed that you would most likely have a specific brand in mind; but if for a moment we assumed that you were less brand sensitive and just cared about the coffee you could pretty much pick it up at any of these coffee shops. Even if you were a loyalist, multiple factors such as long queues, and distance from your flight gate, seasonal delights etc., could easily influence your decision.

The point I am trying to make is that in the highly competitive (Quick Service Restaurant) QSR space, particularly for the café/bakery segment, driving footfalls and loyalty is no easy task. So, when I do succeed in getting a customer in the door, how do I maximize my revenue potential through intelligent product offerings?

Let us explore the aforementioned situation with a real case example that we at NLB were recently involved with. One of our clients is a global café chain operating out of most major airports/malls, and many retail stores. The challenge facing our client was that ~40% of their transactions were only single item tickets – I am sure you have already guessed what precious item constituted the majority of those transactions….the quintessential coffee!

Given this backdrop, how were we able to help our client increase individual transaction value and, by consequence, overall revenue? We simply let the Point of Sale data aided by Machine Learning drive us to the actionable insights…

Define the Problem: Our first line of attack was to leverage the fact that a very large number of customers do come in to the store to get coffee. Left to their volition, clearly, most of these customers would not even think about buying anything else simply because in their mind it is only coffee that they need. They are too busy to think further. But it is also true that 60% of customers did buy more than one item. So what could we learn from those customers to entice the other 40%?

Machine Learning Based Solution: We implemented a market basket analysis to understand product affinities. Needless to say, the model readily reconfirmed that coffee was THE strongest anchor; but we were actually interested in knowing what its strongest attachments were? Applying a simple Bayes’ Rule driven confidence, support, and lift methodology, we estimated that croissants was the show stealer: whenever a transaction had coffee on the ticket, the likelihood of croissant being on the same ticket (i.e., probability of attachment) was a staggering 52% compared to a much lower 29% likelihood when coffee was not present on the ticket! Contrast this with mineral water, for example, where the likelihood of the attachment (croissant) being purchased with the anchor (Coffee) vs. without the anchor showed hardly any difference (9.3% vs. 8.9%). It was obvious, therefore, that our client had to aggressively market coffee & croissant together.

Solution & Delivered Impact: The store reps were trained to proactively ask the coffee buyers whether they would like to ‘add’ a croissant to their order; coffee & croissant special combo pricing was also tested. The latter was a particularly successful marketing test increasing single item to multiple item conversions by 12%. To put this in context, a 12% increase translated in to $67 Million increase in bottom-line for this client!

We also performed a similar basket analysis for lunch orders along with a Pareto analysis to identify the 80-20 contributors to lunch revenue. We discovered that certain items on the lunch menu (e.g., barley salad, fruit cocktail, chowder) were seldom ordered. By eliminating the insignificant revenue contributors and aggressively pricing/marketing the high-affinity menu items, our client was able to improve their lunch revenue by 5.3%.

In passing, it is worth noting that human beings are creatures of habit and particularly so when it comes to their food choices at QSRs because one just wants to stick to items that have worked for them before given the lack of a gourmet appeal. Additionally, the high intensity, crowded, and noisy environments characterizing the café QSRs does not really offer an ideal venue for deeper thinking on the menu. In such a situation, a little nudge based on historically proven order patterns can indeed go a long way!

How many times have you rushed to grab a cup of fresh brew after you just got out of a morning flight in an airport, or on your way to your morning flight for that matter? I mean, think about it, there are options all around whether you look left, right, or straight ahead. Agreed that you would most likely have a specific brand in mind; but if for a moment we assumed that you were less brand sensitive and just cared about the coffee you could pretty much pick it up at any of these coffee shops. Even if you were a loyalist, multiple factors such as long queues, and distance from your flight gate, seasonal delights etc., could easily influence your decision.

The point I am trying to make is that in the highly competitive (Quick Service Restaurant) QSR space, particularly for the café/bakery segment, driving footfalls and loyalty is no easy task. So, when I do succeed in getting a customer in the door, how do I maximize my revenue potential through intelligent product offerings?

Let us explore the aforementioned situation with a real case example that we at NLB were recently involved with. One of our clients is a global café chain operating out of most major airports/malls, and many retail stores. The challenge facing our client was that ~40% of their transactions were only single item tickets – I am sure you have already guessed what precious item constituted the majority of those transactions….the quintessential coffee!

Cafe Chain - Case Study

Given this backdrop, how were we able to help our client increase individual transaction value and, by consequence, overall revenue? We simply let the Point of Sale data aided by Machine Learning drive us to the actionable insights…

Define the Problem

Define the Problem: Our first line of attack was to leverage the fact that a very large number of customers do come in to the store to get coffee. Left to their volition, clearly, most of these customers would not even think about buying anything else simply because in their mind it is only coffee that they need. They are too busy to think further. But it is also true that 60% of customers did buy more than one item. So what could we learn from those customers to entice the other 40%?

Machine Learning Based Solution

Machine Learning Based Solution: We implemented a market basket analysis to understand product affinities. Needless to say, the model readily reconfirmed that coffee was THE strongest anchor; but we were actually interested in knowing what its strongest attachments were.

Applying a simple Bayes’ Rule driven confidence, support, and lift methodology, we estimated that croissants was the show stealer: whenever a transaction had coffee on the ticket, the likelihood of croissant being on the same ticket (i.e., probability of attachment) was a staggering 52% compared to a much lower 29% likelihood when coffee was not present on the ticket! Contrast this with mineral water, for example, where the likelihood of the attachment (croissant) being purchased with the anchor (Coffee) vs. without the anchor showed hardly any difference (9.3% vs. 8.9%). It was obvious, therefore, that our client had to aggressively market coffee & croissant together.

Solution & Delivered Impact

Solution & Delivered Impact: The store reps were trained to proactively ask the coffee buyers whether they would like to ‘add’ a croissant to their order; coffee & croissant special combo pricing was also tested. The latter was a particularly successful marketing test increasing single item to multiple item conversions by 12%. To put this in context, a 12% increase translated in to $67 Million increase in bottom-line for this client!

Basket analysis in QSR

We also performed a similar basket analysis for lunch orders along with a Pareto analysis to identify the 80-20 contributors to lunch revenue. We discovered that certain items on the lunch menu (e.g., barley salad, fruit cocktail, chowder) were seldom ordered. By eliminating the insignificant revenue contributors and aggressively pricing/marketing the high-affinity menu items, our client was able to improve their lunch revenue by 5.3%.

In passing, it is worth noting that human beings are creatures of habit and particularly so when it comes to their food choices at QSRs because one just wants to stick to items that have worked for them before given the lack of a gourmet appeal. Additionally, the high intensity, crowded, and noisy environments characterizing the café QSRs does not really offer an ideal venue for deeper thinking on the menu. In such a situation, a little nudge based on historically proven order patterns can indeed go a long way!

About the Author
NLB Analytics Team

carroll.mcintire@nlbservices.com

732-239-7579