Reid Hoffman | How to Decide

Hi everyone. Welcome to Grey Matter, the podcast from Greylock, where we share stories from company builders and business leaders. I'm Heather Mack, head of editorial at Greylock. On today's episode, Greylock general partner Reid Hoffman talks with his Blitzscaling co author Chris Yeh about one of the most critical yet underanalyzed aspects of entrepreneurship, decision making. Entrepreneurs and executives will do all sorts of things to improve themselves, from taking a class on a particular technology, to reading a book about venture deals, to practicing intermittent fasting. Yet very few of them stopped to ask the simple yet powerful question, how can I make decision making better? Reid and Chris discuss in detail, providing anecdotes from their own careers. This episode was first recorded in 2021. You can read the essay based on this discussion on our website@greylock.com hi, it's.

Chris Yeh, the co author of Blitzscaling and I'm here with my co author and old friend Reid Hoffman, the co founder of LinkedIn and and investor at Greylock Partners. Today we're talking about instinct, data and decision making. Making data driven decisions is rightly a major emphasis in the startup world, but the reality is the data available to drive our decisions is often imperfect and incomplete. We touch on this in one of the key transitions of Blitzscaling, inspiration to data. I'd like to dive deeper into this topic. Why is decision making so important to focus on, both for entrepreneurs and in general?

You know, everyone thinks about decisions as the very macro, the very large decisions, which of course super important, start this company, not start this company, co found it with this person, co found it with this other person, shut it down or pivot. I mean those things and those are, by the way, super important decisions. And they tend to be the kind of decisions like for example, most founders in starting a company, that's many, many conversations, even if it's compressed in time in order to make things happen. On the other hand, once the game starts going, and sometimes the game is going actually even before you make a decision to start the company, actually, in fact, decision making is like competence, it's a sport and so forth and everything is there. So for example, it's not just do I start this company or not, Do I start the company with these ideas? And this entrepreneur, do I do this first? Do I raise money? Do I paper test it? Do I test it with customers, do I move in order to do it? Like there's all of these questions, which is decisions, decisions, decisions. I mean literally, usually there are multiple decisions per hour, not just these macro Ones that you're actually in fact making. And so even small differences, because look, one of the things you begin to be interesting is that compounding is say for example, you've increased your decision making practice and just in metaphorically, you've increased it by 0.5% in its utility function. Well, that compounds that may still be that these are the very big issues that you slow down on and everything else. So focusing on these decisions can reap benefits across the entire entrepreneurship practice, which is obviously fraught. And there are tons and tons of decisions to make.

And it's pretty remarkable when you think about all the different things that entrepreneurs and other folks in the ecosystem do to improve themselves. They'll take a class on, on a particular framework or technology, or they'll read a book about how to choose their venture capitalists or how to think about startup investing, but no one stops to think, how can I make my decision making better? And yet, as you point out, it's the fundamental underpinning of everything.

Yep, exactly. Because the thing with these decisions is that sometimes the small decision that you're making right now gates this much larger decision. And people's normal instinct when you think, oh, it gets to a big decision, is to slow down, talk to a lot of people, you know, get emotionally comfortable with it. These kinds of things, which are almost never the right kinds of answers within the entrepreneurial circumstance that there need to be other ways that you're improving your decisioning, that you're making decisioning and a effective sport in what you're doing. And that's actually part of the reason why it's one of the central things and in entrepreneur skills is that you learn entrepreneurial decision making and how to make decisions in the kind of constrained. And you know, one metaphor of course I use in entrepreneurship is assembling a plane after you've jumped off a cliff. But another one is you're running across a rocky minefield at night in fog, right? And it's how do you make all those decisions? And the decision about how you're proceeding and doing that the right way. And that decision making calculus is the key thing that I actually look for in funding repeat entrepreneurs because if their decision making is getting a lot better in terms of what they're doing and of course applies everything from hiring to, you know, decisions in the business and, and all the rest, if their decision makings get a lot better, they're much more likely to be a good bet, a better bet the second time go around.

