Welcome to our first startup social responsibility study - a data-inspired look at trends in "tech for good" (and, at times, not-so-good). It's common for startups to say they're "changing the world." So we wanted to explore an important, adjacent question:
To do that, we crawled, collected, and analyzed publicly available data from multiple sources1 on 40,461 U.S. startups in 8 cities - Austin, Boston, Chicago, Denver, Los Angeles, NYC, San Francisco (including Oakland and the surrounding Bay Area), and Seattle. We also used text classification (text mining alongside tf-idf) to look at what these startups include and say on their own websites. How do they talk about issues like sustainability and diversity? Are startups concerned about climate change? How consistently do their actions seem to back their words, values, and stated ideals?
As a result, our study methodology isn't completely free of self-reporting, survivorship, and other selection bias, given that website copywriting and directory listings can be incorrect, incomplete, or misleading. Nonetheless, we're cautiously optimistic our data set, in aggregate, paints a directionally correct assessment of today's startup reality (and intent) in areas like corporate social responsibility, sustainable business practices, and ethics - particularly in the set of fields commonly labeled "artificial intelligence" or "A.I." Please consider this first, imperfect attempt a conversation-starter.
Overall, we find:
We identify 4,992 U.S. startups focused on social impact, sustainable products, and ethical business models, including education technology (EdTech), healthcare (HealthTech), clean energy and transportation, socially-motivated technology for cities (GovTech), tech aimed at helping vulnerable identity groups (children, immigrants, LGBTQ+, Black/African-American/Carribean-American, Latino/Latina, etc.), and ethically-sourced and manufactured consumer products. Out of our entire data set, we find 12.3% of startups have a social responsibility or stated "green" component to their business model heading into 2019.
Having a mission or social impact-focused business model or product doesn't consistently correlate with broader, company-led social responsibility efforts like having a dedicated (1) workplace diversity program (and diversity leadership), (2) sustainability or social responsibility department (working on sustainable operating practices, environmental footprint measurement and reduction, employee volunteering and giving, etc.), or (3) comprehensive employee health and wellness program2 that extends beyond core benefit coverage. We find only 8.8% of startups indicate they offer employees comprehensive wellness, 1.6% have a dedicated social responsibility or sustainability team, and 1.4% have dedicated diversity leadership. While we recognize many startups are resource-constrained and may group sustainability or diversity work as a subset of another role's job responsibilities (i.e., Director of HR), we nonetheless see org structure, team makeup, and hiring patterns as general reflections of corporate values and executive priorities.
Looking at startups across several metro areas, we see clear, city-level differences in social responsibility, sustainability, and overall cultural prioritization. For example, while San Francisco is home to the largest overall number of social impact startups, Boston (#1) and NYC (#2) boast higher relative concentrations of startups with socially impact-focused products or business models. All three cities, however, fall well short of Denver on measures of employee wellness and giving back. Seattle meanwhile, stands out in the data for its diversity prioritization. Startups in Seattle are more than twice as likely to have a dedicated diversity function vs. the national average.
Approximately 250 new startups have raised $1 million or more in venture funding within the past ten years to develop A.I. and A.I.-integrated business models and services. Of this cohort, only four (1.6%) list a public ethics statement on their website or designate an internal ethics team or leader. Google and Microsoft (both excluded from our group of 250 due to their size and age) have enacted public ethics statements and internal A.I. governance (Facebook is reported to be creating one as well). Other tech giants are primarily exploring ethical A.I. through external working groups and alliances like Partnership on AI and AI4ALL - though the exact output, outcomes, and commitment-levels remain vague.
At a time when a single Bitcoin transaction consumes the energy equivalent of three weeks of power for the average American household, Juul's introducing thousands of young, non-smokers to nicotine, and Facebook's being weaponized around the world to spread hate speech and shred digital privacy rights, we believe all of us in the technology industry have a moral obligation to thoroughly examine the objectives, incentive structures, feedback loops, biases, externalities, and future trends surrounding our work.
We hope this report contributes further urgency around the conversations we need to be having as entrepreneurs, innovators, and technology professionals - and the proactive preventions we need to undertake. We hope you find it interesting, useful, or - better yet - a message that helps inspire positive action and change.
