Broad Spectrum – finding your place in the world of analytics

Last week I had the opportunity to interact with a highly intelligent, dynamic group of data and analytics practitioners. Initially I felt a little out of place –the technical acumen and brain power in the room was impressive. The meeting was a “visualization potluck” in which attendees were asked to bring a visualization and present it to the group. After a series of very well thought out presentations and much discussion of tools and methods the group leader asked if there were any other presentations from the group.

I meagerly raised my hand and asked if I could share what would be the final visualization of the meeting. To break the ice while introducing myself I made a joke about the fact that I felt like the majority of the group was in the deep end of the analytics pool while I was at the house next door in the kiddie pool – silly joke, but it went over well with the crowd which helped me relax before continuing with my presentation.

I realized later that the joke was more than a light-hearted way to introduce myself – it was, in some ways, a description of the spectrum of data professionals. As data science and the practice of analytics continues to develop and mature, one thing that has emerged with clarity is that analytics practitioners fall somewhere on a spectrum with data scientists at one end and data-savvy business analysts at the other.

The type work being done at each end of the spectrum is going to vary greatly as will the way in which the work is done. In short, the data scientists will work to solve very complex issues, often in a fairly independent setting, and the algorithms built by data scientists could take months, or years to complete. A data scientist is likely working in a research-type role, relying on deep technical and statistical expertise, to answer open ended questions. The business analyst, on the other hand, is going to work on short term projects for days or weeks. The analyst is usually going to be working in a more collaborative, or team setting using their business acumen coupled with data to solve operational problems.

These thoughts on data roles aren’t necessarily new or original, but the visualization potluck really drove home for me what each end of the spectrum looks like and where I desire to be on the spectrum (hint: I won’t be pursuing a Ph.D. anytime soon).

Whether you are trying to find the right data-talent for your organization, or you aspire to develop into a data professional it is important to be mindful of the spectrum of roles and the expectations associated with each end. Hiring the wrong type of talent, or pursuing the wrong type of training for your goals, skills, and aspirations is a no-win proposition.

As for my visualization; I presented the histogram of restaurant score distribution that was in my previous article and briefly recapped the points outlined in the article. The room was very receptive to my graphic and the story the data told, validating my place at the table as a data-curious business analyst.


Grading the grades – a data driven look at Restaurant Food Safety Scores

The food scene in Charleston is unquestionably one of my favorite parts of living here. Charleston has high quality restaurants of all types and for all budgets. Despite the nuances that make each restaurant unique, they all have something in common: a food safety grade assigned by a DHEC inspector on display in the restaurant showing an “A, B,” or “C” grade.

During a recent dining experience in a city outside of Charleston, my wife found a hair in her food. The restaurant in question had their ‘A’ food safety rating card properly, and perhaps proudly displayed, but that did little to console my wife who promptly put down her fork. I typically don’t pay much attention to the ratings, but the ensuing conversation about restaurant cleanliness and food safety ratings inspired me do a little research on what goes into a DHEC grade and frankly, DHEC may have a quality control problem.

According to the SC DHEC website, restaurants are subject to unannounced routine and/or follow-up compliance inspections. During the inspection points are assessed for violations that are uncovered, totaled, and a score is assigned; 0-12 points is an ‘A,’ 13-22 points a ‘B,’ 23-30 a ‘C,’ and anything over 30 points is presumed to shut down the restaurant. Points are assigned at the discretion of the inspector and there are a number of violations that can be corrected during the inspection which then reduces the final point-value of the violation. My first thought is that this score range seems to be quite wide, potentially masking serious violations behind an ‘A’ rating displayed in the window of your favorite eatery.

The data available through the DHEC website is admittedly limited, but what is available is suspicious. Of 1,452 routine (unannounced) restaurant inspections recorded between 2013 and 2015, 97% of the restaurants inspected received an ‘A’ rating, the remaining 3% received a ‘B’ which was the lowest grade in the dataset.

The histogram below shows how scores are distributed for the available data. I expected the data to be skewed, but I was shocked to find that the most frequently awarded score was 0 (indicating no violations were found) given that the inspection report has 56 sections to score and each of the 56 sections has multiple applicable codes – there are literally hundreds of possible violations that could be found during an inspection and yet 12.5% of the routine inspection yielded not one violation!