And that's why we're really looking for people who are decisive. It's the same route, right? People who are good at making good decisions quickly under those circumstances as they're running through that minefield. And of course, the minefield is even more difficult when you're blitz scaling, which means that you're prioritizing speed above everything else. You're not just running through minefield, you're on a dead sprint. So why is decision making so hard for blitzscalers and how should they think about the right way of making decisions while blitzscaling?

This is one of the things that I really learned from PayPal and my experience at PayPal because, you know, I thought I was really good at decision making, partially being smart, thinking of it as sport even before I started entrepreneurship, then got a whole bunch of experience with social net and made some decisions too slowly, made some decisions badly, you know, and what I always do is I reform my decision making framework. I didn't just go, oh, I trusted so and so. And that was wrong. It was like, no, no, take responsibility for every decision you made. What were the set of things that you should do differently as playing the game forward as a sport? It's like, you know, how you actually in fact act, how operate, what do you do? What are the things that make this different? And in blitz scaling, it's really hard because not only of course are you intensely compressed in speed. Usually you're there in a competitive framework. Because part of what's gotten you to blitzscaling is because there is either extant or could be extant competition in a really serious way. You've got this high stakes part of it because it's kind of like, well, this is my economic and emotional well being in this. It's done for the whole group. So like a lot of people all in the same boat together. So like row left, row right. You know, like, you know, that kind of thing as part of it. You don't really have a lot of chance to communicate, don't have the time to all get on the same page and communicate. So you have to make those decisions and operate in that framework where, you know, like, ultimately that's part of the reason why these decision making frameworks like Daisy and Raci and other things are so important and so often, you know, used by entrepreneurial companies and blitzscaling companies because it's like, okay, so and so is making the decision, let's go right? That's how the decision make. But also how does that work out? And then of course, you know, while you would hope for all of the relevant data to make the decision. Frequently, there are obvious massive blind spots in your data for decisioning. Even if you do have some, you're fortunate enough to have invested in or tripped across or made some data in order to make it happen. And that's what makes decision making so particularly hard in blitzscaling context for founders, executives, managers, contributors, and so on.

Well, let's talk about one of those key dilemmas that blitzscalers face, which we describe as the transition from inspiration to data. And the key question is, how should companies manage that transition from going off their gut instincts in the early days to being able to really incorporate data later decisions, and what are some of those key markers along the way?

