Build your company like a product | Sahil Lavingia (Founder & CEO, Gumroad)

How to listen

Welcome to Subscription Heroes. I'm Scott Hurff, your host, co-founder, and Chief Product Officer of Churnkey. In this episode, I sit down with my friend Sahil Lavingia. He's an entrepreneur, investor, writer, and painter. He also founded Gumroad, which empowers creators to sell goods and memberships with little overhead. It's famous for running as a lean open startup with a no-meeting, no full-time culture, and it now generates $20 million a year in revenue. He also published his book The Minimalist Entrepreneur in 2021 through Portfolio.

In this episode, we'll discuss how he relentlessly automates grindy tasks, how he's using AI on internal projects, and more. Enjoy the chat with Sahil. We'll kick things off by discussing his latest price hikes, the drama around this big decision, and how churn has been affected.

How a Pricing Increase Defied Expectations

Scott: All right, let's talk pricing and pain. The pricing rollout was in December, right? And then the prices kicked in on January 31st. You got a lot of crap for how you rolled it out. Any regrets, learnings, anything that you thought would happen badly that didn't happen?

Sahil: No regrets, but definitely some learnings along the way. To be honest, I think it went better than expected. I knew going into it that there were going to be a lot of people mad on the internet. I think that's sort of expected every time you make a big change, especially a negative change. We're effectively telling creators that their taxes are going up next year, and they're sort of allowed to be mad about that.

I think overall it clarified a lot of beliefs I had going into it. I was in the minority in the sense that I thought there was not going to be high churn. A lot of people were like, "No, you're making a mistake; you need tiers, you need volume discounts for large creators, you need to grandfather people, you need more time." All these things that people said we needed to do.

Use the Churnkey churn rate calculator to check your current numbers.

I think the reason the reaction was so strongly negative was that I didn't feel like I had to do those things. I could just say, "Hey, the price is changing; this is how it's changing, hopefully you keep using Gumroad." That was the whole entirety of the email. The reason I did that is that I just wanted people to focus on what was actually changing. When you write fiction, there's a kind of purple prose, lots of flowery language, you distract from the point; and then there's just: this is what happened. I like to be on the latter side of things, especially in this environment where there's a lot of crazy, weird stuff to wrap your head around every day.

Pricing is going up, we're getting rid of the tiers, the fee is now exclusive of Stripe's, so Stripe fees are now on top of the Gumroad fee instead of included. It's a lot of stuff to digest. So let me just give it to you, and then you can get mad, you can process it however you will.

For me, it was: let's monitor churn. Let's actually see how many people choose to leave Gumroad. I hoped that people would freak out, but most people would stick around. We expected about 20 to 50% churn.

Scott: Really? You planned for 50%?

Sahil: Yeah. 50% would have been; it was still the right move, but that was hard. 20% was like, okay, totally worth it. And really, we saw about 3%.

Scott: Unreal.

Sahil: And maybe even less on the people who are actually contributing to our margin. Almost nothing. That kind of confirmed a bunch of beliefs I had. I was like, if we actually see super low churn, then the assumptions I'm making are that I should double down on them more, because these are kind of unique insights that I feel like I'm having that maybe the market across the board is not having.

Did Gumroad's Price Increase Hurt Inbound?

Scott: Did it affect inbound in any way, or were you mostly monitoring the existing base, like, oh no, our top people are leaving?

Sahil: Most of the focus was on the existing base. We expected inbound to probably go down, but that's gone up, too, actually. February was stronger than January, and March looks like it'll be stronger than February and January as well.

Generally, Gumroad is on this very slow and steady up-and-to-the-right trajectory that hasn't really changed, like, ever. COVID, we grew a little bit faster than expected, and then it kind of reverts to the mean. The way I think about Gumroad and position it now is: we solve a problem in the market, which is if you want to sell some stuff really easily, a PDF, a course, or something like that, Gumroad occupies that space in people's heads. That social consensus is so powerful. That is a lot of the value Gumroad provides.

Of course, we have tons of features, and we're constantly shipping new stuff, and we have great design; all that stuff really matters. But if you really think about what Substack offers, what Bandcamp offers, what Etsy offers, what Gumroad offers, what Patreon offers, what Amazon offers, there's a lot of the same functionality. What's happened in the last 10 years is that technology has made it really easy to build some of these services. What's hard is the fact that everyone uses Amazon at the same time, or everyone uses Gumroad at the same time. Gumroad's been around for almost 12 years, which takes time. A new competitor can't just come around and say, "Hey, we do all the things Gumroad does, but we're half the cost." The truth is, there are hundreds of those services already.