The other component of this data that is of particular interest to me is the difference between the number of restaurants that score 12 and the number of restaurants that score 13 – keep in mind this is the difference between an ‘A’ and a ‘B’ rating for the restaurant. The number of restaurants that score 12 points (just enough for an ‘A’) is 17 times higher than those that score 13, or rate a ‘B’. That so many restaurants would do just enough to score an ‘A’ and that so few would fall just one point short seems fishy. Given the perception around restaurant grades, a ‘B’ grade at a restaurant could be cause for a major loss of business if word got out about anything less than an ‘A.’ It makes sense then that an inspector might provide enough leniency to keep a restaurant in the ‘A’ category. In fact, I’m reminded of the time in high school that I managed to get a lucky break and catch a bump from B+ to A- based on (non-graded) class participation. It wasn’t a score warranted strictly by the points I scored, but based on the discretion and leniency of the teacher.

However, this article isn’t about getting a grade I didn’t deserve; it’s about food safety and the reliability of restaurant safety grades. The DHEC materials make it clear that inspectors are allowed to use discretion during the inspection, but the data suggests that inspectors may be too lenient on restaurant grading.

There are a number of things that could be done to address quality issues in the inspection process:

  1. Review score distribution by inspector; this distribution might indicate training issues (inspectors consistently giving very low scores indicating few or no violations) or too much leniency (inspectors that have given out many scores of 12, but few or no scores of 13).
  2. Have two inspectors present for every inspection and have them look for different things. They can tabulate the total score at the end and neither would know if a restaurant was on the cusp between letter grades.
  3. Update the scoring system; given the fairly wide range of scores between the grades you could update the points system, or narrow the grade ranges. In reviewing the report sheet, it seems to me that the point values assigned for each violation seem quite generous – a restaurant could theoretically rack up a number of violations and maintain a score less than 12, allowing the restaurant to keep an ‘A’ rating.

The data found on the DHEC website indicates that DHEC may have a quality control issue among its inspectors – quality issues that could put the public at unnecessary risk.

There is little evidence that food borne illnesses are a big problem in Charleston – anecdotally I can’t ever recall myself, or any of my friends dealing with food sickness after a meal in Charleston. In light of the data, I believe this is likely a result of Charleston restaurants making an effort to keep diners safe in spite of DHEC inspections, not because of them.

Analytics on the Battlefield: The Start of a Years Long Journey

Consider this my “Hello World!” post. This article was originally published on my linkedin profile, but in considering what this blog is about, where to start, etc., I felt like it would be a great jumping off point. Thanks for reading, I hope you enjoy!

Afghanistan was the very last place I expected to have an epiphany about the power of data, but it was this unlikeliest of locations where I realized data could be used in a profound way.

In 2010 I was deployed to Afghanistan as part of a Route Clearance Company that was responsible for patrolling the roads around Kandahar for mines and Improvised Explosive Devices (IEDs). As the Operations Officer for the company, one of my primary tasks was route planning and mission development. After the first few weeks in theater, I began to take a hard look at the routes we were patrolling, taking into consideration the volume of traffic on the routes and history of IED events on each route. Unbeknownst to myself at the time, I was building a heat map of traffic patterns and IED events.

What I found is that IED events were typically occurring in and around the same places; there was little variation where IEDs were being found. By using and refining the heat map, I was able to work with the company staff to plan missions that minimized time on routes with little traffic, no tactical importance, or no history of IED events, and instead had the unit maximize time in the areas where there was most likely to be an IED.

This approach proved to be wildly successful and I credit this use of data both as my personal “A-Ha!” moment about the power of data and as a significant reason why our unit had a successful deployment. My unit cleared nearly 15,000km and most importantly, brought everyone home alive. I was awarded a Bronze Star at the end of the deployment which was due in no small part to my use of data-driven mission planning and mission optimization.

Now, nearly 5 years removed from my deployment, I’m ready to take the next step in learning how to use data to drive strategic business decisions – I’m proud to announce that I will be starting graduate work in Business Intelligence and Business Analytics.

Starting classes doesn’t represent the end of my journey – merely the next phase – and I am excited for what it means to be going back to school, and for formalizing skills in an area I have been interested in for a long time.