So one of the key things to always understand about scale and even blitzscaling is that ultimately, as you begin to get to scale, you must have data. There is really no universe in which you don't actually, in fact, start harmonizing and operationalizing your entire company with data in order to make scaling work. Right? So, for example, to some degree, one way to looking at what the drumbeat of an organization is, is what dashboard and meetings around that dashboard are you managing the company to? Because that's way you collectively make decisions together. That's the way that you understand are you making progress or not? That's the way you say, hey, did this decision or this force or this thing that's going on, how did that affect the. The okrs, the numbers on the dashboard and kind of what's going on with that? And so your plan is to get there now the challenge is to say, you know, when do you start doing the instrumentation? Because you can instrument, and then obviously you don't have product market fit, and you've got all this instrumentation, no product market fit. You're not getting a data, it doesn't really matter. That was wasted energy to put all that time in instrumentation. On the other hand, as you start scaling, you know, everything from the question is which channels are working? If you're spending money, which spends a working and being late to instrument. And to know this data can be really a problem. Now, by the way, again, you don't suddenly flip to all like. So, for example, one of the really early discussions we had at the Airbnb board meeting was one of the Airbnb board members was like, well, we should know what our operating margins are, right? Like we should be instrumenting what our operating margins of the businesses. And this is one of the places where I spoke up and I said, no, actually, in fact, that's not the case. Because look, this is a digital property, digital marketplace. You know, ultimately at some scale the margins will be fine, whatever the margins are. And maybe they're way to improve or not. You know, I get it in the semiconductor industry, if you don't understand your margins or a hardware product, you don't understand your margins in the very beginning, that's a problem. So in that business, understanding those. But here that's mistaken instrumentation. And matter of fact, even as you're starting to focus on it now, it's the wrong focus right now in order to be doing this. And that's kind of like one of the areas where you have as kind of as a board discussion and so which things you instrument when and some things deliberately later instrumented is the right thing. Now that being said, you say, well, we're instrumenting growth and we're just intermittent growth. And we know the growth and we know the growth's working, we know the channels are working. What else do we need to instrument? And by the way, in order to get the growth, usually you have to have the instrumentation, you have to know which things leading to what, more customers, more members, more engagement, you know, et cetera, more time on site, whatever the thing is, you know, the set of things for this. But growth can also mask trouble. The most classic one that most people talk about is leaky bucket, which is, hey, you got this huge input into the thing, but it masks what your churn is and what's going on. Another version of that is what your cost of acquisition is over a long term value. Now of course, by the way, sometimes measuring long term value too early can be a real problem because you go, oh, it's this and it's actually nowhere near that. Because measurements end up becoming the proxy for truth. And you always have to be, when you're looking at the numbers, be thinking, look, not just how true is that now? But how true is that going to be three months from now? How true is that going to be 12 months from now, 36 months from now? And what are the things that will change that truth? Because these are not like, you know, these aren't things like you know, Planck's constant or E equals MC squared or these are truths about the operation of the business where the market can change and your channel acquisition can change and everything else and the number can go, you know, you can invent a new business model or you can add in additional revenue stream and that changes all your numbers. So you've got this invention calculus that also goes along with it. Now, part of the reason why you're instrumenting is that the key thing in almost all blitzscaling is you're really getting to compounding growth, compounding loops. Sometimes it's straightforward and direct, like a virality curve, like what we were doing with PayPal or instrumentation and Airbnb or LinkedIn or Facebook. But sometimes it's also, by the way, loops within the playbook. So, for example, how Uber was looking at getting into different cities and what the playbook launch for launching a city is and what is the look of a city health play to. Now, all of this gets back to what you need to know from the very beginning when you're going to go into a scaling and a blitzscaling company is you're going to ultimately be running this with dashboards. And so starting to think about what your dashboards will grow into being from the very beginning is the right thing, even if you're not fully instrumenting the dashboard. And one of the questions you need to be thinking about is all which sub instrumentations do you need, which other things so that the numbers that you're resolving on for the dashboard are the real numbers and there aren't hidden massive problems which make you under deliver your opportunity. Now, sometimes, of course, you can have supporting dashboards which are the, well, everything working normally is like this and this is like the, you know, looking under the hood in the car and is the engine running well? But you know, that's the kind of thing you need to do. And that all is part of this entrepreneurial journey where you're moving from the I have an idea to running it as a real business through this process sometimes and more often in technology companies.

Blitzscaling and what I hear from you is it's really important to both be very rigorous. On the other hand, still have some humility around the power of data. Data is so absolutely essential, but you have to still make sure you're getting the right data. And you have to understand that the data that is relevant may change over time and that you may still need to make decisions rather than just basing everything on the data. But that is really one of the fundamental dilemmas, which is I'm sure there are times in your career when you've had to make a decision based on gut instincts. But how do you tell the difference between a gut instinct that's helpful and something that might be wishful thinking based on really desperately wanting something to be true?

