Location data analytics has taken on new urgency in the COVID-19 era. This is most obviously true for businesses dependent on in-person activities and foot traffic such as traditional retail. But the ‘new normal’ also includes changes of much wider impact, such as likely permanent shifts in working from home and dramatic acceleration of both pure and mixed e-commerce.
These changes are complex, affecting particular groups of workers and consumers in different ways. For example, service and office workers’ experiences during and after COVID correlate more strongly to the business sector than to ‘collar color.’ These changes are also both rapid and historically unprecedented. This makes them a poor fit for many conventional statistical modeling techniques. Traditional ‘baselines’ at a minimum need to be recalibrated and refactored to consider new cohorts.
But this pandemic is not over yet especially internationally. So rather than to base projections on quarterly, annually or decadal lagged data, best-in-breed enterprises are finding new ways to leverage live geospatial data at scale, and pivot rapidly when new information diverges from prior expectations. They are also empowering their analyst to more easily generate, revise and share scenario-based plans with explicit assumptions.
The Importance of Granular Spatiotemporal Data
We are sometimes so accustomed to working with aggregate data, that we miss important insights. Of course, some datasets are only available or reliable with a certain level of aggregation, and aggregation can be important in privacy protection. Yet the COVID pandemic is nothing if not varying in time and place, and so an important meta-question for every business must be: are we collecting and using data at appropriate scales.
The importance of locational data was actually most famously made clear all the way back in the 19th century when in 1854, a doctor named John Snow manually plotted a geotemporal sequence of cholera cases (see related Wikipedia story). Interestingly, scientific debate about the origins of COVID-19 is also using similar methods today (Bolos 2020). Of course the hand-drawn maps are no longer required, and today’s scientists are coloring their maps based on the gene sequence of the virus mutating over time. But the core method of spatial analysis still applies, and still requires disaggregated data. John Snow would not have gotten far with a map of London census blocks showing cholera cases per decade.
Quantifying New Demographic and Cultural Shifts
The US over the last several decades has seen both intrastate migration to sunbelt cities and pattern shifts within major metros. COVID-19 so far has not appreciably influenced the broader pattern but has flipped the script within metros. Working from home and the continued rise of e-commerce has had enormous implications, not only for offices but also for entire regions. The question impossible to resolve at least before schools restart in the fall is where the ‘new normal’ will settle.
According to Pew Research, about 20% of Americans worked from home before the pandemic and 70% currently do. What will these numbers look like in a year? Well, if employees get their way, 54% will continue to work from home. Employers were not surveyed. But safe to say, the eventual range will be something like 30-50%. That is a shift in home-based work we have not seen since before the industrial revolution. How will this affect your business? Recent data from within the pandemic is muddled, and realistically we won’t in the US have an answer until the new school year settles in. Updating your analytics capabilities now so that you and your staff are trained and prepared would be a wise move.
Which Data Are Available and at What Scales?
Geospatial location-based data is at the heart of our ability to measure and respond effectively to these shifts. The precise data needed by a business will obviously vary. But some general sources are clear.
First, the new census data will roll out late in the fall. This is a “full count” and will be useful especially in capturing large scale demographic shifts. But keep in mind the survey period landed in the midst of the pandemic. If you aren’t already doing so, consider reviewing customer data in this light. At a broad level, are your customers still where you think they are? Has their offline and online behavior shifted to match broader trends, or are your customer cohorts different?
Second, consumer behavior and footfall data is available now. This can come from a variety of commercial vendors, or at county-levels or above from public sources. However, these data will be particularly valuable to look at in the 4th quarter as parents shift to new school and childcare arrangements. These data at disaggregate levels can get big quick. So you will want to make sure your spatial analysis tools can gracefully handle the load.
Third, there has been a lot of churn in real estate over the pandemic. From an internal perspective, this means rebalancing your physical spaces may be important. Market values have shifted, and lease rates will inevitably follow. From a general data analytics point of view, this means your prior location-based datasets are due for a major refresh. Take a look especially at any business points of interest or address datasets. These are likely candidates for at least a refresh, if not a major overhaul.
Intuitive Visual Data Exploration and Collaboration
The accelerated speed with which businesses can explore geospatial data is ushering in new forms of inquiry. Previously, data analytics has been couched in the language of mathematics. But thanks to advanced CPU and GPU processing breakthroughs, analytics has become a dramatically more intuitive pursuit. The blistering speed of geospatial analytics now allows anyone—even the non-technical user—to not only visualize huge datasets, but also to dig into them according to their natural curiosity.
When query response rates are measured in milliseconds, people can focus on exploration and forget about the underlying technology. The exploration of COVID-19 data has allowed people to test hypotheses as quickly as they can be considered. Similarly, business insights can be tested against a variety of conditions to examine their viability and impact.
Think about who in your organization might benefit from improved geospatial tools or training. If you haven’t updated your geospatial IT for a while, today’s “no code” dashboarding environments should come as a pleasant surprise. In particular, an end user not necessarily familiar with spatial analysis or GIS tools can feel comfortable with these newer generation tools in hours and not months. Meanwhile, for your more technical staff, consider what the revolution in ‘notebooks’ and spatial data science means to your organization. Notebooks (such as Jupyter) are a way of mixing scripting languages and interactive graphics that transcend the limits of spreadsheets.
Collaboration is playing an increasingly large role in analytics, and for good reason. So as you are thinking about your analytics, keep in mind how and when these can best be shared among your staff, or with stakeholders or customers. Both dashboards and notebooks have become popular due to their strength in “replicability.” How easy or hard is it for you to adjust an assumption made by a colleague and regenerate a new analysis? Do your current methods work only internally and not for customers? How about on mobile?
The big idea here is that analytics insights are only useful to the extent they can influence outcomes. So it’s great if you have single-user tools which give your analysts deep insights. But much better if they can fluidly share these to accomplish business objectives. This capability supports what analysts call “social speed.” In business contexts, multiple rounds and types of analytics sharing occur. In early stages, there is an emphasis on confirmation and feedback; later, the goal is often to convince peers, stakeholders or customers that a particular approach is valid and has been well considered.
Working at social speed, users can easily create and share annotated dashboards without programming knowledge. Instead of simply sharing a static graphic or table, users can share an interactive interface that includes access to full data where appropriate. This allows the recipient to engage themselves with the idea in question—and build confidence that it is correct.
Broader, Faster, Smarter
As access to these advanced tools becomes broader and more democratic, more uses are available to more people. “No code” analytics makes it easier for large numbers of business and government users to explore critical data.
Machine learning, one of the most exciting developments on the horizon, will help these entities arrive at better answers faster. But the benefits don’t stop there. The combination of granular data collection, particularly from IoT sensors, with machine learning technology will enable computers to present findings that would be simply impossible for humans to accomplish manually.
The digitization of maps at scale has also been useful for many organizations. But until now, those maps have been heavily aggregated and static. Their presentation reflects the presence or absence of an object class over a geographic area with long production lag times. Today we’re seeing fully dynamic, granular maps that include quality information. Instead of a map showing where a forest was five years ago, it can reveal where individual trees are currently, and even if they are under stress or healthy.
This kind of map, called a “digital twin,” is much more actionable. It can provide real-time insight into people movement, environmental changes and, in the case of a public health crisis, infection mitigation efforts. “GeoML” can apply to buildings in cities, individual vehicles on roads, or customers in a retail environment. The applications are endless—and a worthy legacy for what has been one of the most definitive, and consequential, global events of our time.