A couple of weeks ago, I went to my favorite museum in the Boston area to see a new exhibition. While I was there, I noticed a lot of visitors walking through the museum using self-guided tours on handheld devices. I knew that the museum rented such devices to visitors, but I hadn’t thought about how the museum could use these devices to gather information about visitor behavior. Because I’ve been working on HP Vertica’s geospatial functionality, HP Vertica Place, I wondered what kind of insight the museum might be able to gather from their visitors’ spatial relationship with the museum and its artwork.
HP Vertica Place leverages the performance and scale of the HP Vertica Analytics Platform and uses OGC-based SQL functions for computation on two-dimensional planar data. So, I know that HP Vertica Place can help me answer a number of questions regarding point-in-polygon, distance, and intersection to name a few.
To test my hypothesis I needed a dataset. Because I couldn’t find one publicly available, I created my own dataset using a Python script.
Instead of trying to replicate the movement of thousands of people moving through a space with more than 500,000 square feet, I took a more micro approach. Museums are normally compartmentalized into rooms, so I choose to focus on a single room of a museum. In that room, I added six different works of art for visitors to view. Visitors could view the artworks in differing order and could also view the works of art for a varying amount of time. All visitors to the room are assumed to be using a handheld device providing them additional information about the artwork. The handheld device records each visitor’s location data every second. This information is then downloaded from the device after the visitor returns it and is then uploaded to an HP Vertica database.
Using the script I created, I generated a dataset of 2,000 visitors moving through the room from 9:00 am to 6:00 pm.
This graphic shows how the room was laid out:
I then proposed four questions that can provide valuable information about the collection and how visitors engaged with the space:
- Which work of art was the most popular?
- How many people interacted with each artwork?
- On average, how much time does a visitor spend viewing an artwork?
- At the busiest times of day, how physically close are visitors to one another?
To find the answers to these questions I queried my database of location data and polygons representing the viewing area of each artwork.
While working through the first question, I discovered that I could use HP Vertica Place to write one query to answer my first two questions. Here’s the query I wrote:
From the results it’s clear that artwork number 4 is the most popular, but 2 and 3 are in close competition. However, artwork 1 isn’t drawing nearly as many visitors as we’d expect. Why? Is it not properly marketed? Or is the location a problem? The results of this query could help us address those types of questions.
Curators across the country would be thrilled to know on average how long visitors spent in the viewing area of each artwork in their collection. I used this query to calculate the average time spent viewing an artwork:
I know that the handheld device records the visitor’s location each second, so I can infer the amount of time the average person intersects with the viewing area of an artwork. From this inference, I can make an accurate estimation about the length of time each visitor viewed a work.
The fourth question relates to my most common gripe when visiting a popular exhibition or artwork: how close are people to standing to one another during the busiest time of day? For this query, I decided to look at artwork number 4 because it was the most popular in our previous query:
On average during the busiest time of day, visitors viewing the most popular artwork are standing 2.2 feet apart from one another! This type of insight is fascinating. What if this artwork was placed in a larger room with a wider viewing angle? Would more people be able to get a more direct view of the artwork? These questions are difficult to answer without this type of data and analysis, but think about how having this data could improve the overall visitor experience.
Imagine that you’re a museum curator and learn that your most popular artwork isn’t the Monet that your members continually rave about in their visitor surveys. Instead, it’s actually the Renoir across from the Monet that generates the most traffic. Or, you discover that the buzz of a high-profile auction drove more traffic to your three Giacometti statues than the email marketing campaign you did the month prior.
This type of data is available in many different types of business scenarios, not just museums. Location data provides valuable insights into how people interact with spaces. HP Vertica can help you discover these insights.
Do you want to try out this example on your own? Install HP Vertica 7.1.x and the HP Vertica Place package from your my.vertica page. Then, download the dataset and accompanying SQL file from our GitHub repository.