Practices on Acquiring Proprietary Data for Transportation Applications (2019)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

59 Recent technological advancements have led to new types of transportation data. These data often have greater temporal and wider geographical coverage and may contain more details than the traditional data sets (e.g., probe and crowdsourced data). Despite facing challenges associated with obtaining these new proprietary data, state DOTs and MPOs have been using them to meet various needs. This study compiled information on the practices and experience of state DOTs and MPOs on the acquisition and use of proprietary data for transportation applications. Summary of Findings Data and Uses Survey results indicated that DOTs and MPOs have acquired several types of data, such as real-time and historical speed data; O-D data, including truck trip data; bicycle- and pedestrian- count data; crowdsourced incident and jam alerts; socioeconomic data; freight movement data; and digital map and imagery products. These data are used to support a wide range of agency business areas. Among the data reviewed, real-time and historical speed data are the most widely used by transportation agencies around the U.S. for a variety of applications. Agencies employ real-time speed data to support highway operations, including travel-time monitoring, posting alerts on DMSs, 511 services, and incident recovery monitoring. With respect to traffic monitoring, the use of probe-speed data varies among agencies (Athey Creek Consultants 2017). Some agencies obtain such data only for roads that lack sensors or for specific project needs, while others are shifting to probe-speed data for statewide coverage and only deploy sensors where probe data are inadequate. Historical speed data are often used in applications such as performance measures, corridor studies, before-and-after project evaluations, and travel-demand model validation. Highly precise GPS data from in-vehicle systems and mobile phones have found numerous uses in transportation applications. Vendors are processing these data to provide information on O-D pairs—including for trucks—at various spatial and temporal resolutions. These data provide useful information on trip patterns that are not available from traditional data collection methods. However, concerns remain about the potential biases inherent to these data because samples are not randomly selected and can be demographically skewed. These data often require additional staff resources for validation purposes. A growing number of agencies are partnering with Waze under its Connected Citizens Program. These partnerships give agencies access to incident and jam alerts generated by Waze users, which can then be incorporated into the agencies’ traffic monitoring and reporting C H A P T E R 5 Conclusions

60 Practices on Acquiring Proprietary Data for Transportation Applications services. In exchange, agencies send Waze information on special events and planned road work, which are shared with Waze users. Crowdsourced alerts are likely to become an increasingly important part of providing incident awareness. Crowdsourced smartphone applications also benefit from data collection for non-motorized modes of travel, such as for bicycle and pedestrian trips. Several agencies have begun to leverage these emerging data sources to better understand popular routes on networks and factors that affect cyclists’ decision making. The analyses help agencies to identify optimal locations for bike counters and to make informed decisions about infrastructure investment. However, issues found in other data types—such as limited sample sizes, as well as demographic and geographic biases—are also present in the non-motorized data. Many agencies have also procured socioeconomic data, employment data, freight movement data, as well as digital maps and aerial imagery to support transportation applications. These data tend to be licensed from well-established providers. Agency Concerns and Practices Survey respondents and interviewees identified several barriers to and concerns associated with procuring and using proprietary data. They offered reflections on their experiences pro- curing data and shared their perspectives and recommended best practices. Table 8 summarizes these concerns and related agency experiences and practices. Successful Procurement Practices The agency experiences and practices listed in Table 8 address general concerns with regard to proprietary data, while the practices discussed in this section pertain specifically to the pro- curement process. Some practices overlap, but they are discussed in more detail here. Successful practices are summarized in the following categories across different stages in the procurement process. Legislative and Institutional Support • Agencies can take better advantage of emerging data sources and more easily navigate intel- lectual property rights if legislatures revisit and amend existing laws that may restrict or prohibit acquiring or using crowdsourced data collected based on personally identifiable information. • Establish procedures explicitly for proprietary data acquisitions and applications, which should cover data contracting, sharing agreements, and quality-assessment strategies, as well as market evaluation. • Incorporate proprietary data into DBPs as an integral component to fulfill departmental business needs and to fill data gaps. Promote coordination and collaboration among depart- ments within agencies and other state DOTs to make the best use of agency resources and to reduce the cost of proprietary data acquisition, storage, and sharing. • Identify funding sources for data acquisition. If possible, establish a regular budget to maintain data purchases or subscriptions, given that the data meets the agency’s business needs. • Ensure that agency staff have necessary expertise or skills to acquire and work with proprietary data. This may include training, IT, and legal support. Before Issuing the RFP • Establish a workgroup consisting of staff from different offices and divisions within an agency to identify data needs. Determining to what extent data needs overlap should be a key focus of conversations. Forming workgroups is also useful for making different work units

Conclusions 61 Barriers and Concerns Agency Experiences and Practices Data and Service Quality • Include clear data specification in the RFP, including temporal and spatial coverage and sample size requirement. • Request sample data from vendors for evaluation. • Take customer service into consideration during vendor selection. • Include service-level agreement in the RFP and contract; perform regular data-quality audits. • Specify an exit strategy in the contract. Cost • Request clear cost structure from vendors, including future renewal pricing options. • Use a standard cost sheet to facilitate comparisons among vendors and products. • Coordinate with internal units and collaborate with partnering agencies on data licenses to achieve economies of scale; explore pooled-fund options for data procurement. Staff Expertise and IT Resources • Involve agency IT and data analysts in procurement process. • Consider open-source, off-the-shelf tools for data processing needs. Finding the Right Product • Use an RFI to gather information on the current market. • Promote intra- and interagency collaboration to coordinate data needs. • Develop a clear vision for data uses. • Specify broad agency data needs and ask vendors to propose services to meet those needs. • Consider analytical tools in addition to data. Legal Issues (use restriction, non-disclosure, and privacy) • Involve agency legal counsel in contract negotiations. • Specify terms of use in the contract, and try to include all agency internal business areas while balancing costs. • Specify data-sharing plan in the contract. • Specify how to handle open records requests in the contract. Table 8. Identified agency concerns and practices. aware of new data potentially becoming available, which can prompt brainstorming about potential uses. • Circulate an RFI—if possible—to gather information with regard to the data market, including data availability, vendor experiences with other agencies, and pricing and licensing arrangements. For example, the I-95 Corridor Coalition leveraged this approach, issuing two RFIs before it formally initiated procurement. The first RFI described the Coalition’s vision and objectives to solicit feedback from prospective vendors. A second RFI was prepared

