On Thursday August 22nd, along with one thousand other Sydney-siders, Missing Perspectives hit the ICC in Darling Harbour. No, we weren’t there to see Elton John. We were there for something much nerdier: Stripe Tour.
For those uninitiated with Stripe’s unique role in the global economy – I invite you to consider the mechanics of putting on a high school play. Some students gravitate towards cast roles, wowing audiences with their acting, dancing and singing. By curtain’s down, everyone knows their names.
Others – equally talented, but in different ways – handle the behind-the-scenes, ensuring the lighting, audio, and set is on point for the whole production to run smoothly. That’s Stripe, but for business, or as their team puts it, “building the economic infrastructure of the Internet”.
We met with Stripe head of information Emily Glassberg Sands, who across a single week can be found context switching between rolling out new technologies like AI to figuring out how to serve Stripe’s evolving user base, which includes Stan, Leonardo.ai and Midjourney.
Emily grew up in Boseman, Montana, and attended a rural school called Monforton. During this time, two seeds were planted that have sprouted in her life and career today: her sensitivity to inequity, and strong female leadership.
Many of Emily’s Monforton classmates lived in the nearby Forest Park trailer park, and grew up with predominantly working class parents. By nine, Emily was struck by the unfairness of opportunities already emerging between different students her age. Her Mum explained that while every child certainly has the capacity to be “gifted and talented”, not all have the support or resources needed to bring those fruits to bear.
In response, a young Emily teamed up with Jane Fonda’s nieces to start a “Girls For Change” group to help empower girls 8-13 to build community, think about their education, and consider college opportunities.
This memory and sensitivity to socioeconomic deprivation is a feature that Emily shares with Stripe co-founder and CEO, Patrick Collison.
During a media roundtable, MP asked Patrick if Stripe’s palpable focus on the company’s users and customers throughout the morning’s presentations – and increasing their wealth-generating capacities with a focus on revenue – was deliberate.
Patrick, who has a quick, positive demeanour, seems to come alive on this issue. It’s easy for companies to become solipsistic, he said, so Stripe’s focus on users and customers is very much an intentional choice.
Going deeper on about wealth generation, the Irish-born entrepreneur noted that he lives with the cultural memory of Ireland during the 1950s and 1960s as “the poor man of Europe”. It wasn’t until the 1990s that the country gained some economic momentum as the “Celtic tiger” entering European markets. For Collison, improving the entrepreneurial capacities of Stripe users from Guatemala to Australia as a hedge against backsliding into poverty, appears to be a driving force.
“You have to remember that the average global [personal] income is still $10,000,” Patrick Collison said. “That’s not enough to live the kind of good life that the people in this room are likely accustomed to, so it’s clear that globally, we are faced with this collective challenge… building better products and services is the best answer that we’ve come up with so far.”
But back to Emily. From making bold moves out of academia into the private sector by convincing online education startup Coursera that they could use data science to improve the company, to her work at Stripe today, Emily Glassberg Sands is one to watch.
Missing Perspectives: Firstly, thank you for making the time today, we really appreciate it. It was great to see you on stage this morning, and I think this has been a big learning experience for us. To simply be at a global fintech conference is really exciting.
Emily Glassberg Sands: This is our biggest tour! We do Stripe Tour in Paris, Berlin, London, New York, Singapore, but this is our biggest. We had a thousand folks in the room today.
You can feel the energy. One of our MP contributors, Cara Davies, wrote an article about how she started and sold a Gen Z fitness app. She recommended a book to us called The Unfair Advantage – in it the authors contend that all entrepreneurs will have a unique set of unfair advantages. So Emily – as a woman in a significant role at a very entrepreneurial company – what are yours?
I think my unfair advantage is I’m obsessed with understanding why people and businesses make the decisions they make, and where those decisions are suboptimal, helping them make better decisions. This was a bug I got back in college when I met this amazing female playwright in New York named Julia Jordan. I was like, 20, and she said Emily… did you know that only 1 in 5 plays in production are written by women? Can you help me figure out what the heck is happening?
And I spent a year doing a series of studies, including an audit study where Julie and her friends contributed four never before seen scripts, and I sent them out to hundreds of artistic directors and literary managers around the country just varying the pen name on the script – is it written by Mary Walker or Michael Walker – and asking them if they wanted to put it in production, and asking them why or why not?
And what I found was depressing, which was the exact same play when purportedly written by a woman was less likely to make it into production. But the really rewarding part was that the theatre community cared. Like, that insight led to action. They actually did want the best plays in production, and once there was that awareness around it, people started talking about it, The New York Times wrote about it, and now a decade and a half later, half of the plays in production in the US are written by women.
