i would need an internal organization analysis on SafeGraph

Strengths Weaknesses Value Chain or Resource Based View analysis Sources of competitive advantage (if any). conclusion about the company's overall competitive position in the industry: does it have a competitive advantage over its rivals or not?

Launching SafeCraph Hoffman and Perez launched SafeGraph in San Francisco, CA in 2016. While a variety of companies existed that leveraged a company's own data to provide solutions, applications, or analytics (cg, Snowake, Databricks), SafeGraph would purely source and sell external data. In so doing, the company would undertake three basic activities: Take in data from a variety of sources; use machine learning to fill in holes or remediate duplications; and deliver that data in a userafriendly way. The duo spent a few months honing in on what type of data they would sell, steering clear of personally identifiable information (P11) since it was an area of increasing scrutiny over privacy protections. Hoffman recalled, \"We considered several candidates: data generated through the Ad Tech system; anonymized CPS-level data from smart phones; and healthcare data. We rejected getting involved in people data.\" The team decided to pursue data related to physical places because they believed it would be increasingly important as technology companies intersected with the physical world. " At the end of the day, we are (3152' nerds, so this made sense to us," Hoffman recalled. The cofounders' initial strategy was to target a niche market and become dominant quickly, later expand to adjacent markets, price aggressively, and access customer relationships through acquisitions. SafeGraph raised a $16 million Series A round of private financing in 2017. Early investor and Managing Partner of Ridge Ventures Alex Rosen commented on his view of the market: \"There's lots of not very valuable data businesses. And there's a handful of very valuable data businesses, most of which fly under the radar. I love that Data as a Service businesses aren't in the limelight, It means fewer competitors in the space and less investors in the industry. But if you look, there have been multiple multibillion dollar businesses, built across different areas in data. Think of a company like Dun 1: Bradstreet or others." He continued, "While there are plenty of valuable small niche software companies, there are no valuable niche data businesses. You either get big or you're irrelevant. But there's a huge amount of work building a data company of scale and not many companies can or are willing to invest in that. Those that do will have what I like to think of as an execution advantage. If anyone can do this, it's Auren." According to early data scientist hire Ryan Fox Squire, "Auren had a strong recruiting muscle. Given his industry connections and work at LiveRamp, he had a great reputation and was able to bring in over a dozen employees, mostly engineers, in the first few months." Hoffman reected, "Relative to the highflying Silicon Valley SaaS businesses all around us, selling data was less overtly dazzling. Working at a data company is like being an unsexy archivist at the Library of Congress. You know your job is important, but you also know it is a supporting role that helps other people shine." Hoffman favored technical talent and high energy over pedigree. The majority of the early team were engineers and product specialists that hailed from other data and tech startups. Evolution of the Product SafeGraph initially focused on selling anonymizecl, aggregated device-level GPS data it procured from application companies that had opt-in location tracking information from mobile phones. These app companies were those wllere location was core to their product. for instance, exercise. Such companieswere thought to produce more accurate data because they were drawn from true GPS points (versus poor quality data derived from apps that used cell-tower lookups for genlocating users, a phone-haery-inefficient process often disabled by users). Early employees spent the majority of their time reaching out to create direct partnerships with these application companies, securing data sources guided by the ethos of: \"more supply, is better supply." SafeCraph's first product, Movellient, was a time series of longitude and latitude coordinates that represented the movement of people at places in the U5, and Canada where "people spent time or money" (leaving out residential locations). Ninetyrflve percent of the initial customers were players in the Ad Tech space who wanted to use the data to improve their locationrbased marketing. Squire commented, \"We got some good traction, but it became clear that our customers wanted to understand more context around the data beyondjust the coordinates." The data's inherent limitation was the user's lack of familiarity with what Movement coordinates actually represented in the physical world. For instance, what good would it be to know if a device visited the location represented by the coordinates 78 .570868,35,864862 on a daily basis at 4:00pm, if one did not know that the string of numbers represented the Starbucks at Triangle Tewne Mall in