That's not what makes Gumroad valuable. Literally, the name is valuable. I always use the example of Substack or Patreon as being basically the same service but for two completely different sets of people, and that matters. That has a lot of value. That's just hard to replicate. People focus so much, especially with anchoring bias, on the functionality, and on why something costs 10% or 4% or 5%. But at the end of the day, your iPhone is an awesome piece of software, but it's only as valuable as the network you get access to through it.

One thing I did think about going into this was: there is a lot of competition, Stripe Connect, all these other ways to do it yourself. But to me, that almost gave me more confidence in the decision, in the sense that we needed to find this out. Because the truth is, if we're not providing that much more value, then what's the point? Why are we investing all of these engineering resources and design? No, let's just let everybody go and let this thing run as a Stripe wrapper.

But if we can charge 10% and there really is a ton of value we're driving, then we should go hire more people and keep investing in the product. And there were all these people on Twitter telling me, "No, Gumroad is not worth it." Cool, show me. And then I will react. Luckily, the numbers at least show that there is a lot of value in having a product that does this and solves this problem.

Why Software Companies Can No Longer Afford to Keep Prices Low

Scott: And Stripe's not rolling out, I think you mentioned in another interview with Arvid Kahl, how there were so many creators getting sales through your Discover tab, and you marked that up previously, and they're like, "I didn't even realize you raised prices because I'm still seeing the same take."

Sahil: Yeah, and there is a lot of value in that. Honestly, it's something that I think as an early founder I probably didn't understand or weigh very strongly, because I was like, "No, the best product wins." But if you really think about what you use daily, you have certain things for certain purposes. I think it makes sense to lean into what you're really good at. You can do things like Discovery in ways that you can't if you're just trying to be the cheapest service.

We've been living in this zero-percentage-rate growth environment, and we've all just been trying to offer our services for a really low price, almost like charity. But now we're in an environment where part of the value Gumroad offers is that we will exist in 10 years, in 20 years. And if it costs money to make sure that's going to happen, that may also be worth it. The worst thing that could happen is that this thing just completely goes away.

Scott: Right. And I think software has been a huge beneficiary of cheap financing. But compare that behavior to a restaurant; if eggs go up, your pies are going to go up in price, too. You feel that immediately. But with us, software's inputs have been murky. Money's been cheap, so people take that for granted. People aren't used to raising prices on both sides.

Sahil: The core input into cost for a software company is staff, engineers, basically. And if you do the math, engineers were effectively free in the sense that the market was boosting your stock price every time you hired one. Sure, you were paying somebody $200,000 a year, but the market was increasing your stock value by a million. Facebook, Google, and all these companies were just hiring as many people as possible.

Now we're in a completely different environment. If you look at a P/E ratio of 30, like Coke, you're saying: if you can save $200,000 a year, that's actually worth $6 million, 30x, because that's literally like the P/E ratio of Coke. For every dollar of profit, you get $30 in stock price. So every software engineer you let go is huge.

Companies are still hiring engineers because those people will eventually contribute to software that compounds over many years. But it requires a lot of complexity to understand. It's not like: "This egg costs a dollar, so my omelet goes up in price." It's: "This software engineer, because of zero-percent interest rates, was effectively free, and now all of a sudden these people are still super expensive."

For Gumroad to pay 20 engineers $200,000 a year, that's $4 million. And that requires way more people to understand the actual business model, the macroeconomics. I was having a call with our engineers, and they were like, "What? Why are you talking about interest rates?" And I'm like, "Yeah, the reason you get paid a lot of money is because of zero-percent interest rates." You can be blissfully unaware of that fact, but it's still true.

Now all of a sudden it's: hey, cash flow. Like, if you're a barista, you've always had to live like this. A barista doesn't get to make $400,000 a year because the coffee shop's growing 40% year over year. They get paid $30K or $50K or $60K. Now, all of a sudden, software engineers are more in that bucket; it's more like this zero-sum game where the stock price isn't the main thing. It really is: are you able to drive the value? If you are, you're worth $200,000 a year. If you're not, you're effectively worth zero, just as much as the value you can provide yearly.

That's hard for software companies to adjust to, because all of a sudden, you're actually not being valued like a software company, you're being valued like literally any other business, just a box that generates cash flow. That's hard. That takes a lot of the advantage of being a software company out of the business.

And then when you do need to raise prices, you're effectively taking the first step out of line, because everyone else has been artificially low for so long. You take the hit.

