Google Maps Platform provides developers with business information, ratings, reviews and more for over 200 million businesses and places around the world. Points of interest (POI) data is constantly changing–new businesses open and existing businesses shut down or change their hours of operation. Changes like these can impact not only a business’ customers, but also other businesses relying on them. For instance, a change in the hours of a grocery store will directly impact the ETA commitments of its logistics partners to their end users. These changes may happen more frequently during disruptive events such as a pandemic or a natural disaster. During the COVID-19 pandemic, there was an unprecedented change in business hours and government agency guidance that resulted in frequent and unexpected changes to POI open/closed hours and COVID-related services. In this blog, we will explore key investments we made to keep our POI data fresh during the pandemic and how we continue to build on these investments to deliver up-to-date POI data to our customers globally.
An agile approach to maintaining POI freshness helped us respond rapidly to constant large-scale changes
To help people and businesses get the most up-to-date POI information that kept up with real-world changes during the height of the pandemic, we quickly developed a model that looked at various signals to detect regional changes (such as a government mandates for mask requirements) affecting all POIs in an area, as well as POI-specific changes. We used machine learning and evaluated signals collected through imagery, AI, and user input.
For example, we looked at the attribute describing whether a business required customers to wear masks. During March 2021, this was usually true for dining places in NYC. However, when the restriction was lifted in mid-May 2021, fewer businesses required masks during the summer months. Based on user input, our model detected the lift of the mask mandate. We also used Duplex conversational technology in Argentina, Australia, Chile, France, Japan, Mexico, New Zealand, Peru, and the United States to call businesses to confirm their information such as opening hours or whether they offer takeout and delivery. This enabled us to quickly scale our efforts to provide the most up-to-date POI information to our businesses and consumers. Another example is how we used machine learning to automatically determine if business hours were inaccurate, and instantly updated them with AI-generated predictions.
Our approach ensured that businesses using Google Maps Platform to develop and deliver critical services during the pandemic had access to up-to-date POI data on whether a place was open, closed, or temporarily shut down–and its updated hours. For instance, access to updated hours for grocery stores and restaurants allowed delivery companies to provide more accurate ETAs to customers on their delivery orders. Local governments and organizations were able to provide up-to-date information on hours of operation for hospitals and vaccine centers–with resources like the Japan Vaccine Locator. Companies like Costa Coffee were able to aggregate and provide the operational status of their local stores (constantly changing driven by regional mandates) to customers. And several national chains were able to use their aggregated operational status data to define their regional and national operations strategy.
Rigorous data quality assessment set the foundation for a strong POI freshness machine
Over the years, we have developed several mechanisms to measure and improve the quality of our POI data. For instance, we look at data points such as existence (whether a business is open/closed or temporarily closed) and hours, latency for how soon real world changes are reflected in our data, and POI coverage for how quickly we add newly opened businesses to our data. These areas and several others are part of a rigorous framework to measure POI freshness. They help us identify areas that need attention and deliver rapid quality improvements to our customers. Efforts that helped us adapt during the pandemic are also helping us improve our systems over the long run. For instance, our efforts to measure business changes during the pandemic have enabled us to reflect real world changes on the map in record time.
For a certain POI, we often receive multiple edits from various sources and sometimes these edits contradict each other. Over the course of the pandemic, people have submitted millions of temporary closure and reopen reports. Google has invested in building advanced systems to synthesize and corroborate human intelligence, user feedback, third party data, and imagery evidence in order to deliver the most trusted information possible to our users. These investments enable us to reconcile edits from various sources.
We continue to build on our efforts during the pandemic and are making targeted investments to improve POI freshness across regions. Besides tailoring our ML algorithms by country and type of business, we also frequently team up with regional experts to add locally relevant POI attributes. Besides benefiting our customers in the region, these efforts have been essential for some of our global customers entering new markets, and have helped them improve customer satisfaction while innovating rapidly as they build on a platform with extensive global coverage.
We are constantly launching new products and features to help businesses seamlessly share information with their customers. Read more about the new attributes and Places details, such as special hours, editorial summaries, and ability to sort reviews, which are now available on Google Maps Platform.
For more information on Google Maps Platform,visit our website.
By: Bonnie Pericolosi (Product Management Director)
Source: Google Cloud Blog