Raleigh, NC? Such a consumer that habitually enjoyed an afternoon caffeine break might respond to a certain type of advertisement from a retailer at the mall. The SafeCrraph team thus recognized an opportunity to provide data at the level of a point of interest (POI) such as a physical place (store, restaurant, or park) as opposed to merely at the devicerlevel, In its quest to purchase a POI data set to marry to the Movement data, Sachraph executives encountered limitations. The team evaluated dozens of companies that sold data about physical places, but did not feel that available data sets were more than 50% accurate. "The existing set of P01 data vendors were selling into an older marketing use case where 50% accuracy could perform really well. But we wanted to source data for data science and machine learning so for these purposes, underlying data needed to be at least 95% accurate otherwise errors would compound really fast,\" noted Hoffman. Moreover, many existing sources for business listings, such as Dun S: Bradstreet's 9rdigit DUNS company identifier, did not contain gcospatial information, A DUNS entry might be: 783266950, which represented Home Depot's Office in Atlanta, GA. Such listings often placed several companies at the same location (eg, inultiplebusinesses in a building with different suite numbers or a Starbucks Within a Target). Other business listing vendors might have slightly different names, spellings, or address format (e.g., RF, Chang's, at 740 5. Mill Ave in Tempe, AZ might appear as PF Changs, or PF Chang's Corp. with the street address possibly spelling out the abbreviations as South and Avenue, The company realized it needed to build its own solution to answer the seemingly straightforward, but in fact quite complex questions: "What is the business and allure exactly is it located?\" In early 2018, it hired machine learning specialists to infer the shape of buildings from satellite imagery, city records, and building draWings. Additionally, it crawled for data from thousands of open sources on the internet. The resulting output was dubbed Places and it was used internally to tie the CPS movement data to business listings in a new product called Visits. This product enabled customers to gain insights about foot traffic at commercial sites that would inform decisions on how, when and to Whom to serve up advertisements. For instance, Visits data could help a sneaker company target advertisements to people who recently visited a sporting goods store. The SafeCraph team built the Visits data brands by-brand, At first it took about one month to create a Visits data set for a brand, a process which grew exponentially faster as internal expertise grew, The colnpany published a Zirpage technical white paper to describe how Visits was developed, Accordlng to Hoffman, listening to early customer feedback and staying nimble was key to early product development. While Places had been created for internal use as an input to the. Visits product, he saw an opportunity to sell it in a meeting with a significanttelecommunications prospect, VP Sales '3: Customer Success Jason Cook recalled, "In true CEO fashion, Auren heard they were doing a bake off and decided to throw our hat in the ring. We ended up winning and then had to scramble to tackle all the engineering required to make the backend work for Places as a standalone prod uct," The 2018 Final: Prom DeviceLevel m Place-Level Data Over the course of 2017 and 2018, the company continued to fincl steady success selling its Movement and Visits data to early adopters in the Ad Tech and retail sectors. However, Hoffman was mindful of how difficult it was tor his team to source reliable places data, and saw an opportunity to fix this problem for other data usersi Cook likened device-level data to the "water that flows through ones plumbing, The homeowner likely does not care where the water supply comes from as water is water.\" He continued, " We want to instead be the pipes in your home. Once you have the pipes in place. there's very little reason to replace them and its labor intensive to do so even if you wanted\" Hoffman concurred, \"Geospatial data i' similarly foundational to an organization, so we decided to focus on being really good at one thing, places data." Thus the decision was made to spin out the device-level Movement product to a new San Francisco-based company, Veraset, which would continue to operate independently and in turn become a data vendor to SafeGraph. The Places product was renamed Care Plums and it eventually grew to Coinprise listings at nearly 7 million places. in the U.S. and Canada, representing over 5,500 brands and 3 million nlom-anci-pop stores. [l omitted extraneous locations such as ATM machines, PO boxes, or home-based busine, es. The Places product was delivered to customers as a series of rows (that defined the business entity location) and columns (that described attributes related to that entity that might include: address, phone number, operating hours, business category based on North American Information Classification System (NAICS) code. Accordingly, row data tended to remain stable as stores infrequently moved, while some column data changed