SaaS Pricing: Who Raises Prices First and What Happens If You Don't

Scott: Yeah, exactly. It's kind of a prisoner's dilemma; everyone has realized what we're doing is not super sustainable, but who wants to change first? You're going to be the first person to get shot.

Sahil: Exactly. The first people to move are generally those who have already been canceled by the mainstream, or have experience with that, or are running much smaller, leaner companies with less bureaucracy to wade through, or aren't running public companies.

Some businesses are kind of lucky in that they can change prices without announcing it. The Ubers and DoorDashes of the world have already adapted to this new reality; they're just way more expensive now than they were two or three years ago, and you're deciding on every single one-off transaction.

At the end of the day, you have to look at your business and say: Does this work? At the end of the year, do we have more cash in our bank account or not? For Gumroad, if we didn't change the pricing, we were effectively just going to slowly die. We'd slowly run out of cash. The product would get better slowly, but would never really 2x, 3x, or 5x. You'd never get ChatGPT within Gumroad or anything cool like that. But with the pricing change, if it were to go well, we would actually continue to be able to potentially grow this thing and have it continue to compound.

On the other hand, we were making a bet that if the market said no, we would have just accelerated the demise. But we only made that bet because we had to. There was this existential dread: what if AI comes around and the creator economy is no longer interesting? Interest rates are at 4.5%, we're not growing super fast; we had this amazing 12-year journey as a company and made no money. That would have been weird. So if not now, when?

When Does Your Pricing Model Stop Making Sense?

Scott: You mentioned it was an ongoing realization. Was there something that triggered it?

Sahil: There was not one thing. It's one of those things where you really don't want to consider it, so you put it on the shelf. For example, we had one user, Andrew Tate, who was making a ton of money on Gumroad, and we realized that the way our tiers were structured, he was making a lot of money. It was a high burden on support, but it wasn't really generating any profit for the company. It was a slow year for the creator economy. We also had a bunch of users who'd been using Gumroad to sell things that weren't really good fits, a lot of SaaS products.

People did it because we were super cheap. They'd start and think, "Sure, Stripe can do some of this, but requires a bit more work, so I'll just use Gumroad, and then once I hit scale, maybe I'll move off." But because of our tier structure, we always made it economically worth it for people to stay. The problem is: imagine you start doing a million dollars a year on Gumroad and you're like, "Hey, Gumroad, can you build this feature?" And we're like, "No, because we don't make any money off you." So, our most powerful users, we had no incentive to even help them.

It creates this dissonance. If you run a country, you actually want everyone to feel like they're paying some money and getting something for it. You don't want to make it super efficient where one group subsidizes another. That just leaves a bitter taste in everybody's mouth.

That's also why we didn't grandfather users. What's fair? What's simple? If you're a new creator signing up with Gumroad, the price isn't changing that much. For everybody else, it might be, especially if you've been on the platform for a long time. It's almost like I'm hurting the most loyal users; we're giving the best deal to the people who've been around the longest. But at the end of the day, just because you've been pious longer doesn't mean you get to go to heaven before this person who just joined. Everyone gets to enjoy this thing.

I kept going back to: what's fair, and what's simple? 10% is just super easy to get. If you made $100,000 on Gumroad, I couldn't even tell you before how much Gumroad costs, because it's like: "Well, it's 9% until this amount, and it also includes Stripe, which you wouldn't pay..." But now it's: Gumroad is $10,000. Is it worth $10,000? If so, great. If not, it's not.

How to Handle Customer Backlash When You Raise Prices

Scott: I mean, I know you joke that you're not empathetic, and maybe sometimes your writing can be too blunt, but in a way, this is an act of empathy. You have a company delivering value to thousands of people, you're really close to them, you know what's best for the company and for them in the long run, and you're doing the hardest thing now.

Sahil: It's sort of this J-curve, right? It'll be more expensive in the short term, and they won't really feel like they're getting much more value. It's like, you're not doing anything more than you were doing yesterday. But two years from now, I hope people will look back, just like when some brand you love redesigns and you're like, "This is the worst thing in the world," and then you look back, and you're like, "Oh my God, it was so bad before. Can you imagine if Instagram still looked like that?"

Maybe it's Discovery, Marketplace features, AI integrated into the library, all sorts of stuff we'll be able to do because we had a team, because we were able to pay them well. If that ends up making you more money as a creator on Gumroad, hopefully that should be much more than the few percent you gave up along the way. It becomes a positive-sum game again.

However, I understand it's on us to show people that. People are allowed to be pissed and mad.