And it wasn’t just that study, but a movement that came about with this recognition that the market was operating inefficiently, and businesses want markets to operate efficiently, the theatre community wants markets to operate efficiently, and individuals actually want to make rational, optimal decisions.
I think my comparative advantage is I have an obsession with why people make the decisions that they make, and in cases where those decisions are inefficient, help them be better. And a lot of what Stripe is doing is empowering business to make better decisions, whether that’s optimising checkout for them, or blocking fraud, or providing insights about what markets to expand to, what payment methods to add, what products to go after, what regions to consider, because there is a lot of imperfect information in the world, and you only see the slice of your own experiences, and being able to look across a broader swathe and play that back… it’s a bug I can’t get out of my head.
It must have been such a fortuitous run in, because it sounds like that meeting with the playwright and the work you did has propelled you into so much of your work now.
Totally. I was like, I want to go to grad school, and I want to do research. And one of the things I did there was look at the fact that half of jobs were found through referrals. And I was like – why is that? Is it a selection thing – the fact that you’re referred to signals that you’re better? Is it a collaboration thing – you do better work because you know people at the firm already? Or is it a little bit of nepotism? I ran a series of studies on that in grad school, and it’s that same sort of obsession with understanding why people make the decisions they make that took me to Coursera. Coursera was a little 40 person startup at the time …
You’re underplaying it. That was quite a significant turning point for you in your career to go from academia to…
Industry. And this was 2014, so like, before academic economists went into industry. Now, academic economists actually go to industry, and become data scientists. But at the time, people didn’t really know what data science was, and economists weren’t going into industry.
So what happened? I had a friend from college who was at Coursera, and I was talking to him about what the company was doing, and I thought what better way to level the playing field in the job market than to level the playing field in education. If you could actually get anybody access to world class education from the likes of Stanford and Harvard and wherever else – you can really elevate them, and the opportunities that they then have in the labour market.
So yeah, it was a bit of an unusual choice, and actually, I didn’t initially apply to be a data scientist. The only job posted on the site was a sales manager, which is basically like a sales person, a role where you go and convince universities to put their content on Coursera.
So I applied for it, and I made it to onsite for it, and then I got to onsite and I kept pitching my interviewers on like “you know… you could measure the value of a Coursera credential. You can figure out how to communicate to employers how useful it is that people have these skills. And you can close the loop of like – they’re not just learning the thing or developing the skill, they’re actually being rewarded for that skill in the labour market.” They rejected me initially, but within a few months of me pestering them, Andrew and Daphne ultimately got to a ‘yes’ and I joined Coursera as our first data science hire.
You’re in a significant role – it can be overdone to be like, there’s a lack of women in STEM, or there’s a lack of women in data science, and honestly from a data perspective, I don’t have the full picture. What do you think about your role today? And how could we get more women into roles like the one you have at Stripe today?
I joined Stripe about three years ago, originally to look after data science, and I didn’t know a lot about Stripe, but David Singleton our CTO reached out through a mutual friend and from my very first conversation with him, I realised two things.
First is, Stripe was already helping millions of businesses be more effective and efficient and reach customers in new ways, and they were capturing a tremendous amount of data in doing that, which meant there was a lot of opportunity to do even better for those millions of businesses through data. The second thing I realised is how beneficently Stripe has the luxury of being able to act. So there are lots of businesses that can’t operate beneficiently, in the case of Stripe, anything we do that helps our businesses grow, helps us grow.
And why is Stripe in that unique position?
Stripe is in that position because our core product is our payments product, and so we monetize on payments that pass through Stripe, so if we help a business grow its revenues… [MP readers: The main way that Stripe makes money is by taking a cut of every successful payment processed through its platforms]
You’re linked to that in a win-win kind of situation…
It’s a win-win. And then the GDP of the Internet grows. Anything you can do that helps businesses grow, helps us grow. I joined Stripe originally to look after data science, today I lead the information organisation, which broadly speakingensures that Stripe can effectively use it’s data, so whether that’s machine learning infrastructure, data engineering, analytics or data science, or building our new GenAI bets, and the second thing is I’m accountable for the self-serve business, so making sure that the small and medium enterprises and startups that come to Stripe have a great experience.