So this is one of the reasons why, as a general way of strategically operating an entrepreneurial Business or matter of fact, most projects. What I recommend is this framework, you know, Chris, which is having an explicit set of bullet points on an investment thesis that gives you a set of things that you can try to measure. Now you won't be able to get data by that, but part of by the way trying to measure that is talking to your smart friends and getting feedback, asking them a question of what, what won't work here, why will it break, why will it fail? And then getting either a plan for that that you think is contrarian potentially and right, or, or something else, and making that work is partially what's key to, to evolving and investment thesis. Obviously the more that you have useful data to tell it will be have a higher degree of accuracy to inform your decisions. Now that being said, frequently, most often you have no data. And so then you go, okay, how do you make those decisions? Well, there's a few things that are beyond this kind of investment thesis framework and they're related. So one is first have a network cross check. So part of the reason why going and talking to everyone, smart new network. Part of the reason why I recommend entrepreneurs to not kind of say, hey, I'm going to hold everything super close to my vest and not talk to anyone. And the invention, the secret idea is the secret sauce. But actually go to talk to people. Now you don't publish it because you only talk to people when you get something valuable back in the conversation. But your competitive defense is you're the person who's moving on it, you're the person doing it. So by the way, you don't go talk to your competitor, you don't go talk to like, and you don't publish it because you know that, you know, gets it into the hands of other people who might be directly competing. But you go and talk to smart people and you ask the questions about like, what's wrong, what's broken, et cetera. You know, what could be better as a way of doing this? And that's one thing that can help you do it. Because if people also share the no, no, that sounds like a good bet, then more likely you're on a helpful gut instinct than a wishful thinking. Now the other thing is to part of a classic decision making theory which is really good to look at this kind of stuff. There's an excellent lab at the Harvard JFK School run by Professor Jennifer Lerner and other things, which is what are the decision making biases? It could be a sunk cost bias, could be a confirmation bias, it could be a Set of different things that, that say, okay, where might biases be leading me towards this being wishful thinking. Because if you've done the cross check and then you still end up with look, I think this is a good bet. I think I've got a good idea here, a good theory of the game, a good activity, then that can still be the right thing to do.

Now I'm really interested in this notion of the biases that can creep into our decision making because I think it's true that we can always run into pitfalls, blind spots, biases, things like that. What are some of the notable examples you've seen, either from your own career or just the startup world in general, of the kinds of biases that we really have to watch out for? And what are some of the things that have happened when people fail to watch out for them?

Well, there's this concept from science and psychology called type 1, type 2 errors. Type 1 errors are false positive, type 2 errors are false negatives. And you can use data decisions and get to both type 1 and type 2 errors. In terms of what's going on now, one of the ones that I had a little bit of personal experience on because Greylock was in a Vestron Groupon was how overheated the channel was with a belief that the, oh, look at the growth curve. You just project out the growth curve and you think this is going to be really amazing. And they hadn't fact that they had such a massive and effective and high driven sales engine that they were essentially buying all of the one off products from all of the SMBs that wanted the customer growth in a way that wasn't good for the SMBs in a sustainable way, wasn't good for the customers in a sustainable way and was kind of like a flash, you know, kind of sale that wasn't particularly good and wasn't particularly sustainable. And that was an instance obviously of a, of a false positive. Because you look at the numbers and you go, oh my God, I've never seen a business with, with numbers like this is really amazing. And they hadn't really done the. Well, where does this really lead to sustainability? What is, what are the parts that of this bucket that are deeply leaky, you know, et cetera, et cetera, you know. And another one that was really compelling but interesting is like one of the things that you would frequently do in these data analyses is cohort analyses and one of the things that when in the investment decision in Dropbox was that the cohorts actually got structurally better over Time, that was a very super rare thing in businesses because usually what happens is on a cohort, usually asymptote like time or customer or region, asymptote dots down to a kind of a persistent thing. And what you're doing really well is where your new cohorts have the same asymptotic leveling that the old cohorts do. And then that really adds into a very good business. Now Dropboxes was increasing and that's really great. The challenge was overly generalizing to how much that would lead to a bunch of other products. Right. And what it ended up being. Now we were very happy with investment and both Groupon and Dropbox Greylock made money for its LPs because we invested in early enough in the compounding cycle. But both of them could have, I think been much bigger than they ended up being, maybe because of an over reliance upon just, oh, this data will just carry us there. Versus there need to be reinventions that happen in order to make that work. Now most often mistake that I see in general startup world is to make claims of organic growth versus instrumented. And I'll take a personal failing on this case, which is I had invested as part of Greylock in an educational company called Edmodo, which is still going and I think helping teachers and students now owned by Netdragon. And I think they're doing a good job with it. But part of the reason it wasn't an independent company was because we had presumed, because the folks had seen LinkedIn and Facebook, that they had instrumented their viral growth the right way and they'd instrumented how this would spread between teachers and students and all the rest. And that we thought this would be a new social network platform primarily empowering high schools and below and maybe at some point get to colleges and so forth. And the problem was that they had organic growth from teachers who were just looking for a free LMS tool, but they didn't actually have instrumented growth. And once that organic growth dropped away and they realized they hadn't been doing any of the instrumentation that these kind of classic social networks do to build the dashboards and then build the features that support the dashboard, which can, by the way, sometimes irritate users like notification emails and so forth. So, hey Chris, you've got a new connection on LinkedIn. It's like, oh, I'm being spammed. But those things actually play into the re engagement and so forth kind of as a way of operating. And that Edmodo hadn't done any of that. And that was one of the things that kind of led to it being better to combine Edmodo with another platform. You know, all of the excellent work that netDragon is doing versus being an independent one.