I think of it similarly to the SVB bank run reaction. Most likely, it would have played out the way it did anyway. But maybe it doesn't hurt to have a bunch of people on Twitter for 24 or 48 hours just in case. And so when I see all these people really mad and emotional on Twitter, I also have to remind myself: there's a rational reason they're doing it too. Imagine they all rationally said, "Okay, this makes sense, I get it," and no one said anything on Twitter. What feedback loop is that for me? Maybe I could have made it 20 or 30% or something.

My pro tip to people doing that, by the way: I'm still reading DMs, I'm still reading emails. There are many higher-signal ways to get at me than just tweets. I kind of mentally had a list of people I really needed to listen to; maybe 100 people who've made tons of money on Gumroad, whose opinion I respect, who are going to be financially hurt by this change. If I see tweets from them, I'm reading them.

Those people generally were not super angry. Some of them were like, "I'm going to leave, this makes no sense for me to stay. I'm actually glad you did this, because Gumroad was amazing for me to get started, and I kind of just stuck with it. You finally gave me the push."

What is churn costing your business? Get your free metrics now.

Async Work, Notion, and No Zoom Meetings: How to Stay Focused and Keep Shipping

Scott: How do you operate with that deluge of inbound messaging, from customer support to pricing tips to whatever, because you're so transparent? How do you get the headspace to actually execute on things or synthesize all that stuff?

Sahil: I do try to consume all of it, like a fire hose. It's kind of like AI models: you don't really know why it's saying something, it's just somewhere in the training data. I think it's good to just get as much training data as possible.

Working async is super helpful because I never feel like I have to react immediately. I can always synthesize. And I always write everything out. I don't do Zoom meetings, I don't do a lot of audio. So everything I write gets edited, gets a bunch of team feedback. By the time it gets out there, there are always many layers of feedback.

The first product instinct is often: "Hey, this person on Twitter said this thing sucks, here's my attempt at making it better." I put it in Slack, and the team is like, "Ah, we really shouldn't accept Solana right now; think about this a little bit more." When you have that time, I consider it a mania. You have this "we need to do this, and the world would be so much better if we do this right away!" And when you just work through it, it goes away.

Working through Notion, working through these tools forces me to present my thoughts in a much more structured way. It has to make sense within the other things that have come before it. So I'm generally not just blasting Slack, saying, "Hey, we need to do this." It has to fit within the internal roadmap within Notion, which has other things already in the queue.

It's almost like Gumroad forces me to work a certain way. The product has a certain set of goals and requirements and needs, and I have to make sure those things get met. Only then can I do my "fun stuff." For example, the biggest one right now is Discovery. We need to be way better at Discovery. Gumroad.com effectively needs to become like Etsy, where it's a marketplace of content.

I know I need to do that. It feeds into a bunch of other tasks that have fed into it, and then I know once I do that, things start to work out as you build the product. I've been doing this for 12 years, so I just have a lot of training data, a lot of feedback on what the North Star is.

We are building kind of like a faster horse; the Steve Jobs line about "if you just did what customers were telling you, you'd just have a faster horse, not a car." But I really believe in this iterative process of building. I consider Gumroad like an infinite painting. We just make the painting better every day. For example, we automated username generation: when you sign up with an email, we tell GPT and say, "Based on this email, come up with a username for this person," and it does something nice. That's a little bit of friction they no longer have to deal with.

Gumroad's Remote Team Model: Global Contractors, No Full-Time Employees, and Why It Works

Scott: I mean, I'm really jealous of how you've structured the company. It's async, it's written-word first, you're not expected to be on call. And 1099s throughout the world. Was the genesis of this from having to scale down so hard after the early days?

Sahil: I realized today, actually, that anyone at Gumroad can basically do anyone else's job, because everything is writing. An engineer can go to support, a support person could do some design in Figma, everything is there, informationally.

It started basically because in 2016, after the layoffs, we were traditionally venture-backed, and I had to shrink the team down to just me. I was effectively doing all the support tickets every Friday, getting to inbox zero, then fixing bugs during the week. Just not super fun, for like a year.

Anyway, we hired Steve, who had done our support before the layoffs, but as a contract position this time, flexible, hourly. Then we kept doing that. We had some engineers and a designer. Fast-forward to today, and we have about 21 people all over the world who work like this and make a good living. Everyone gets paid $75 to $100-and-change per hour. If you do the math, that's $110K to $260K a year. And this is anywhere in the world, we pay the same. Then we double it in equity: you get $110K in cash plus $110K in equity at the last valuation of the company, which is publicly $100 million.