In terms of how we can get more women in tech or data science, I think that data science can be a really powerful entry point to technology roles for women. One reason is that social scientists are actually really good at data science, because a lot of what you need to understand is causal relationships. And you know, economists don’t skew heavily female, but political scientists do, and psychology PHDs do, and there are a lot of highly quantitative empirical fields that actually have almost equal shares of women and men that end up doing very well in data science. I don’t know what current industry stats are, but it’s probably like 35 per cent female, whereas engineering would be 25 per cent female.
The other thing that I think is important when I’m recruiting or retaining, on average for women it’s about more than just the technology, and more about the why, the so what, the impact that they’re having. And you know, obviously building economic infrastructure for the Internet has a tremendous so-what, but it’s increasingly important I think for women to emphasise that impact, and to motivate them to stay the course.
In a simple way, can you explain what a head of information role actually is?
A head of information means different things at different companies. At Stripe, it’s about making we’re using our data effectively for our users, and supporting in a scaled way all the small businesses that operate on us who don’t have access to a sales team and high touch services, and therefore really need an in-product experience or a product-led growth experience.
And in terms of how your week looks, who are the main people that you interact with?
The vertical product teams, the core payments team, the billing team, and the connect team.
I also spend a lot of time with the go-to-market team because I talked about self-serve users, but really it’s a continuum, a lot of our startups may start on Stripe as self-serve users, they get going directly in the product, we’re serving them with pure technology experiences, and then they grow very quickly.
The Leonardo.ais, the Open.ais, the Midjourneys – they grow very quickly – and all of a sudden they’re in a more high touch world, and so there’s lot of handoff between the two, and ensuring that as users grow, they’re supported with the right set of services.
What’s your relationship like with OpenAI?
OpenAI are a customer, and we are a customer. So we use their models, and they use us for payments and subscriptions and so on.
And LLM Explorer, which is a tool you created for your staff to use internally, does that fit in the Stripe GenAI experiment bucket?
Yes. The model landscape is evolving very quickly. And there are a number of different model providers, and a number of different model providers that use Stripe. I think our approach was Open AI is amazing and GPT 5 and 4 are awesome, and, different models work better or worse for different use cases, right? Like any given use case has a whole set of requirements around cost and latency, like how quickly it has to return an answer, and models perform better or worse in different contexts, and so, from the get go, the LLM infrastructure that we’ve built supports a bunch of different models on the backend. Today we’ve exposed over a dozen models to Stripes, some of those are from Open Ai, and others are by other model providers.
Okay. So on the governance side with AI – I struggle to even frame this question because when I was trying to research AI regulation, honestly it sort of felt like it’s happening in real-time, and sometimes you get regulation that’s like, Armageddon is coming we need to stop now, and some regulation is like, AI is the key to innovation. How do you navigate that landscape?
I honestly think it’s not either or. It’s like, well what are you using the AI to do? And that informs how deeply you need to be interrogating the policy and regulatory implications. So, as a simple example, our capital product is extending loans to people. It is incredibly important that we don’t make loan decisions on the basis of demographic information. In fact, our entire capital product is seeded on the premise that people should have access to capital not on the basis of traditional demographic attributes but instead on the basis of the actual health of their business. And because we see their payments data, we’re able to extend loans on the basis of their actual realised payments flows instead of traditional signals of credit worthiness.
That has a few implications – we don’t have to stress about whether or not the AI is biased because we’re actually using payments data not any demographic information. But two, it allows us to actually give loans to businesses that actually wouldn’t on the basis of traditional credit signals look creditworthy.
It depends on the domain, and there are some domains where it’s incredibly important to think very deeply about the implications for the users, and there are other domains where it’s just not risky in the same ways?
How does it work when you guys get together to have a discussion about what you will and won’t do? Who is in the room?
Legal, policy, product, engineering, AI teams.
And literally – if you have a sticky situation – do you get on a Teams call? Or do you like, write essays? How do you bring that cognitive diversity to bear?
One of the things I love about Stripe is the use of the written word. And the reason I like it is that the best ideas bubble to the surface, rather than the loudest voice. People are very much rewarded for how they reason from first principles, and the results that they produce.
In this case, as in most cases at Stripe we have a RAPID, so it’s clear who is making the recommendation and who needs to agree, and who’s providing input, and who’s making the decisions. And the thinking behind that RAPID or that decision is crisply articulated generally by the recommender, but with other folks weighing in. We don’t weigh in completely in the document, but I’ll often walk into a meeting about this or anything else, and it’ll be like 15 minutes of reading time at the top, and someone has put together a really thoughtful recommendation and brief, and then 45 minutes of discussion, and then decisions are often made offline after there has been time for reflection. But the written word levels the playing field in a way I think little else does.
So in a way, it’s not just extroverts being able to move the groupthink towards that.