You know, what's fascinating about these examples is that in each of these cases, and you talked about Groupon and Dropbox and Edmodo, in each of these cases, there was data, there was a metric that was doing really well and that gave you a lot of comfort, led you to believe, hey, this is a great investment or this is going to go on. And again, let me emphasize that you talked about some investments that made money by going public, and you talked about an investment that was then acquired by another company. So it's not like these are total disasters, but it was the case that you had this data that looked great, but you didn't go into the next level and really say, well, what's behind this data? What's underlying this data? What are the different things around it? And I think there's a lesson there around data doesn't have to be wrong in order for it to mislead you. And it's up to you as the decision maker to actually make sure that you have the data that you need to accurately assess the situation.

Yeah, that's exactly right, Chris. I mean, I think the things around understanding data, too often the simpler thing is once you have data, data speaks for itself. Data is its own fact. You know, data is like EMC squared or 2 plus 2 is 4 and that kind of stuff. And actually, in fact, data is almost always in these contexts, something interpreted doesn't mean that there isn't some deeply factual nature to it. How many customers you have, what is your revenue, what is your revenue for this quarter, what is your revenue for the last three quarters? But you go, look, I've had this crawling growth for revenue for the last three quarters. What's Q4 going to be? Well, you say, well, I predict off these lines and all the rest. But you go, and it's the same organization, the same sales pipe. You go, look, that may be a very high confidence prediction. That's great. That's data you may factor into your decisions, factor in your, your budgeting, like what you're going to do for cost structure, factor in other kinds of things. But you know, the fact is how you understand the story, what it's moving towards in the future, you know, all of that plays into it. And so having always an active like kind of theory of the data, perhaps the one that I actually Thought was kind of most interesting on a LinkedIn basis was something I observed from Jeff Weiner, which was really stunning and amazing. Jeff is such a. Calls himself an infovore, but he's also a datavoor. And so part of how he managed LinkedIn extremely effectively was to have a bunch of different dashboards. And he was such kind of a maestro, a conductor of these dashboards, that one of the things I saw happen multiple times is he'd be looking at a dashboard, look at a number and go, that number. Something's wrong. That number should be different than the number is. Call the product manager, the general manager in charge of that number, and say, something's wrong right here. Like this, this number, like, being. That the number is being reflected. Something. Something's wrong, right? And always, you know, in these cases, he would be the first, like, noticer of, oh, look, the email notification system is down, or deliverability changed, or the instrumentation wasn't checked in the right way, or the log file was delayed and it's being processed or something that was like, oh, that was wrong. Because he was like, okay, this is what I would expect to see from all this. And since I'm not seeing it, something has happened here. Most often it's in our internal systems, but something else, and that's paying attention. But by the way, that's because he has an active theory of it. He looks at it and goes, okay, this is my active theory about how all these numbers work. This is what's happening, this is what's going. This is the range of things. And if I don't see it, I then pursue and I update my theoretical model of what's happening, of how the system works, of what the game looks like, et cetera.