daily. Great care was taken by the product team to determine the column labelling and formatting for units of measure such that they were consistent and easily understood by users. For example, one might define visit frequency by day, week, or month, so it was important to define the unit of time by labelling a column: visitsibyiclay. Cook recalled, "ll was great. Companies that used to ask us, 'What are we supposed to do with these millions of rows of lat and longs?' now have criiitrrt about the data," Data panels were delivered through a monthly bulk downloadable CSV file. (See Exhibit 2 for sample data feed.) As SafeGraph's Ad Tech business was solid and growing, the company saw opportunity to expand to new sectors. Yet feedback from prospective clients revealed that the large tile size or the Visits data was proving daunting and their internal teams sometimes lacked the sophistication to make sense of even the contextualized data. Thus, the company developed a new sub-product clubbed Places Patterns, which aggregated the foot traffic data in a more user-friendly lorrnat and smaller tile size, (An average file size for a week of Patterns data was 1.2 GB compared to 45.5 GB for a week of Visits data). Patterns data provided quantitative information such as how many devices entered and exited a store, or how long the device remained in the store, it also supplied qualitative data such as a list of numerical codes representing the home neighborhood of the POI visitor (these digits represented the "census block group" (CBC) codes, the smallest unit of population grouping reported on by the US. Census Bureau), One could overlay information about the demographics of a car; to understand age, education level, o income level. Another example was data on other brands that device visited on the same day as its visit to a given POI. SafeCraph's customers were able to perform a range of analysis, and could also lay on other data such as weather, to see how rain impacted purchases. Another subproduct developed was Geometry, which defined the actual physical shape of the place by using spatial hierarchy inetadata. Such information would reveal when a POl was located within a l'Ol, for example a clothing store located Within a shopping mall. Product Roadmap Looking ahead, Lauren Spiegel, VP Product, saw opportunities to expand SafeGraph's offering by adding rows to include new categories of business entities such as industrial parks, warehouses, or offices, or even international locations. Spiegel often participated in customer Feedback calls and one such request involved adding cemeteries to the available rows, she commented, "Our internal systems make it fairly easy to add these new types of places.\" Other opportunities stemmed from adding additional attributes and columns. Yet, since SafeGraph delivered its data through periodic files, their ability to glean insights from how users applied or interacted with the data was stymied. Spiegel hoped to address this by launching an API such that she and her team would View frequency and other attributes of AH queries. The Plircckey Standard and the Gcnspaiinl Enthusiast Community While SafeGrapli Places data filled the need for reliable POI and aggregated Visit data, joining data sets from multiple vendors without a standard join key remained an issue in the industry. Along with roughly a dozen companies committed to universal access to data, SafEGroph launched a free universal identifier for physical places called Plurekrzy. SafeGraph's commitment to Placekcy furthered its mission to democratize access to data for all as it enabled its own customers to glean even more context by expanding thc network oi datasets to which SafeGraph data could be joined. The intention of the founding Plucekcy partners was to use the lsrdigit Plucekey as a standard Way of naming a POI at a certain location thereby making the sharing oi data about that place caster, For example, the Philz Coffee at 1775 17'h Street, San Francisco, CA could be represented as: 232-222@5vg-7gt-5lf. The first six digits identified the \"what\" of the location whereby the first three digits told the address encoding, the second three digits told the POI information (such as business name). The last nine digits represented the \"where" of the business based on transforming longitude and latitude information to the open source hexagonal grid system developed by Uber known as H3, These grids were more useful measures of location than other commonly used zone characteristics such as zip codes or neighborhood zone districts which were arbitrary and could be subject to change. The resulting Placckey told the "What and where" of a business and acted as a so-ealled \"join key" that enabled easy joining with myriad other data sets. AP] access allowed interested users to query the Plucekcy of a given business. In 2020, SateGraph began to append the Plucekey to its POI data. Within a year, several hundred geospalial enthusiasts communicated on a Plucekcy Community Slack group to share research, generate ideas, and collaborate on projects. Marketing and Sales The data industry was crowded with a multitude of players ranging from the tech behemoth Google in