I really like this flexible type of working where I never feel obligated to create a bundled job for somebody. I create work for people and say: this is the work that needs to happen. It's not a job with benefits and an office and a water cooler and healthcare and all this other stuff. It's just freelance.

And it's been honestly amazing. In the beginning, it was just sort of efficient to run. But now I really think it could be a model for a lot of the way companies could work in the future. The fact that we work in this async, written way means we can hire people anywhere in the world. We don't require anyone's calendars. We can hire new moms, all sorts of different kinds of people who could not really work at this kind of company before, can now work here. They can work 20 hours a week. I think that really broadens the labor pool.

And in an economy where interest rates are at 4.5%, if you can say: we're not hiring software engineers in San Francisco making $500,000 a year, but we can hire software engineers in Malaysia who are amazing at $100,000 a year, that's a $400,000 delta. Going back to the 30 P/E ratio, that's $12 million in market value. It's a huge multiple.

Great for companies because they can save money and run more efficiently. But also great for workers all over the world, they get paid way more, they don't have to move to the US, and they can invest in their local community.

It's a very low-stakes setup; it's just freelance. You can hire people, you can let people go, you can create more work. It just feels like much less "bundled and fight-or-flight" as capitalism sometimes does. You get let go, and you feel terrible, and there's so much emotion bundled into it. I was like: this is just work. Gumroad does this thing, it's awesome. You can work on it if you want, but it's just work. I don't want to pretend that it's some grand purpose we all have.

Like that meme: "born at the right time to work on B2B SaaS." We are doing something valuable in the world. But also, we're doing it because it's a good job and it creates value for everybody involved. We don't have to pretend this is some cult to create AGI or something. There are those companies, too. If you really want to go all in, move to the Bay Area and work at OpenAI. That's there for you. But that's not most companies.

I also think the equity piece is going to be really important. Long term, all of these people are generating value for Gumroad, but we don't need them. We could let them all go in a year, and then all of the value goes to the shareholders. The only way to fix that long-term is to make sure every single person who works on the product gets equity for it, and that equity entitles them to dividends.

To me, this is really necessary to fix the system. Dividends generally alienate people because it's like it takes 30 years to get a 1x, a 3% dividend yield, or something. But yeah, 30 years is a long time. I've been working on Gumroad for 12 years, and I can imagine it will take 20 more years to really generate hundreds of millions of dollars in value. And that's what it takes.

We've been living in an environment where the way you get rewarded is by selling your stock to somebody else. But if that weren't an option, if you had to actually just benefit from the cash flow, the way almost everybody else has to, what would that look like?

That's what I'm exploring now: can Gumroad be sustainable? Yes, we changed our pricing. But in the long term, we're not looking for an exit, not planning to IPO, and not planning to sell. We just make a few million dollars a year in net income and distribute it to the people who own the company over time, like Coke, AT&T, or Starbucks. Companies that have been around for decades, figured out their business model, are generating profits, and returning dividends. And I guess that if you earn equity in those companies, you won't be interested in starting a union. You'll have the best form of union there is: you're literally an owner.

What Is Flexile? The Contractor Management and Equity Tool

Scott: That has ripple effects all the way back to the stuff you work on every day. But the problem is that it's hard to structure, especially keeping track of people from Malaysia to Buenos Aires. This is my segue into talking about Flexile, your new company, which I don't think a lot of people know about, but you're effectively productizing this philosophy, so anyone can do it.

Sahil: Yeah, that's what we're trying to do. Flexile, flexile.com, started as a contractor management, onboarding, and payments tool, and will slowly expand to basically become an operating system for flexible work. If we want this to be more popular, it has to be a lot easier to do, and it requires quite a lot of software: 20 people in 20 countries, everyone working different hours, different roles. Over time, I want people to be able to select how much cash versus equity they want. There should be a slider that rewards them for taking more equity. All of these things need a UI; an accountant has to figure everything out one by one.

My dream is that an engineer can look at their dashboard and see: "I'm getting paid $100 an hour plus $100 an hour in Gumroad stock. Gumroad does this much in net income and is going to do this over time. Assuming flat growth over the next 30 years, this is how much of a dividend payout I'm going to get." And then, engineers should be able to decide what they want to work on based on that. Like, "This task says support people currently spend 50 hours a year doing this manual thing, and if we automate it, that's $2,000 a year in value." Over 30 years, that means a certain amount of dollars directly in their pocket.

Scott: So this becomes like self-organizing workplaces.