And that story about Jeff and his datavore tendencies is fascinating because I think it really illustrates something which is Jeff may have started off with having a feeling, a gut instinct, hey, that looks wrong. But he wasn't content with just sort of saying, hey, it feels funny. He's like, let me figure out what is the actual model that exists so that it's not a funny feeling, but rather an actual deviation from the model that would ultimately attract his attention.

Yes, exactly. And so it's like understanding the data. And that means that you always, with data, have interpretive overlays on the data. And by the way, there's always some question about the. Not just the future accuracy and predictability from that, from this data, but also like, okay, well, what goes into that number? What caused it? Like, you know, there's accuracy questions. But then there's also a, well, sure, that was the actual number of people signed up, but by the way, you were not accounting for the following thing that was happening that caused that number to be what it is now.

In your day to day life as a venture capitalist, I'm sure you see a lot of pitch decks. And in these pitch decks there's all these charts and graphs that use macroeconomic data, industry trends, research reports. This kind of data is readily available, although imperfect at times. Have you used this kind of data to drive important decisions? Or is it just something that people put in pitch decks to convince venture capitalists to invest?

Well, look, for example, if you don't ultimately not going for a large market, it's not a venture return, right? There's a bunch of other characteristics that also lead to venture returns, network effects, other kinds of things. But ultimately, if you don't have a sizable market, you won't have a venture return. Now the question is, what are the tells about it being a sizable market? So I tend to think that a lot of this macro kind of data stuff usually engenders a bunch of mistakes. It can even engender mistakes where you go, oh, we predict that the total addressable market, the tam, is too small and actually, in fact, it's bigger than you think. This is like the canonical case in my head is Uber, which is the, oh, it's the black car market. And you're like, well, that's not a very interesting market. But actually, in fact, when you go to. No, no, actually, in fact, it's completely reimagining passenger transport and making taxis work the way they should. If they weren't this constrained weird regulatory thing by which cities are, you know, both managing them, but also extracting the rent and therefore between the two, limiting innovation, that this is the way to have a highly active train transport into a network and then that changes your TAM altogether massively for the up. So I tend to think as an investor, paying detailed attention to these numbers is usually wrong. What you are trying to do is get a sense of, well, okay, if the number is high, do I believe that there's actually really that kind of market there? Or is this some weird I bank or analyst? Just like I called 100 people and they all said they'd buy this. So 100 times, 340 million people live in the country, hey, it's got a huge market and not something kind of dumb like that in terms of the high side, but also sometimes too low. Like, okay, do I think this will actually grow to being an interesting market. And sometimes, you know, especially in the kind of investing we do at Greylock with a seed series a series B, it's the, okay, what might get added to this, what might change the TAM number and so forth, and could it grow in some interesting way and kind of lead to somewhere kind of interesting? So I broadly ignore those numbers. I think most of my partners do too. But we're always paying attention to, you know, is there a scalable market here that's reached on a risk adjusted basis? Because those are the venture companies.

Now, we've talked about some of the very common data sources. I'm also curious about the opposite. Which are some of the more unusual or uncommon data sources that you've tapped to get helpful data? I know there's all these classic apocryphal stories about going to visit the parking lot of a company on a weekend and count the number of cars there. I don't know if you've ever dived in someone's garbage or anything like that. I. What sort of interesting, unusual things have you done?

