Subscribe to DSC Newsletter

Thought for Food: Analyzing the performance of a leading restaurant chain

The original piece, "THOUGHT FOR FOOD: Analyzing the performance of a leading restaurant...", appeared originally on the CrowdANALYTIX blog.

The restaurant industry in the US has changed dramatically over the years, and the need to understand and respond to changing customer preferences is more important than ever. As a result, many restaurants have become adept in collecting and tracking sales and customer survey data to optimize pricing, food quality and guest experience. But is that enough?

More and more customers are speaking their minds on social media and other open channels, sharing their preferences directly through a review or implicitly through location-based applications. This information is more valuable than any data the restaurant can collect internally, primarily because it’s unbiased and allows direct comparison with competitors.

However, although the consumers reveal terabytes of data every day, most of what they share is not necessarily actionable. Knowing what to act upon and knowing whether all possible sources of information have been considered isn’t easy. Social media monitoring tools are good at aggregating information in more appealing visuals, but they don’t tell you what strategic decisions should be taken.

Planning the attack

Olive Garden locations across the USA

Here, we present a case where we were able to address the following strategic questions for Olive Garden and potential investors in Darden Restaurants:

  1. Olive Garden experienced 22 consecutive quarters of declining sales since the recession. What caused this and can something be done to reverse the trend?
  2. Are the offerings (food and service) no longer appealing to the consumers?
  3. Is the competition stealing away market share? If yes, how and where?

Our approach included a restaurant level “bottom-up” model where we gathered data across 800+ stores at the ZIP level, covering 200+ factors from various categories, such as:

  • Macroeconomic (e.g Total Per capita income in the past 12 months)
  • Vicinity / Restaurant industry (e.g Coffee shops count within 3 miles)
  • Demographic (e.g Percentage of population earning between USD 25000 and USD 49999)
  • Social / Consumer trends (e.g Consumer review ratings on popular social media networks)
  • Competition (e.g Popularity score of Panera Bread in the vicinity)
  • Weather (e.g Frequency of rain)

Let’s bring in the artillery

Comparing Olive Garden Locations' Performance competition card

We posed the problem to the CrowdANALYTIX solver community by launching a contest called Olive Garden Restaurant Comparison Analysis, where we provided the RAW public data we collected. 262 solvers worked over a period of 3 weeks to submit multiple approaches and derive insights.

One of the many challenges with this data was the large number of factors across very few time intervals. To address this, data scientists had to implement methods to reduce the data dimensionality, such as:

  • Score creation:
    • Housing Costs Weighted by Housing Type (owner/renter)
    • Percent Score of Bad Weather (sum of 5 weather percentages)
    • Income Structure Score (combined 6 income percentage variables into a single score)
  • Scaling the variables, reducing the effect of mismatched data):
    • Standardized Housing Starts region-wise (0-100)
    • Sum of Standardized Frequency of Bad Weather (0-100)
  • Binding some continuous variables into classes to aid in analysis and comparison:
    • Performance of Each Competitor (Very Poor, Poor, Good, Very Good)
    • Percent of Reported BMI > 30 (1-4)
    • Counts of Each Competitor (ex. 0, 1, 2, 3 or more)

So what did we learn?

Transformation of the data and variables in this manner enabled us to uncover robust and valuable business insights. Further using an inductive approach, we analyzed correlations and conducted business sanity checks, whittling down factors for building an “Industry and Macro-economic Model”.

Unfortunately, we can't discuss the final results here, but we can say that the analysis confirmed some long-held hunches and also revealed some irreversible shifts.

 

Views: 1658

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Follow Us

Videos

  • Add Videos
  • View All

Resources

© 2017   Data Science Central   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service