At this point, Stripe’s director of communications for Asia-Pacific Leigh, who was sitting to Emily’s left, leant in to quietly suggest to her “do you want to explain what a RAPID is?”.
Oh yeah! RAPID is just a decision-making framework. In a RAPID, you state explicitly what’s the decision to be made, but then as importantly, who are the people accountable for different components of that decision. The R A P I D simply stands for different roles that individuals have.
Cool. Zooming back out from the granular to the more general, are there any women mentors you could credit in your career or personal life – you’re a mother as well – that have played supportive roles in getting to where you are today?
For sure. The reason that female playwright wanted to talk to me, actually had nothing to do with this junior in college. She knew I was working with Cecelia Rouse, who was an economist who wrote an awesome study called Orchestrating Impartiality. This one particular paper found that when orchestras implemented a rule for blind auditions, where you actually have a screen between the person auditioning and the decision makers, and they rolled out a carpet to mask the sound of heels, you consistently saw a jump in the number of women accepted into orchestra.
Cece was an amazing mentor. Cece is actually why I went to Harvard to get my PhD, because she wrote Orchestrating Impartiality with Claudia Golden, who just won the Nobel Prize in economics. She’s an economic historian, but she does really interesting labour stuff, and so not just Claudia, but also her partner Larry Katz, have really been luminaries in understanding labour market inequalities. I had the privilege of having both of them as advisors during grad school.
But then I think I’ve just gotten lucky throughout. My first manager at Coursera Tom Doe was like, awesome, it doesn’t have to be a woman. People can cheer for you no matter who they are. And I was really fortunate to have people in industry with very different strengths than I did. I’ve always reported to traditional engineering leaders or product leaders, not data leaders. It’s interesting when people can see your strengths and support your strengths even when they’re not their strengths.
On the personal, my sister-in-law has four kids, and she has helped me raise my two kids. My husband has also been an amazing partner. I joined Stripe with a five-month old and a two-year-old. Stripe is an amazing company, and it’s also a very intense company. There’s an intense focus on the urgent needs of the users, and how quickly we can solve them, and I love that, it’s invigorating, but it also required me to have a lot of support early on when I had two little kids and was starting a brand new job after eight years at another company.
It’s important to have that infrastructure around you. Where to next for you in your career or at Stripe? Who are the next customers that you’re going to be serving? Let’s say in the next three years.
First, I think Stripe was originally sort of payments infrastructure of the internet, and has expanding to be broader financial infrastructure for the internet, and that creates a lot of opportunities around how we can use our payments data to help customers solve a broader set of problems, automate a bunch of the office of the CFO, and help them optimise their business well beyond the payments layer. AI is playing a super important role.
We also started primarily serving startups, and now we’re over a decade in, and increasingly also serving large enterprises. And as we go through that transition, one of the things that stands out is the modularity requirements of our platform, and the ability to support multiprocessor integrations.
And on the AI front, I think I’m personally really interested in training foundation models on Stripe data, and we’re seeing that pan out in payments use cases. And exploring how we can go one layer up, and provide really automated business recommendations to help you grow your business. And then a level above that, can we actually provide insights about where the economy is headed, what inflation looks like, where the small business index is going, the sorts of things that companies use to steer their one to three year strategic decisions.
Using AI to improve the quality of the signal?
Yeah, basically making it near real-time. Can we near real-time estimate inflation, as opposed to waiting six months for the consumer price index to come out? Can we near real-time understand the health of small businesses rather than waiting for the small business index to come out? And then what decisions are businesses able to make differently if they have that near real time data.
It’s a big responsibility. Thank you for your time Emily!
Emily Glassberg Sands on… 3 women data scientists to get across ASAP
- Ya Xu
“She just left LinkedIn, and I can’t wait to see what she does next. But I think she was there for like a decade leading the end-to-end data org. I love her in part because she was a super senior individual contributor (IC) who moved to management quite late in her career and ended up running the whole org in no time.”
- Dr Fei-Fei Li
Dr Fei-Fei Li is a Stanford professor, investor, founder, and heavily into data science education. She also has an incredible story in her own right – moving from China to the US as a teenager with less than $20 before riding the AI wave in her own incredible career journey.
- Daphne Koller
Emily and Daphne Koller crossed paths at Coursera. She’s an Israeli-American computer scientist and the founder of Insitro. Emily says the unifying theme of Daphne’s life’s work is that it’s all about trying to make people healthier through science.
Editor’s note: The MP team were invited courtesy of Stripe to attend Stripe Tour in Sydney.