small data startups that flew under the radar. Each ofSafeGraph's products faced competition from a different set of companies. Direct competition often came from Sears companies with a small data offering. According to SafeCrraph, these companies tended to put less focus on their data product, resulting in lower accuracy and less frequent updates. Indirect competition stemmed from noti- geospatial data providers like creditcard companies that sold piles of transaction data, used to identify trends in consumer sentiment and behavior. Yet those data points only told one part of the story and lacked information about a consumer's physical movements surrounding those purchases. \"In reality, credit card data is complementary to places data, and companies get the most value from using them together," Hoffman observed. (See Exhibit 3 for chart of Coospatial Data Ecosystem.) Spiegel identified SafeGra ph's unique value proposition, \"Our data is easy to work with. Our users understand how it is organized and labelled based on the format we deliver it in. We build incredible trust based on the accuracy of our data and prospective clients appreciate that when they compare us to other vendors." Cook continued, We have two advantages over our competition. First we are a datarfirst and dataronly company whereas data is at best secondary or tertiary priority to most of our competitors. So, we are focused, understand the science, and can invest in Significantly improving the data as a product. Second, since we are a tech startup, we have a high velocity of internal communication by virtue of our Slack channels that connect our Sales, Product, and Customer Success teams. We move fast to respond to our customers. We also release our data monthly, while our competitors usually release every 3 or 6 months. The early market for SafeGraph's Movement data was Ad Tech, where customers wellrunderstood the value proposition and Hoffman's close relationships smoothed customer communication. When SafeGraph later spun out Vet-riser, SafeGrtiph was able to market Places to the same customers in the Ad Tech space as many of them had appetite for that data as well. But before the company put further efforts to sell into new verticals, it pushed to find ptirt'i'terships with ecosystem players such as consultants and analytics companies that supported brands. These partners would educate the brands on how to perform rich analysis tin the data to info rm husine ss decisions. One such channel partnership was struck with a GlSsoftware company, Esri, which sold small sets of SafeGi-aph's data to cumplement its own offering. Cook described SafeGraph's \"land and expand" marketing strategy whereby once a market was established they would focus on extending to other verticals. To start to familiarize other verticals, SafeGrriph hosted booths at industry conferences and offered weekly webinars to interested parties that visited their website. With each passing experience, the SafeGraph team learned more about what potential customers wanted to do with the data. Yet as Cook explained, "There are Just so many use cases and personas that could benefit from this data, even within the same company." His team frequently targeted corporate professionals with titles such as Product Manager or Data Scientist Companies that were known to already use several highrquality analytics tools were thought to be more likely to be interested and have the internal capabilities to leverage the data. SafeGmph posted curated data set samples on their website to provide prospective buyers an opportunity to experiment with at no cost. One example included foot-prints of 61 hardware stores in Denver, CO. The company also created and disseminated blog and other content on platforms such as Facebouk, Linkedln, and Twitter. Evan Barry, VP Marketing commented, "The challenge is that it is hard to put into words what the data does and since we don't offer a dashboard interface, we don't have an easy way for customers to evaluate our product We sometimes create visualizations for our content, but we need to be careful that prospects do not get confused and think the visualization is our product and not the data. \" Prospective buyer visitors to the company homepage were invited to upload their own CSV files to compare, contrast, or merge with SafeGrriph'b data. The team tracked how various content messaging resonated with prospects as they moved through the funnel. Those that showed interest were offered a free data download so they could experience the 8 data themselves. The next step was the evaluation process which started by the Sales team understanding how the potential customer wished to use the data and then identifying the success criteria. In some instances, SafeCraph Sales turned away customers if their expectations were not aligned With what the data could deliver. Ironically, there were several instances Where this stance led to conVersion because it fostered trust with the client. The evaluations were on a self-serve model and Sachraph chose to offer little hand-holding during the process. Hoffman decided to keep the size of the sales and marketing teams fairly modest relative to fast growing Silicon