While the startups world is, by definition, a data-rich ecosystem, insights on startup social responsibility and ethical technology have, to date, been limited. In an attempt to better understand this emerging set of trends, we accessed information on 40,461 U.S. startups with a headquarters or primary office in Austin, Boston, Chicago, Denver, Los Angeles (LA), New York City (NYC), the San Francisco Bay Area (SF), or Seattle.
For the purposes of this study, we define a startup as a company:
Our data set is SF-centric, with 19,541 (48.3%) of startups based in the San Francisco Bay Area, which includes SF, Oakland, Palo Alto, and Mountain View.
Source: Angelist, CBInsights, Crunchbase, LinkedIn, GlassDoor, and Brightest
When it comes to socially responsible and "green" startups3, the Bay Area is also home to the largest overall group of tech companies targeting social good.
Social Impact Startups by City
Last Updated: January 3, 2019
Source: Angelist, CBInsights, Crunchbase, LinkedIn, and Brightest
However, given its deep pool of entrepreneurs, startup incubators, and venture capital (not to mention historical traditions of climate, anti-war, and LGBTQ+ activism), on a relative basis, SF lags behind several of its peers in social entrepreneurship focus. We find 21.6% of Boston startups are social impact-focused, led by the city's 138+ healthcare startups (including Kyruus, SHYFT, Wellframe, and PillPack), compared to 13.3% of Bay Area startups. NYC, meanwhile, leads the industry in 'green' and sustainable consumer product startups (42), we hypothesize due to the city's overall influence in consumer culture and advertising.
LA, by comparison, emerges as the weakest major city for social entrepreneurship, with less than 1% of local startups directly tackling education, health, wellness, or sustainability challenges4, despite a strong personal wellness culture and a wealth of sun for solar energy generation.
Social Entrepreneurship Focus Density by City
Last Updated: January 3, 2019
Source: Angelist, CBInsights, Crunchbase, LinkedIn, and Brightest
It's also worth asking what (all) does startup social responsibility encompass? Sustainability and ethics certainly don't start and end with operating in a "green" or socially-motivated industry. Far from it, in fact.
From our vantage point, when we look at the most socially responsible startups, we see common patterns in terms of company-level commitments and actions. Committed startups typically do most - if not all - of the following:
When it comes to environmental footprint, it's hard to apply a one-size-fits-all estimate for an average tech startup. Ultimately, it's up to Operations and DevOps to quantify it properly. We can however, summarize some ballpark approximations using Amazon Web Services (AWS), which controls 35% market share of the global cloud hosting market, and offsets 50% of its power use via renewable energy sources. For reference, it's estimated AWS's data centers consume over 4 million MWh of power per year, the equivalent of 380,000 U.S. households or a city the size of Cleveland, Minneapolis, or Oakland5.
In 2016, Oppenheimer analysts estimated AWS used 1.3 million servers. At AWS's current growth rates (and operating margins), we estimate AWS will reach ~2.2 - 2.3 million servers in 2019, if it hasn't already. If that estimate is correct, an average AWS server consumes approximately 1,742 kWh of power per year, based on Amazon's disclosures. At 0.4 kg/kWh of source carbon, one AWS server produces 0.7 metric tons (MTs) of CO2 per year, or a net 0.35 MTs (0.7 * 50%) if we credit half its power offset by renewables.
Annually then, AWS's carbon footprint from servers could be in the range of 1.6 million MTs, or 800,000 MTs post-offset today. Amazon has indicated "long term" plans to 100% divest from fossil fuels, but to our knowledge has yet to set a specific target date.
AWS Annual CO2 Footprint Comparison
Last Updated: January 3, 2019
Source: Brightest, AWS, EPA.
Using AWS's own calculator, we might generalize and assume the average startup has half a million (500,000) monthly active users (MAUs) and serves five million pageviews a month. This hypothetical startup might run three dedicated m3.medium Linux servers, plus 5GB of static file hosting on S3, 25GB of log file storage, and two PostgreSQL database instances with 200GB of memory. We might assume then, that the average app and software-only startup's "cloud carbon footprint" is roughly 2 MTs of CO2, or one MT passed through from AWS after accounting for its renewable energy procurement. This logic could be used to calculate an input for a startups's DevOps or server environmental footprint [similar logic would apply to Google Cloud or Microsoft Azure], alongside other footprint areas like hardware (laptops) and office energy consumption.