I have definitely not dived into anyone's garbage. And I guess there are some universes in which that's legal in some way, but, you know, whatever. Still seems low integrity at the very least and smelly. Yes, well, but, you know, ethics and integrity. Look, I think the most interesting thing to do is to pull together synthetic data in various ways that's completely legit and gives you a signal where most people don't. So actually, in fact, you can buy data on various kinds of credit card receipts and, you know, buying patterns of credit cards and buying patterns of credit cards. So things. And so buying that data and laying that data against your investment decisions in various ways is kind of a smart thing to do. Similarly, both inside LinkedIn, various decisions we made, and outside doing analyses of job postings like, so you can get a sense of, for example, basically most companies, because they actually need to hire the talent, will do job postings. Those job postings will give you a sense of what the major projects are and what the details they're working on. Like, for example, like you said, an investment decision. And you're looking for investing in, you know, productivity tool like engineering developer tools. Well, what are the, what are the job listings? Say, you know, kind of list and kind of aggregate. So they're another source that you can actually in fact do analysis for that can give you useful things. There's obviously various insights on mobile, various companies you can buy them from and find out, you know, downloads and other kinds of things. And in order to get trends and all of these things are useful, you always have to have kind of a theory about, okay, like again, as we've been talking about this entire podcast, you know, what's the uses and limits of the data, accuracy, predictability, et cetera. And then in investing, you always have to think, you know, and frequently, of course, you have to think about this in entrepreneurship too. Is competitive edge like, well, where do I have, you know, is this a differentiable competitive signal that I have and other people don't have that makes me better at decisioning this. Do the investment after the investment pay a higher price, not pay a higher price, et cetera, et cetera. And you have to think about where does your data get you better decisioning signal than your investor competitors?

One final question for the podcast. I'm going to take us all the way back to the beginning because we began talking about instincts and inspiration versus data, and so thinking about companies like LinkedIn or Airbnb, where, as you mentioned, there's now plentiful data, vast amounts of data, terabytes of data. What kind of role does inspiration continue to play even when there's data aplenty?

Well, not surprisingly to you, Chris, and not surprisingly, probably the close listeners of this podcast, the answer is always there is the theory of the data. Because all data is interpreted the kind of, where does that data go and how does it, and how does it operate? You know, what does the data mean in terms of trend line? Like, what does that data mean in terms of the future and what's the theory of the game and how that future is changing, inflecting up, inflecting down, stability, et cetera, et cetera. Then there's a question about, like, what are the things you can do that change the data curve? Like, here's a new marketing method, here's a new sales method, you know, here's a new virality method, you know, you know, here's a new support tool that reduces our customer service costs. Like, so, so all these numbers are numbers we can operate on in various ways. And how did that change those numbers in the machine? And so basically the answer is inspiration. Innovation always play a serious role. Now it's an important thing to asterisks. And back to your earlier question, not to just kind of go, well, we're going to change this and our margin structure is just going to be better. It's like, okay, no, we have these three ideas about, it's a little bit like the mods here in inspiration are like the ABC planning from the startup of you, which is, look, I got three ideas. The first one doesn't work, I'll try the second one. Second one doesn't work, I'll try the third one. I'll try refinements of them, you know, as a way of getting there. But I've got an active plan for how to do it and that's a good enough reason. But you still have to pay attention to Is your active plan a good one? Is it the right one to execute on? And as you're executing on it, is it working? Are you refining it? If it's not working, are you pivoting.

And adjusting well, Reid, as always, thank you so much for taking time out of your busy schedule to share some wisdom and insights. That concludes this episode of Gray Matter. You can subscribe to gray matter on soundcloud.com greylock partners. You can also find new episodes and blog posts on greylock.com youm can follow Greylock on Twitter reylockvc. I'm Chris Yeh and on behalf of Reid Hoffman, thank you for listening.