62 Practices on Acquiring Proprietary Data for Transportation Applications based on vendor input and sought additional comments from vendors. Circulating two RFIs aided the Coalition’s efforts to develop a targeted and refined RFP. • Make sure in-house data analysts, IT, and legal experts have a role in acquiring data. They provide valuable input on data integration, processing, storage, and management as well as critical legal support to protect the agency’s interest. • Improve collaboration with partner agencies, and explore the viability of pooled-fund acquisition. This may also reduce the cost for each agency because of economies of scale. A number of respondents and interviewees consider this approach beneficial. Two agencies indicated that they have adopted this approach when acquiring certain data items. What to Include in the RFP • Provide clear data specifications, including required data feed format and temporal and spatial resolution. Agencies will benefit from asking prospective vendors to supply data in a format that minimizes the effort they need to expend integrating data with existing software systems. • Use standard cost sheets to simplify price comparisons among vendors. For instance, Missouri DOT not only required vendors to structure their cost proposals by start-up cost and recurring subscription for several predefined data sets, but they also asked vendors to propose per centerline mile cost in case the Missouri DOT decided to only purchase portions of a data set with the budget available to them. • Ask proposers to provide pricing information for different licensing options, such as license for agency internal use only and license for all public agencies statewide. • Use the service-level agreement to ensure data quality. State in the RFP and contract that payment is contingent upon data quality (e.g., availability, accuracy, and latency). I-95 Corridor Coalition, Missouri DOT, and Ohio DOT have such provisions in their contracts. However, this requires agencies to devote resources to routinely audit data. • State data terms of use explicitly (e.g., specify if data will be used for a single or a specific set of applications, for the agency’s internal use only, or for all public agencies in the state). • Specify data-sharing needs. Agencies must be specific about their plans for sharing the data, aggregated data and statistics, reports, visualization, and other derivative works. • Ask vendors to discuss the integration efforts based on the agency’s need. Request vendors’ past work on integration with other public entities. • Outline specific terms of technical service, including customer service response time. Product and Vendor Evaluation • Follow agency guidelines or state regulations, if any, with regard to proposal evaluations. Over two-thirds of the surveyed states have formal guidelines for evaluating vendors and their products. For example, at Minnesota DOT, a panel must be convened to develop RFPs if the total contract value will be $50,000 or more. The agency’s policies enumerate several important factors that should be considered when selecting contractors, including costs, experience and background of both the vendor and its personnel, past work examples, level of understanding with regard to the contract and its specifications, and overall strategy or methodology. • Develop a list of follow-up questions for vendors with regard to their original data sources and methodologies used for processing data. These questions foster transparency and help the agency better understand the strength and weakness of a vendor’s data and approach. • Request sample data, an on-site demonstration, or both. Agreement Negotiation • Specify terms for renewing contracts, and negotiate a renewal cost structure up front. This cost structure may be in the form of total annual cost or maximum rate of increase

Conclusions 63 per year. This step offers clarity on expected future costs, allowing agencies to prepare more accurate budgets. • Work with in-house legal counsel during contract negotiations. The goal is for agencies to ensure full compliance with federal, state, and local laws, especially as it relates to how open records requests for proprietary data will be handled. • Negotiate agreements with private vendors to obtain the most favorable terms possible. • Specify and confirm data-sharing policies. • Articulate an exit strategy clearly within contracts. Areas of Future Research Agencies are likely to face similar challenges during data acquisition, validation, and appli- cation. Survey results show that agencies purchased the same or similar data sets for the same intended uses. Agencies that are new to proprietary data acquisition can learn from early adopters. Hence, communication with peer agencies often proves valuable. Peer exchanges can be an effective approach for interagency information sharing. Efforts at the national level may be needed to develop guidance or standardized processes for proprietary data acquisition, validation, and integration. Areas of future research are identified as follows: • Develop standard proprietary data license models and application guidelines for those commonly used data types. • Investigate unit cost of proprietary data based on past procurement to assist agencies in future decisions on acquiring data. • Develop guidelines and methodologies to help state DOTs and MPOs: (1) validate propri- etary data; and (2) integrate the proprietary data with their own network, such as state DOTs’ linear referencing network and MPOs’ travel-demand model network. • Conduct more analyses on bike and pedestrian data. • Conduct case studies or peer exchange to identify successful practices on proprietary data uses, management, and governance. • Conduct case studies or peer exchanges to evaluate the benefit, challenges, and best practice of forming partnerships among agencies, including state DOTs, MPOs, transit agencies, and local government to pool resources and share data. Today, innovation in the technology sector is transforming the field of transportation. As connected and autonomous vehicles and mobility-on-demand services continue to expand their user bases, the data needs of transportation agencies will continue to evolve. In the meantime, new challenges will certainly surface during the process. Prompted by these proprietary data in large volume, many agencies have begun turning toward big data tools or cloud comput- ing services to handle their data processing needs. As noted in NCHRP Synthesis 508: Data Management and Governance Practices (Gharaibeh et al. 2017), this transformation may create additional uncertainties, such as data security risks.