Again, we should clarify that these are directional, outside-in estimates based on the available data. If you can improve our calculations, feel free to get in touch.
Within the tech community, some larger companies like Apple are already leading by example. To be clear, Apple still emits millions of metric tons of CO2 each year, earns more in U.S. federal tax breaks than it invests in sustainability6, and has much more work to do. But importantly symbolic work is happening. Under VP of Environment, Policy and Social Initiatives Lisa Jackson, who reports directly to CEO Tim Cook, Apple is aggressively and transparently reducing its carbon footprint. Today, Apple consumes 100% renewable energy across all its facilities (while peer-pressuring its suppliers to do the same), re-engineered its manufacturing process to reduce its high aluminum footprint, and even designed an iPhone dissasembly robot named 'Daisy' to more efficiently recover and recycle materials from old devices7.
Source: Apple
Google's latest sustainability reports show CEO Sudar Pichai is moving the company in a similar direction. While Apple and Google are long-removed from their early Palo Alto startup days, given their size, prominence, and cultural gravity in the tech community, we expect more startups will divest from fossil fuels and emulate their programs in the coming years - particularly mature, well-funded ones.
Some startups, including Calm, Squarespace, Flatiron Health, Flexport, and Impossible Foods are already on their way. For most of the industry however, these programs appear under-prioritized or, at best, just getting calibrated.
In a competitive market for tech talent, more startups are investing in comprehensive employee wellness programs, including benefits like maternity, paternity, or family medical leave, child care support, free healthy food, tuition reimbursement, fitness subsidies, and on-site wellness facilities.
We find 8.8% of startups operate such programs, with Denver slightly leading the Bay Area in best metro area work-wellness balance.
Percentage of Startups Offering Comprehensive Employee Health & Wellness by City
Last Updated: January 3, 2019
Source: Angelist, CBInsights, LinkedIn, GlassDoor, and Brightest
By comparison, most startups do not have a dedicated Social Responsibility or Sustainability team. Here again, the Denver startup community shows regional progressive leadership, forming this type of team at a 68% higher rate than the national average. Nonetheless, zooming out, the simple, self-evident truth is the majority of tech entrepreneurs today are ambivalent, or at best, inactive, on the threat of climate change.
Percentage of Startups with a Social Responsibility or Sustainability Team by City
Last Updated: January 3, 2019
Source: Angelist, CBInsights, LinkedIn, GlassDoor, and Brightest
The startups in our data set also show low rates of diversity leadership, consistent with First Round Capital's recent findings that only 21.6% of founders say they have "a formal [diversity] plan or policies" at their company. We looked at an even more conservative, org-focused measure: does the company have a leader with the word "Diversity" in their job title? Here, we found only 578 (1.4%) startups that consider workplace diversity important enough to institutionalize it.
Notaby, Seattle, one of the smallest startup communities in our study, scores higher marks than its peers on relative diversity. 3% of Seattle startups have a diversity leader, roughly twice the rate of SF of the national average. Austin and Boston meanwhile, stand out as the most socially conservative.
One final area of research interest for us is ethical technology, specifically the development of ethical machine learning, commonly dubbed "artificial intelligence" (A.I.). While we remain far and away more concerned about climate risks, because of A.I.'s protean but potentially dramatic capacity to shape the future of work, transportation, media, privacy rights, and, among the most concerning, policing and warfare, we consider careful, human-centric development of A.I. critically important. Using machine learning for lighweight personalization, task efficiency improvement, and improved medical diagnosis is one thing (and can be immensely beneficial); employing it in areas of legal or moral consequence (law enforcement, immigration, harassment, credit access) requires much higher value scrutiny.
If A.I.'s lets us think and work faster, we fail if we let it accelerate the same flawed social programming pressuring the working class and worsening global warming.
Thankfully, we do not appear to be completely alone in this stance.
Our principal, near-term concerns with A.I. are the ways machine-intelligent processing power might reinforce existing social imbalances - income inequality, erroding labor rights, over-policing, profiling. Indeed, some of these harms and unintended consequences are already happening. A.I. applications and organizational decision-making logic must be designed with an emphasis on positive social value, risk prevention, and bias elimination. And we need informed public policy to match.