Sahil: Yeah, and you could probably get to the point where you're rewarding extra dividends by tracking contributions. Elon had an interesting quote about money. He basically said money is really like information, it's a ledger of IOUs. The best money is the most truthful and the least laggy.

In Gumroad's case, if I can publish everything, these are the numbers, this is the database, this is the work I think is worth doing (which is me really just aggregating customer inbound feedback and trying to tie it to KPIs and metrics), then in theory people can just say, "This is what I want to work on."

Honestly, that's kind of why I don't work that many hours. Like, I do a lot of work, but it's behind-the-scenes work. Reorganizing stuff. I don't actually have to talk to anybody, because when people have work or have time, they just go into Notion. They go into the icebox and say, "There are 25 things tagged 'building needed'. I'll just click and see which ones Sahil wants done."

That's how we work. It means there's no requirement for meetings, no requirement for calendars. It's just me going into Notion and typing stuff up. And because I'm doing that on such a lag, there are about 94 tasks in there; there's probably six to nine months of work already queued. If I died, Gumroad would continue to work. Payments would continue to happen because all the bank account notifications automatically show up in the right Slack room. Everything is so automated and asynchronous that I'm really not needed. I need to DocuSign some stuff once a quarter.

And our CFO, our accounting firm, they handle a bunch of stuff. Cooley knows when to do stuff. So much of this is automated at this point that I honestly wonder: what is everyone else doing?

How to Automate Customer Support: Gumroad's Ticket System Explained

Scott: You're seriously the king of automation of anyone I know. Like, at Gumroad, you do about 6,000 tickets a month from creators, and I think in the next two quarters you'll probably automate half of them just using GPT.

Sahil: Yeah, the technology is there. You just have to take a very product-oriented mindset to everything. How do we automate this completely? A ticket comes into our system, we pull it out, we look at it, and we pull in our help docs. We pull in our prior help center conversations that we've answered successfully. I'm now working on adding more context from our database, so GPT will know, like, this email has bought these products in the past, and then it'll just say: based on all this information, answer this question.

Like, half the tickets are something like: "I bought a product on Gumroad, I don't remember what it was, I bought it with an American Express ending in 4926." I really believe that's a question GPT could answer. There's enough in that single email. I kind of know where I need to get the rest of the information. I have to go to the purchases table and filter by those things. The problem is it's fuzzy, so a human has to say: "Okay, this is the card number, this is the card type," and build in that very robust SQL search logic. Very deterministic versus probabilistic.

But guess what GPT can do? Given this email and given our purchases table, create a SQL query. You might still need a human to approve that the SQL query is going to run, but that's 20 seconds per ticket; something a human currently has to do. It's not work that requires any intelligence. It's just copy-paste, opening two tabs next to each other.

And that's one example. Half of the support is basically stuff like that. "Hey, I wasn't able to get paid out." Okay, click their payment in Stripe. Oh, they didn't upload the right identity verification. That's completely API-driven today. APIs will give me all of that information. Imagine the prompt just has information about the user, their purchases, their products, and it just figures out: "Oh, by the way, it's probably the fact that your Stripe identity verification hasn't been completed."

It's a systems way of thinking. And we're fortunate to be in a position where we've automated so much that it's just how we do things. The reason we're able to automate everything is that we have automated everything, so it gives us time to continue to automate, because we're not overwhelmed all the time.

Even like our financial stuff, we probably spend about $8,000 a month on close because it takes a lot of time to do accounting on $16 million in GMV. But my guess is we could probably automate a good chunk of it. There are probably tons of tasks in there that could be automated. It'll just require like 28 hours of an engineer, the CFO talking to an engineer, me typing up the notes, figuring out the inputs and the outputs, connecting QuickBooks to our Stripe account, having an intermediary layer where Stripe talks to our database, the database pulls in Stripe and PayPal, does some finangling, then sends it to QuickBooks. It's just work. But it can be done.

You have to prioritize the fixed cost of doing that now versus the five minutes it's going to save in perpetuity. Going back to the 30-year time frame: if you can get people to think 30 years, they'll make those kinds of decisions. Most people are thinking about it like right now, "I just need to do the dishes." They're not thinking: "What if I automated dishwashing?" But hopefully someone is thinking about that problem.

It's like building Ctrl+C / Ctrl+V back in the '80s; how much collective time has copy-and-paste saved humanity?

When and What to Automate: Gumroad's Rule for Cutting Repetitive Work

Scott: How do you know when it's not worth it? When are the 100 hours to build the feature not going to be worth the time saved?