Valley SaaS startups so as to maintain financial exibility as they searched for productr market fit. This careful and methodical approach allowed the product team to work closely with customer support to keep an open dialogue about feedback learned along the way so that they could continuously improve the product or dream up new use cases. Buying cycles ranged from 4 to 6 months, but in some instances endured almost one year. SafeGraph Customer Segments 2021 By 2021 SafeCraph sotd to nearly 100 customers across over a dozen industries with the top three being Ad Tech, retail. and financial services. Three of the newest customers were new to purchasmg external data; (See Exhibit 4 for customer breakdown.) Ad Tech, At the most basic level, Ad Tech customers used SafeGraph data to make decisions on targeting ads based on where and what POIs consumers visited. The idea was for data to reveal the impact of advertising by knowing which campaigns drove in-store traffic. For example, Bitlups, a company that serviced outofhome and outdoor adve ng marketplace (e g., physical and digital billboards) for brands and agencies faced an attribution challenge -- not knowing how viewing an outdoor ad impacted a consumer's shopping behavior. SafeGraph's data helped Billups turn anonymwed GPS data into contextuali'zed store Visits by virtue of its building footprints data. In this way Billups was able to help its client choose ad placements based on proximity to certain I'Ols. It also freed Billups from having to maintain its internal POI data set about store openings and closings, Another customer might serve up certain inapp experiences or notifications depending on where a consumer was located relative to a POI. Similar information might enable an advertiser to launch campaigns on devices at competitor outlets, Retail, Stores and brands used Sachraph data to better understand a variety of attributes of their customers. They also analyzed the data to make decisions about real estate management such as opening and closing locations, decisions SafeGraph believed used to be made \"by gut or anecdotally "5 l'hese decisions were informed by not uniy foot traffic at their own stores, but also that at its competitors. A brand might also analyze the data to drive decisions about inrstore experiences. Neoway, a consultant to the consumer-packaged-gootls industry, used SafeGraph data to advise a client on the right mix of beverages stocked at restaurants, bars, and stores in a given area. Neoway analyzed Patterns data on consumers home neighborhood and cell phone type (iOS or Android) to drive recommendations on product mix by location. Financial services. These organizations often relied on traditional official data sources such as Bureau of Labor Statistics Unemployment and Consumer Price lndices, to predict market movements and inform investment decisions However, in recent years, large Financial institutions started to leverage alternative data as they built out their own data science teams. For instance, they used transaction data front credit card companies to understand consumer spending and behavior. SafeGraph had targeted hedge funds and private equity firms with limited success before the COVID- 19 pandemic. Yet as the Chief US. Economist at Goldman Sachs noted, as economic activity turned on a dime with virus shutdowns anti reopening, financial services players needed to look to new sources of information that was updated quickly.7 Soon SafeGraph signed Goldman Sachs, Morgan Stanley, American Express, four large private equity firms, and two of the largest hedge funds. "they used SafeGraph data to research sectors and understand growth patterns based on store openings and closings. Ieffries, Inc.'s Chief Financial Economist commented, "I think it's [alternative data] going to be part of the normal toolkit [after the pandemic] Maybe I won't be refreshing my spreadsheets everyday like I do now. . .when things stabilize maybe I'll go to once a week But I certainly don't think we're going to stop looking at these data.\"5 Leveraging Noantiying Customers In addition to selling to enterprise customers, SafeGraph forged relationships With journalists and academics as part of the Company's mission to democratize access to quality data SafeGraph offered free access to its data to these groups and in so doing built trust and name recognition. Similar to its approach with paying customers, SnfeGrapli often pitched data storytelling article ideas to the press As a result the company developed non-Commercial agreements with several of the most lauded news media outlets including: New York Times, Washington Post, Writ] Strcctj'mtmat, and NPR. Hoffman's long- term view was to have a mutually beneficial and profitable relationship with those organizations. He felt a certain kinship to the media, having once commented, \"SafeGraph is a 21st century news organization (Without the op-ed department) whose product is data."9 The company also offered a SatbG mph ir Good program featuring free access to academic researchers. The company's data was used in papers by professors from top universities. One such study from St. Louis University examined consumer response to corporate political statements by examining foot traffic at Walmart after the company ceased selling certain types of ammunition and banned the open- carry of weapons. Its finding: foot traffic to Walmart stores in highly Republican counties fell by 8.5% and increased by 2.7% in highly Democratic counties.