Up-to-date data for measuring economic conditions: the Louisville Restaurant Status Check

Amy Deora
City as a Service
Published in
6 min readMay 13, 2020

Local governments across the country are looking for ways to understand what the economic fallout from COVID-19 means for their communities, and how to best target resources that will help in recovery. In the following post, we share a potential framework for developing real-time data assets for cities, and introduce an open source library that can help local governments conduct a simple “health check” of their local businesses. Using this framework, we were able to gather current operating information on nearly 400 businesses in an afternoon.

Our goals for this project

At Stae, we help local government combine administrative data sources with open, crowdsourced, and vendor data to better understand their communities, and improve operations. In more “normal” times, there is a wealth of available data to help organizations understand social and economic conditions. But this crisis presents a new challenge: because the world has changed and continues to change so quickly, data from last quarter or last year just isn’t relevant anymore — we need regularly updated data, and we need it faster.

So we at Stae have set out to find sources of data that reflect the economic and social health of our communities, and are nearly as timely as the sensor data we often work with from scooters and streetlights. To that end, we conducted an experiment to see if we could create a timely data asset that might provide insight into impacts on one type of business (restaurants), in one particular community (Louisville, Kentucky — one of our favorite partners). Our goal was to provide a “real-time” dataset that could help Louisville understand which restaurants, in which neighborhoods, are weathering the storm with only takeout services, and to also track whether restaurants are opening back up when allowed.

This is still a work in progress, but through this exercise, we tested out a framework to help cities collect time-sensitive data on key indicators, empowering them to take action quickly:

  1. First, start with traditional sources of city administrative data.
  2. Then, enrich those data with crowdsourced data collected in the community; and
  3. Finally, use an inexpensive method of primary data collection to fill in the blanks.

Here’s how we used this framework to better understand the status of the restaurant industry in Louisville:

First, start with what you already know — City Data on Business Licenses and Health Inspections

Our first step was to use data the Metro government already had on restaurants in operation — in this case from data on health inspections, which are required of all food service businesses. These data are available as open data on the Metro Louisville Open Data site. These data provided an authoritative list of the approximately 2,300 restaurants operating in Louisville.

Then, layer in crowdsourced data from the community

Next, we added in data from a crowdsourced resource. Lots of local organizations are developing great resources to help call attention to what businesses are still open, and what their hours are. Louisville Food and Dining, in partnership with Louisville Tourism collects data via email from businesses and community members on who is open, and who is not. These data are presented on the Food and Dining website in a format that is great for residents looking for a meal, but not in a format that allows us to map the data, look at trends over time, or understand this sample of restaurants in the context of the local restaurant industry.

To do this, we needed to merge the Louisville Food and Dining information with the Louisville food inspection data, and format the data for easy analysis. To do this, we used the Importer tool on the Stae Civic Intelligence Platform. Through the utilities within the importer, we were able to link the two data sets without any sort of ID, using “fuzzy” matching techniques to match restaurants by name and address. We were also able to clean the data, and geocode it by address.

Using the customizable Code Importer tool to merging business data with crowdsourced data on the operating status of local restaurants

Now we have one data set with one row for each restaurant in Louisville, merged with information on 225 restaurants known to be operating, based on the Food and Dining crowdsourced site. And, because of the build in geocoding feature of the Stae Civic Intelligence Platform we can view the restaurants on a map to see what neighborhood they are in.

Map of Restaurants reporting they are open for takeout in Louisville.

Finally, fill in the gaps with a phone survey

Next, we set out to collect even more information about the status of restaurants, as unobtrusively, and as cheaply as possible. To do this, we created a one-question phone survey that restaurants could answer in 15 seconds — asking them to indicate whether their business was open with normal hours, temporarily closed, permanently closed, or open with limited hours.

To field our survey, we needed phone numbers. To get those, we combined the business license data (again, from Louisville open data) with our dataset. To capture results, we developed a new feature on our platform that uses the Twilio API to make the phone calls and feed results back into the Civic Intelligence Platform. This allowed us to field the survey in minutes, and analyze results immediately. We’ve open sourced this project so any of our customers (or anyone else) can perform their own surveys.

We called restaurants at a variety of times of day, to ensure we’d capture businesses open at different times. We received 163 responses to add to our dataset. As a side benefit, we were also able to flag invalid or disconnected numbers in the City business license database, and while not picking up the phone is not proof of closure, it can be an interesting data point in itself.

The results!

Now we have a centralized dataset that includes all restaurants operating in Louisville prior to the pandemic, merged with information we were able to collect in a single day on 395 restaurant establishments via community crowdsourced resources and a 15-second phone survey. We can visualize these data on a map, and within the context of business and other civic data in Louisville. We also have a working data pipeline that will allow us to easily update this dataset over time, allowing us to look at trends in restaurant closures and/or re-opening over time.

While this is by no means an authoritative source of data on what’s open and what’s not in Louisville, we found that most restaurants who responded to the survey are open with service modifications. Only 19 restaurants (approximately 5% of our sample) reported a complete closure.

Looking specifically at neighborhood-level impacts, we looked at two key commercial areas in Louisville that are home to the highest numbers of restaurants — the Bardstown Road corridor, and the Central Business District (downtown). We have affirmative “open” crowdsourced/survey response information from nearly half of Bardstown Road restaurants, but fewer than 10% of CBD restaurants. This suggests that a higher proportion of downtown restaurants may be temporarily closed during the pandemic.

We see more restaurants currently reporting as “open” in the Bardstown Road corridor (in red) than we do in the Central Business District (CBD, in green) , even though the CBD is home to many more restaurants.

Next Steps

We at Stae love to help cities leverage community-collected data, and are excited to work with the wealth of new crowdsourced data resources communities are developing. We’re also excited to build out our new data collection feature. Although we used a phone survey in this example, we can also use the same method to implement rapid email and text surveys as well (in line with rules around automated calling).

If you want to see if your favorite restaurant in Louisville is open, explore our final data set here, and reach out if you’d like to learn more about creating real-time sources of data for your community.

--

--