Sahil: Honestly, we're pretty extreme about it now. It's kind of like: if it happens twice, we'll automate it. We've become so committed to it. A good filter is: is this no longer fun? What is the grindiest part of working at Gumroad? Right now, it's kind of like: you get an email, somebody can't remember a purchase, it's not a hard thing, it really is just a SQL query, but it takes like 20 or 30 seconds to type it in and wait for the search to render. But AI could do that asynchronously in 15 seconds, with the ticket pre-loaded for the person when they get in.

So it's kind of just: what's the lowest-hanging fruit? Whatever that is, every month or every week, we fix one of those things. I told the support team: we're going to automate everything over time, but we'll never really get there, because there's always going to be more stuff. We'll just start with: what are the top five tickets you get that you think could be automated? Then we'll just keep doing that over and over again.

With onboarding, for example, I just tell the product team, "Go sign up for Gumroad and make a list of things that are annoying." We don't even have to have a backlog. Every quarter, we can just run this exercise.

I actually think backlogs are not good because they create reasons not to do the important work that just came up. So I just tell the team: automate yourselves. What's annoying about your job? We have engineers; try to give me stuff that isn't automated. I really believe most things could be.

I think a couple of years from now, we might get to a point where I type in a Notion ticket and GitHub just opens a PR based on what I wrote based on the Notion card and the GitHub history and all the other stuff. We have 10,000 pull requests on Gumroad. I guess that there are many bug fixes or features I could describe, and ChatGPT could almost figure them out.

Why Shipping Product Is the Best Marketing Strategy for SaaS Retention

Scott: Sign me up for that. I think one powerful thing is that you've taken a lot of time to experiment, so you know what GPT can do. You've got Scribbler, you've got Ask My Book, where you can literally talk, and it's you responding, it's crazy. There's that experiment where you're personalizing the earnings creators make: "What can I buy with $400K?" Do you see this as one way to keep Gumroad in the conversation? Like, what's the zeitgeist? How are you staying relevant?

Sahil: I was thinking about this last week; why are we still shipping all this stuff? Because we don't really need to. But what I realized is: shipping product is marketing now. There's so much noise, there's so much fundraising, there's so many videos. All these ways to pretend to be doing stuff for the customer. But customers want to use good products that they think are getting better over time.

For me, shipping is like just reminding customers: if you use Gumroad, I guarantee it will get better over the next year. Every week, you're going to see a tweet, a post, an email, "this is better, this has changed, this is new, there's a new feature, there's Discovery, here's some AI." That will be constant. Partly because I just like building stuff.

That's something a lot of other companies can't really offer. When was the last time that the startup shipped a feature? It's hard. It's hard to hire really good designers and engineers and empower them to ship a lot of functionality. Tons of stuff just get stuck. So I think it's kind of like a long-term retention strategy: when you use Gumroad, you are using a startup. You're using a product that is actively being worked on. Kind of like when you go to GitHub and see "last commit: 2017." You're like, I'm not going to use that. Versus something with a merge from four days ago and 2,000 stars: you just trust it more.

Going back to what you said before: an advantage I have is that I will actually go use the technology. The Whisper API that just came out, the ChatGPT API. I'll actually go build stuff with it. Over a weekend, one Friday night, maybe during the day. And then you can take Gumroad and smash it into this thing with actual knowledge of the constraints.

Otherwise, people are like, "Oh, I can just say 'Create a Gumroad product to make me $1,000,' and it'll just do it." And it's like, well, no, that's not right. It's not a genie in a bottle.

But kind of, right? What you could do is potentially say: "I want to create a book," or ask for the name of the book, so you come up with like Fitness for Founders. And then you could hit Gumroad's database and say, "Okay, here are 10 other books on Gumroad, here are their names, here are their descriptions." And then you ask GPT: "Based on these and this title, come up with a product description for this book." And it would work. But then there's plagiarism. Creators are not necessarily opting into their content being used for these purposes. So that's why we haven't done it. But at least I can, within one or two hops, come up with a much more reasonable implementation of GPT that has a business use case and that I could actually ship.

Because often what happens is: "We need to enable secondary market transactions on Gumroad using Solana." And it's like that will never get shipped. We can get some designers to do cool designs, we can have engineers know the schema, but is that actually going to show up in the summer product? No. This is what often happens with startups: they keep getting obsessed with the shiny object and never actually ship any of the boring stuff.