\"l Another study from the University of Texas- Dallas and Boston College showed that traffic to Starbucks fell 6.8% relative to surrounding cafes after the chain instituted a policy that allowed nonrpaying customers access to restrooms.\" Business Model SafeGraph's costs were similar in composition to those of a SaaS company, with a few key differences, Apart from headcount, Safe-Grapli's primary costs were its licenses with approximately a dozen data vendors, which were typically structured as annual subscriptions, (SafeGraph crawled thousands of websites for open source data free of cost). Data acquisition expenditures were placed into cost-ofgoods-sold (COGS) on the income statement. On the balance sheet, SafeGraph held almost no hard assets apart from employee laptops as the team was fully distributed. (Hoffman had shuttered the San Francisco HQ a few years prior to access a national talent pool, and during the pandemic, decided to base the company in the lower-cost, remote-friendly city of Denver, CO.) SafeGraph's trove of data appeared on the balance sheet as a long-terln asset given that data sets could be repurpused and resold many times with minimal ongoing maintenance costs. SafeGraph originally offered its device-level Movement data on a volume-based pricing model where users could filter the data they consumed. The company later adjusted its pricing for its new products to a \"full file" modet that allowed a simpler product delivery, reduction in cost (to filter the files), and a reduction in data costs for the majority of customers. Having learned from this experience, SafeGraph also started out offering Places data at a single full file rate. Yet, over time they ran into issues with a certain subset of customers voicing the fee was too high. The company addressed this in two ways. One option was for the customer to still purchase the full file, but their usage rights would be restricted to a subset of the data. Another option was for customers to purchase smaller perrrow data sets a-la-crirtc in a 'sclfscrvicc' store on the website using a crcdlt card at $0.10 per record (for instance Core Places data on 1,000 cafes in Chicago cost $100.). The team maintained a flexible approach to pricing and set prices based on their assumptions about a given entity's incremental value derived from the data, which would impact their Willingness to pay. Indeed price elasticity varied by industry. Deals were typically 12-month annual subscriptions with customers invoiced annually (some deals were on autorenew). With a bulk discount, the size oi an average animal su bscription was roughly $100,000. StifeGraph estimated that fastrgrowing data companies were capable of achieving the high profit margins (7mm typically associated with high growth software companies since data, like software, could theoretically be distributed an infinite number of times With nearly zero incremental costs. White cash flow positive, the company was not yet profitable Protecting the Data SafeGraph dealt with potentially sensitive data and felt a responsibility to steward that data. SafeGraph's customers signed data agreements that specified the rights of the data buyer in terms of how long they could use the data as well as whether they had the right to resell the data, Since data is easy to copy, data companies often watermarked their data. That is, they added very small bits of inaccurate data informally called \"honey tokens" that would help them identify if their data had been copied and resold. This practice was similar to that used by cartographers of old who added fake locations to their maps so they could identify fraudulent copies, Instances oi discovering their honey tokens in unexpected places was exceedingly rare and in those cases, SafeGraph approached the data leaker to see if they wished to revise their terms of use, and the unauthorized data user to see if they wanted to become a customer. The COVID-19 Data Consortium In March 2020, as the pandemic unleashed its global spread, governments reacted by issuing mandates for citizens to remain at home, shuttering locations where people congregated such as businesses, schools, and places of worship. Hoffman and team quickly recognized that their data could help demonstrate the impact of these measures on people's actual movements. Such information would allow governments, scientific bodies, nonprofits, and research entities to map that date to other information such as the virus's spread and economic trends, which would prove valuable to informing policy decisions. Driven by an instinct to help society and alignment with the company's mission and values, the decision was made to provide any such group free access to SateGrapl-l data through its website and the COVID-l9 Data Consortium (beyond the already-no-cost academic access). The Sachraph product team worked around the clock to leverage Patterns data to CGTIJUI'E visualizations for the COVID-'l'? Toolkit and it released this information weekly on Tuesdays. The Impacl' of COViD-IEI on Foot Tm'fc Dashboard showed line graphs illustrating changes in movement by industry, brand, or region. For example, by mideMarch, the. impact of shelter-in-placc orders was severe for brick-and-mortar vcnucs with drops in visits ranging from an to 80% at restaurants, hotels, cafe, and bars, On the other hand, foot traffic rebounded by 30 to 40% in the same time period at locations where consumers stocked up on groceries and supplies such as durable goods stores, The team also launched colorrcoded maps such as the Geographic Response to Shelter-inducer Dashboard that demonstrated by state and county the degree to which citi7ens adhered to physical distancing (by measuring the number of devices leaving home). As governments began easing restrictions and slowly reopening businesses in the spring. SafeGraph released the Reopening tile Economy: Industry Flint Traffic Dashboard which showed foot traffic by state and industry and ranked states on how close their foot traffic was compared to normal, For example on the Weekend of the 2020 Thanksgiving holiday, the southern states of Mississippi and Alabama were ranked #1 and 2 for closest to normal and western states of Washington and California were #48 and 49. While these dashboards proved useful in illustrating movement trends, the uptake of the actual SafeGraph Places and Patterns data was swift and widespread, Eventually over 1,000 entities took advantage of the COVlD-l9 Data Consortium including the US. Center for Disease Control, the National Institute of Health, and the states of California, Ohio, and New York. The Federal Reserve Bank, which traditionally relied on weekly or quarterly economic statistics, hungered for immediate data and made use of the daily metrics. One policy advisor at the Dallas Fed remarked, \"All the usual models and the economic statistics that we would look at were essentially almost entirely useless because they're just way too delayed."u Relationships with the Data Consortium's government and nonprofit partners were maintained by a group of community managers, separate from the enterprise sales team, However, when consortium members wanted to delve further into certain data, they were referred to the enterprise sales team. At the start of 2021, ten months into the Data Consortium, Hoffman took stock. \"l've never worked as hard as I did those first six weeks," he acknowledged \"The Consortium Was a success. lt's translated to attracting paying customers faster than i anticipated.\" Barry noted, \"There was a snowball effect, When one state gowmment noticed another one Was using our data, they approached us to sign on. That was a new phenomenon because our other clients, say financial services, tended to hold theirdata sources close to the vest to maintain competitive advantage." Mentions of Sachraph in the news skyrocketed, Academic papers using Sachraph's COVlD-19 data were being written and published in a matter of weeks as opposed to the more typical year-plus timcframe. Word spread quickly in the academic community and soon over 300 peer-reviewed papers mentioned Sach-raph, increasing the company's visibility and respect. For example, a paper published by researchers at University of California at Davis used data trom SafeGraph, Place 1Q, and Google Mobility to demonstrate that people's mobility away from their home varied by income level. 1' Despite the program's success, Hoffman recognized that maintaining the Consortium was costly in terms of his team's time as well as the cost of the associated data acquisition and storage, and he needed to decide how to proceed. He weighed three options: (1) continue free access indefinitely; (2) institute a freemium model where some data would be kept free, but begin to charge or upsell for new products; or (3) convert all access to paid subscriptions at the one-year anniversary approaching in March, The third option could be implemented by subsidizing early government adopters so they could prove out the model. (See Exhibit 5 for charts of Data Consortium growth and user breakdown) Looking Ahead With his laptep's microphone on mute, Hoffman crunched on his trail mix as colleagues from three time zones logged onto the video conference call. The pause allowed him to muse about SafeGraph's future. Industry observers estimated that by 2025 the geospatidl analytics field would grow to $96.5 billion, representing a 12.9% CAGR from 2020.\" He believed data would be a winnerrtakesrmost opportunity and he expected Safe-Graph would be a large and enduring player in the space. Yet so iar, SafeGraph's customers were the early adopters and even still, they were only scratching the surface, Crossing the chasm and rapid scaling were thereiore key. lncrernental growth would not enable the steprfunction growth needed. The team would have to continue to push hard to continue to secure supply to data, scrub that data well, and provide it to customers in ways that enabled them to use that data for their own insights. He knew that prioritizing and sequencing the customer segments and setting the right pricing model would be key