For example, Germany's Federal Ministry of Transport and Digital Infrastructure's Ethics Commission - a group of independent transport experts, legal scholars, information scientists, engineers, philosophers, theologians, and consumer protection representatives - recently worked together to author 20 ethics rules for domestic self-driving cars. One rule is a zero discrimination or profiling provision in unavoidable accidents. In other words, if the A.I. algorithm knows the car it's driving is about to crash, it's forbidden from making any decisions on the basis of passenger or pedestrian features like age, race, or gender.
Being engineers ourselves who recognize the multiplicity of open source software and know full well developers don't always see eye-to-eye with philosophers, social scientists, or vulnerable identity groups, we see a much greater need for independent ethics oversight in the A.I. space. Of the 250 startups who have raised $1 million or more in venture funding within the past ten years to develop A.I. and A.I.-integrated business models and services, we were only able to identify four (1.6%) which list a public ethics statement on their website or designate an internal ethics team or leader - DeepMind (the semi-autonomous British A.I. company owned by Google parent entity Alphabet), MightyAI (training data as a service), MedoPad (medical diagnosis), and Shield (autonomous drones, including military use).
Ethics Focus Among A.I. Startups
Last Updated: January 3, 2019
Source: Angelist, CBInsights, Crunchbase, LinkedIn, and Brightest
Needless to say, if half our sample set of A.I. companies who do have internal ethics oversight sit in weapons manufacturers' algorithm supply chain, we're not encouraged by this trend. And while we do see signs of proactive ethics leadership from tech giants like Microsoft and Google (Alphabet), its impact on the various trajectories of the field is unclear.
One related, early-stage development project we like is Ethically.ai, a python library for auditing the ethics of text-based models for gender and racial bias. Ethics-focused open source contributions are needed, and we hope to see more initiatives along this and adjacent themes. After all, as the linguist J. R. Firth famously noted, "a word is characterized by the company it keeps."
Altogether, we see pockets of promise in the startup ecosystem that give us hope for the future, alongside a lot of sluggish or counterveiling trends that suggest we collectively could - and need to - be doing a lot more. More entrepreneurs outside of cleantech need to be stewarding the Earth's climate destiny, and more responsibility is needed at the front lines of company, culture, and algorithm-building to fight widespread bias across race, gender, and class.
If our goal as startups is to disrupt the status quo, we need to be asking "why," "how," and "for what end" along the way. If we're going to build the power to reach people around the world in seconds, let's use that power responsibly. And if we're going to call ourselves stewards of the future, let's make sure it's the right one.
It's not enough to just change the world - we owe it to ourselves and future generations to change it for the better.
Thanks for reading.
1 Data sources for this study include Google, CBInsights, Angelist (angel.co), Crunchbase, Glassdoor, LinkedIn, and each startup's primary website, with additional data enrichment and analysis by the team here at Brightest. All data gathered between 12/22/18-12/27/18.
2 We define a comprehensive employee wellness program as one that includes four or more of the following benefits: maternity, paternity, or family medical leave, child care support, free healthy food, tuition reimbursement, fitness subsidies, and on-site wellness facilities.
3 A startup whose primary business is education technology (EdTech), healthcare (HealthTech), clean energy, sustainable transportation, socially-motivated technology for cities (GovTech), tech aimed at helping vulnerable identity groups (children, immigrants, LGBTQ+, Black/African-American/Carribean-American, Latino/Latina, etc.), or one or more ethically-sourced and manufactured consumer product(s).
4 For clarity, we do not consider electric scooter companies like Bird and Lime socially impact startups, since it's unclear scooters are replacing meaningful fossil fuel-based transportation (i.e., short drives) compared to short-distance, carbon neutral activities like walking and biking.
5 AWS January 2018 Sustainability Report via Amazon
6 "Apple’s Three-Month Tax Savings under President Trump’s New Tax Law: $1.68 Billion", Institute on Taxation and Economic Policy (ITEP), 2018
7 As reported in Apple's Environmental Responsibility Report - 2018 Progress Report, Covering Fiscal Year 2017.