I spent some time building Ask My Book, and I think allowing people to search the Gumroad library and ask their Gumroad library questions could be a cool use case. It basically saves engineers and designers tons of time. I'm the one-man prototyping machine, I'm the startup, and Gumroad is the company. I do all the experimentation myself and then bring the more fleshed-out ideas to the team. I can do that because I can design and code, which means Gumroad can scale all the way down again if need be. If I had to, I could let every single person in the company go, change all their hours to zero in Flexile, and all of a sudden, Gumroad is making $3 or $4 million more a year.

I feel comfortable saying that because everyone knows. I'm just super clear about what the situation is. When everyone has that clarity, people feel comfortable. Like: "Am I adding value or not? I have six months to prove it. I shipped this, but it didn't really work. Sahil knows it didn't really work." There's no fog of war.

Sahil Lavingia's Book Recommendations

Scott: All right, two final questions. What have you read recently that's made an impression on you?

Sahil: I'm currently reading The Snowball. Have you heard of it? It's like an authorized biography of Warren Buffett. What I've learned is that people present him as a very normal guy, drives a Toyota Corolla or whatever, but he's obsessed with making money. He is just really driven by being the best investor ever, from a very young age. Reading it, I'm like: oh, you just have to kind of be insane. You have to be the guy who, at a dinner party with your wife's friends, says, "Hey, I've got to go upstairs and read about this thing because tomorrow I have to wake up early to trade stocks." That made me feel good about my own weird social interactions sometimes.

I also recently finished Better to Have Gone, which is about this commune that gets set up and all the things that go wrong. It just gives me a lot more perspective. People always talk about exiting the US and creating a network state or whatever, and it's like, that's hard. The number of things that can go wrong. Someone falls sick, and you're like, "Oh my God, we need a doctor, fly in the doctor." That stuff is just: oh yeah, this is not easy.

Sahil Lavingia on Weightlifting, Strength Ratios, and Founder Mental Health

Scott: Last thing, doesn't have to be about business, could be a weightlifting tip. Something that's kind of changed your life overall, a high-octane tip you give other people.

Sahil: Oh, definitely. Weightlifting has changed my life. I got really into it last year after the COVID gym stuff opened up properly. I focus on a very specific metric, which is strength ratio: the amount I can deadlift plus the amount I can bench plus the amount I can squat, over my bodyweight. Trying to increase that as much as possible.

I love having something that I feel like I'm almost completely in control of. If I go to the gym and deadlift 350 pounds five times, three times, the next week I will almost definitely be able to do that with five more pounds. This kind of progressive adaptation. You really push your body to the limit, then you show up again, and you can do five pounds more than what was previously your limit. There's something really cool about that. The fact that your body physically changes that fast. You can now do something that literally a week ago you physically couldn't do.

There's something very nice about that when there's all this other stuff in your life that can go up and down. Knowing that you can just show up to the gym and do this thing and realizing that the other people at the gym, that's the work they've put in. There's really no skipping it.

And I think: just trying to spend more time in the sun was really important for me too. Being in an environment where you're using your whole body. I guess the macro tip is: we spend so much of our time in our brains and our eyes, looking at screens. But actually, we have hands and feet and tons of senses, and if you don't use them for a long time, guess what: they're going to be weird about it. They're going to say, "Hey, I want to get used; go use me, please."

I think things like depression are our bodies reminding us: "Hey, I exist. I need some sun, or I need to go for a walk," just like your dog gets upset if you don't take them for a walk. I think you do too.

Final Thoughts: Building Gumroad for the Long Game

Scott: So basically, the title of this episode will be: "Sahil Never Skips Leg Day."

Sahil: Leg day once a week. That's what I committed to. It's still my weakest exercise. I can squat 240 now. My goal is 315. It's gonna take a while. If I went under 315 today, I would just crumble. I just know I'd fall. But hopefully a year from now I'll be able to do it. Which is kind of crazy, right?

Scott: Awesome, dude. All right, well, thanks for coming on.

Sahil: Good to catch up. See you, man.

Don't miss out on future episodes, get alerts for new drops at subscriptionheroes.com or follow us on your favorite podcast platform. Special thanks to Churnkey for sponsoring the show. Learn how to make customers happier while boosting revenue at churnkey.com.

About Sahil

Sahil Lavingia is an entrepreneur, investor, writer, and painter. He’s the founder of Gumroad, which empowers creators to sell goods and memberships with little overhead. Gumroad runs as a lean, open startup with a no-meeting, no full-time culture, and now generates $20 million per year in revenue. He published his first book, The Minimalist Entrepreneur, in 2021 through Portfolio.

Links: