November 10, 2016
Ep. #5, GraphQL At GitHub
In Ep. #5 of JAMstack Radio, Brian and Ryan talk to GitHub Platform Engineering Manager Kyle Daigle about what makes GraphQL such a great qu...
I was just telling the guys that about 80% of the time that I do this talk, it goes pretty well, 20% of the time it's a train wreck, so we're hoping for the 80 tonight. But the thing that makes it good, when it is good, is when you guys speak up and when you talk about situations in your real life that might push or pull on some of the ideas that I'm going to share with you. So please do that, and that's why there's mics there.
Let me give you two seconds of context on this talk. I started teaching product design and development at Stanford in 2006. When I started doing that, a few students would say, "Well, how should we set a price for this product?" And I was like, "Oh, well maybe we should think about it this way," because I had run pricing at eBay for about 6 years. And during that time, I learned a lot about pricing. I screwed up a ton of things, almost killed the business twice. So I learned a lot about pricing.
Then that pattern of people asking me about pricing kept repeating. It took me about 6 years to realize that pattern existed, until I distilled what I thought about pricing down into a one hour or forty minute talk. That's what this is. I've used it with students, undergraduates and graduate students in the School of Engineering at Stanford, business school students, startups, founders and practically any domain you can imagine, whether that's a physical product for consumers, a software service for businesses. These ideas actually port really beautifully to a lot of different situations, and that's why I think they're pretty useful.
So why don't we dive in and we'll explore a little bit about pricing together over the next forty minutes or so. I want to start with the idea â€” what really stuck with me was when I realized that, holy smokes, most people think pricing is a math problem.
Most people think that pricing is a Rubik's cube with an answer, and it's not. It's actually a judgment problem.
What hangs people up, and why companies really struggle with this issue, is not because they don't have smart people who know how to do math, or know how to draw a demand curve, or even know what that is. It's because they're missing some key judgment tools; some tools to improve their judgment. That's what this whole thing is focused on. This whole talk is focused on pricing, not a math problem. It's a judgment problem.
And to start this out hopefully on a pretty easy foot, we're going to talk about toast. So, what I want to do is I want you to imagine the perfect slice of toast. And I want to ask you, how much would you pay for the toaster that would give you the perfect slice of toast?
Audience: 20 bucks
20 bucks. Most common answer I've ever heard is 20 bucks. Some people say 30, some people say 40, but 20 is a very good guess. Now, what if I told you that that toaster had a mechanical timer that allowed you to get exactly the right level of golden deliciousness that you wanted on that piece of toast?
Do you see what I'm doing? I'm reaching into your brain, and I'm playing with your idea of what is value. I'm moving around the assumptions and beliefs that you have about what makes for good toast, and what makes for a good toaster. And I'm going to do it again. This toaster that I want to sell you has faster warm up, and that ensures that there's perfect browning from the first slice to the last slice. You know, that terrible situation where you toast the first one and it doesn't quite toast. And you toast the second and the third and they turn out to be charcoal? This solves that problem completely.
You feel what I'm doing, right? I'm going into the command line of your human OS and I'm playing with it.
I'm starting to tap into ideas that are very deeply embedded in your brain. These aren't things that you were taught, these are things you have instinctively in you as a human being, who makes discriminating choices between one thing and another thing that might be better.
I'm going to keep doing it. This toaster has defrost setting for frozen bread before toasting. This is getting pretty good. The bagel setting allows you to warm one side and toast the other side. Now how much would you pay? It's not 20. It might only be 21, but I bet you there's somebody in here who would pay 30 or 40. This toaster has an award-winning Pro Heat â€” capitalized P and H â€” elements that are guaranteed for two years. This toaster is superlative quality, hand assembled in England. That's a sentence you don't see very often, at least in this century. It's iconic design, it's made of stainless steel, it cost the manufacturer $130 to make this toaster. Now how much would you pay?
Audience: 50 bucks
50 bucks, you are definitely not the target. The target is different than you, and we're going to talk about that in a second. This is the Dualit brand toaster, and it's available at William Sonoma. It's the same toaster that Queen Elizabeth has in her kitchen. Here's what it looks like. How much would you pay for this? How much would you, if you channeled your inner shallow person with a self-esteem problem and a nice kitchen. If you channel that type of person, how much would you expect to pay?
120 is a common guess. Another? I'll tell you this toaster retails for $260. And that is a beautiful thing. The fact that this exists, and the fact that you can build this slide with these points, and the fact that somebody will take out of their wallet, happily, the means to pay you $260 for this thing is like a sign that everything's actually going to be okay in the world. Not obviously because toasters matter, because they don't. And not obviously because these particular individuals matter, but because it gets at how the typical person, the typical human who makes these decisions can be made happier, can be made better in their daily lives. Their daily lives can be made better.
So how do we think about pricing? What we just did is an illustration of why the standard model is wrong. The standard model being, the model that we all learned if you took an undergraduate economics course, that was the standard model â€” Microeconomics 101. We're going to touch on a couple of these things, but everything we just talked about demonstrates why that is not enough. You remember downward sloping demand curves. You might remember this idea of elasticity, and we're going to talk about that in a second. You might remember supply curves, and the marginal cost of production, and this very concrete notion of buyers and sellers of rational economic actors.
But the standard model is really very poor at managing pricing, particularly in new markets where demand is highly uncertain, particularly in software digital services where the variable costs often approach zero. And especially in markets where humans make decisions based only partly on reason. My argument is that's most markets. So how can we think about it in a different way? Well, the behavioral economist, the psychologist who migrated to the econ department, give us some really useful tools here, and we're going to explore those a little bit more in a second.
Let me skim across the top and tell you that the behavioral economists make a great compelling argument for a two-part reasoning system.
The first part might actually be that rational decision-making analytic machine, but overlaying that is actually an intuitive judgment machine that makes decisions very quickly and without a lot of evidence or thought.
It's that combination, that outer shell, of intuitive judgment kind of surrounding the rational interior core that we want to talk about. What we're going to do is use four very simple tools to improve your pricing judgment. There's a lot of other ways to approach pricing, but this way happens to work. This way leads you to a path where you focus at the beginning of the process on mapping the substitutes for your product.
You have to understand the substitutes and complements of the product to get started. You design for both the intuitive and the rational portion of a person's brain; the decision-maker's brain, whether that's an enterprise, a small business or an individual. You create value-based prototypes rather than cost plus prototypes of your pricing. And lastly, you build assortments.
Always, always, always build assortments to probe the demand curve and increase conversion.
So what does all this mean? Let's rewind to undergraduate days and if you took an undergraduate economics class, this is maybe a little bit of a trip down memory lane. If you didn't, don't worry, you don't need to; it's too late anyway. The kinds of people that you run across when you look at the standard model are these four: Three are dead, Paul Samuelson is still alive. Adam Smith, Daniel Bernoulli, Milton Friedman, Paul Samuelson, these are like the gods of the classical model.
What they helped us understand in the beginnings and the roots of economics was this very elegant model of supply and demand magically coming together. In this model â€” and we'll make all these available to you if you really want to have it so you don't have to feel compelled to draw it or memorize it. In this magical land of price versus quantity, there's a place where the suppliers are happy and the demanders are happy, and that becomes the market clearing price.
Underneath that standard model, is some really nice feature. There's a nice feature of the demand curve in that it's downward sloping. All that really says is that at higher prices, less is demanded and at lower prices, more is demanded. A downward sloping demand curve is a fundamental part of its standard model. Another compelling part of the standard model is this idea of elasticity. That is you can, if you can draw a demand curve, actually measure the percent change in quantity as the price itself changes. So this is really satisfying, elegant, very nicely packaged; unfortunately, it's highly incomplete.
The nice features though of the standard model exist on the supply side as well. The supply side, if you were able to zoom way in on a supply curve and uncurl it, you'd find marginal cost of production and variable cost per unit buried in that supply curve. That's all great stuff and really worth a refresher once in awhile.
These key assumptions that this is somehow a road map or a guidebook to determining pricing strategy is just not helpful when it comes to actually managing pricing.
The nice feature of the supply and demand model is that these buyers and sellers are sort of rational. That they are using reason and analytics to make their decisions on the margin and doing so with perfect efficiency.
The problems with the standard model are obvious, particularly in new markets, particularly in software.
Particularly interesting and really interesting is that the disutility people experience in real life from a loss is much greater than the gain in utility they get from a similar sized windfall.
That's an important idea that the behaviorists add to the equation.
Let's actually dive in and talk a little bit about two of these behavioral economists. Actually, they wouldn't call themselves this. I think they'd both self-identify as psychologists. This is Amos Tversky and Daniel Kahneman. Kahneman won the Nobel Prize in 2008, I think, for economics; first psychologist to ever do that. Tversky would have won as well, but he passed away before the prize was given out and they don't give the prize to the dead.
So Kahneman and Tversky â€” pioneers in the idea of understanding how people make decisions. Based on many, many thousands of observations, they came up with this really compelling model. The model says that we have a two-part decision-making machine in our heads. The two parts are what they call system 1 and system 2, but we'll call them intuition and reason. And system 1 operates at a very fast pace. It's basically the equivalent of trusting plausible judgment.
I encourage you, if you really want to dive into this, go read Kahneman's speech that he gave when he won the Nobel Prize. It contains a beautiful description of this two-part model of decision-making. Like I said, it's fast, it's basically trusting plausible judgment. It's almost an automatic reflexive kind of an experience. It's where your brain quickly accesses prototypes or analogs that can help it make sense of a situation that it doesn't understand. It's a very, very low cost machine. It runs without a very high cognitive load, and it also happens to be prone to error.
The second part of the decision-making machinery in your head, Kahneman and Tversky describe as: many ways the opposite or the complement, careful, very deliberative. It moves more slowly, it's much more effortful, it takes a toll. There's higher processing costs, higher cognitive load, and it's really good at monitoring the flaws of the first system of the intuitive system, but only when you call on it. Only when it's prompted to do an investigation. Otherwise, it will sit there and defer judgment to the intuitive part of the brain. This is essentially why these people are rightly revered because they said hey wait a minute standard model, it's actually much more complicated than that.
How would behavioral economics help us in this situation? Well, this is sort of very rough idea of what a decision tree might look like for a consumer. It could be a business, could be a large or small business, could be a consumer, could be an individual person. If they're confronted with the choice to buy something or not, Kahneman and Tversky would say if they were here, the first thought that occurs to them is, "What does it remind me of?" Because it passes through that intuitive part first. Kahneman and Tversky would say that the intuitive part of the decision making comes back and says, "Oh, it's kind of like that other product that I bought that time. It reminds me of that."
Those memories, those analogs or those examples that people have at their fingertips are either substitutes for this product or complements to this product.
On the flip side, if they can't come up with anything, nothing comes to mind; they flail around for tools, they look for stored analytic techniques that they might use to make the decision. In this place, the intuitive brain has failed â€” the intuitive part of the decision making has failed to give you an answer or a hypothesis. So you turn to the analytic and rational brain and you call on some of these stored analytics. These analytics might be things like: a break-even calculation or an expected value calculation, a simple spreadsheet budget, a net present value if you really wanted to get fancy. But these are the stored highly accessible analytics in the rational part of your decision maker.
The big ideas here are: substitutes, complements, and stored easily accessible analytics.
Those are the kind of things that we're going to keep coming back to here.
How do we sort this out? My proposed synthesis is that we keep the core truths of the standard model. Demand curves generally do slope down. Supply curves are a function of marginal costs of production. There is an idea of utility in all this. But if you add to that the insights of the behaviorist, and you add to that some practical tools, even at the level of a simple worksheet that you could work through a pricing problem on, you get a much better way to manage pricing. Again, just to recall the very first slide, the assumption is that pricing is not a math problem, it's a judgment problem.
The Four Tools. The first is to map your product's substitutes and complements. They anchor your buyer, again, whether you're talking about a consumer, a business large or small, they anchor your buyer's thinking in the intuitive brain. The second is that we design a pricing hypothesis, or a pricing prototype, with the buyer's intuitive and rational sensibilities in mind. Remember, every single decision maker, if Kahneman and Tversky are right, and a lot of people think they are, everybody's walking around with that intuitive processor tied inextricably, maybe you can say handcuffed, to the rational processor. So, design with that in mind.
Third is prototype of value-based price for your product. We're going to talk about what value-based pricing is instead of cost plus pricing, and lastly, assortments. Assortments of good, better, best help you probe the demand curve for whatever your product looks like.
I guarantee you the only way you're going to know what your demand curve looks like is by probing it with assortments. You are not going to be able to sit down with a spreadsheet or any tool and mathematically calculate your demand curve.
Substitutes and complements. These are those highly, very powerful anchors to your buyer's intuitive processing. The question you have to ask yourself is, for a substitute, what would the buyer do without your product? How would he or she solve this problem if your product were illegal or didn't exist? The most common answers obviously are: The most common substitute is doing nothing, living with the pain of not having your product. Another one could be buy a competitor's product if there's a direct one. Jury-rig. A do-it-yourself solution. Eliminate the underlying problem. You get the idea.
On the complement side, the test for you as a founder is to say, what are the other products I might expect to see side by side in my user's daily life? An example here is, left shoes go with right shoes; but probably more compelling for this group is people who have a Heroku account probably are also on GitHub.
You can probably go through in your own mind and say, for my users, what are the associations and complements that I see in their daily lives? I guarantee you there are some really compelling a-ha's in there. We're going to do an example of the toaster in a minute and I think it will help flush this idea out.
The second tool in the tool kit for you is to map out their intuitive and rational thinking.
For your user, for your target user, there is a set of anchoring ideas in their intuition and a set of anchoring analytics that they would do in their rationale. What you want to be able to do is prototype a pricing and a positioning that gets at both.
You might be able to describe in great detail what other product analogies come to mind for them. What other products have they bought that is like yours?
The other thing you might take stock of when you're making this list of intuitive considerations is, what's their starting endowment? Are they coming from a place of abundance and comfort, or are they coming from a place of scarcity and struggle? The reason that matters is because people, as Kahneman and Tversky beautifully said, are much more hesitant to lose something than they are to take a similar sized risk for the upside.
Another way to say that is that the pain of losing x utility is much greater than the pleasure of gaining that same quantity of utility.
Does that make sense? This is sometimes called loss aversion and it's a pretty powerful idea. Simply by taking a little piece of paper and writing down these ideas for your target customer, you unpack huge insights for how to think about the pricing challenge with respect to your product.
On the rational side, you might actually have a simpler time doing this. You might actually be able to say look, they're dealing with a budget of x. They really value security over performance or they really value processing speed versus error rate. There's different ways to unpack their rational considerations. There might be a competitive price value calculation. You might be able to do an NPV.
The important point is, it's not so much what analytics you're prepared to do or support, but the fact that the analytics live side by side with the intuitive.
Third is value based pricing. This picture is a picture of a guy named Robert Dolan. Robert Dolan taught marketing at Harvard Business school for many years. He then went to become the Dean of the University of Michigan Business School and now he's back teaching marketing at Harvard. He's a genius, I worship him. He created I think one of the most interesting and useful concepts in marketing. He identified this idea that what matters is not so much the distance between the cost to produce something and the price to sell it, but he wanted us to focus on the difference between the price that we sell it and the perceived value to the end user.
His point is if they bought it, the difference has to be positive. The perceived value has to be greater than the price, otherwise they wouldn't buy it.
Dolan's point was, let's not build our businesses around cost plus a margin, let's build our businesses around perceived value because the gap between the perceived value and price is the incentive to buy. Let's build this up from the bottom so that it's really crystal clear.
This is Robert Dolan's genius. It starts with zero dollars and the idea that cost to produce is somewhere north of zero. He adds to that and says, we've all been taught that you don't have a viable business unless price is above cost so let's go ahead and just stipulate that that's true. For many of you who are providing software or services based businesses, variable cost is practically zero or it's very low. Remember, Dolan developed all this pre the Internet. This is old stuff. This is classic stuff.
Dolan's genius was to say there's this idea of perceived value that sits on top of price. If he were here, he'd say, "I really wish people would focus not so much on the incentive to sell, which is that gap between price and cost, but rather the incentive to buy" because that's really where the magic happens in product marketing. Does that make sense?
Now, to unpack that, let's go back to the toaster. This idea of perceived value in the toaster, and it kind of recalls some of the things we did in the very beginning when I rattled off that list of features. Dolan, if he were here would say perceived value is kind of an amalgam of a lot of different things. It's got substitutes in mind. It's got analogs. It's probably got some vanity and insecurity baked into it too, particularly for people who feel that they need this toaster as a validation of them as people. I'm one of those people. I have this toaster. In fact, I would buy a more expensive one if it were available, because that's how I feel better about me.
So you get the idea that the aspirations, the humanity, the way that rationale and analytic work together are what determine the perceived value in Dolan's model. I can guarantee you that you can draw this for your product, too.
The last one we're going to talk about is building assortments for the win. Assortments exist on three main levels. The first idea we're going to talk about is horizontal assortment, vertical assortment and optional add-ons. I would say this probably gets a little bit more controversial, and maybe it doesn't fit so much some of your businesses. But I'd be amazed if it didn't have some applicability because I see this everywhere.
Vertical assortment means good, better, best. And it means that for whatever product category you're in, you offer multiple versions of that service where the bottom version has less features and less awesomeness than the top version. Now, you might actually set this assortment, the good, better, best assortment for a lot of different reasons. Most of the time people do good and better because they're probing the demand curve. They just want to understand how many people are willing to pay a little bit more for a little bit more.
At the other extreme, people tend, not always, but people tend to pick a best to stick whether it's on their site or in their product catalog or in their showroom â€” they tend to pick a best in order to anchor the brand. We're going to talk about some examples of that in a minute, some real live examples. Then this middle piece of better still, or a decoy, it tends to, although not always, be used to increase conversion by increasing the real or perceived utility of the products on either side of it.
You get the idea that good, better, best is a pretty powerful lever, and I think it's actually indispensable, whether or not you ever sold a single unit of best, to probe and understand how your customer thinks about your product.
Now, horizontal â€” I got these out of order, sorry, I said I was going to do horizontal first. Horizontal is really much more basic than that. Horizontal is, think of it like apples, oranges and pears. They're all delicious, they all are nutritious but they're just different. And sometimes people are in the mood for an apple. Sometimes they're in the mood for an orange or a pear. People tend to do horizontal assortment to expand their addressable market. Does this make sense?
Let's talk about real live examples. Prego. There's a terrific Ted Talks by Malcolm Gladwell that you can find online. Gladwell talks about the person that invented this assortment strategy for Prego. This is an example of horizontal assortment. This is apples, oranges and pears and this is to say that traditional marinara sauce, traditional tomato sauce, is not enough. What we need is tomato, basil and garlic. We need something called marinara, which I always thought was traditional. We need roasted garlic and herb. In fact, we aren't going to stop there. We're going to keep going. We're going to offer you three types with meat, one with mushroom.
Do you know what they're doing? All they're doing is assorting their product in order to capture more of the demand curve. They do that both by disadvantaging their competitors and by advantaging their own product line. Does that make sense? You don't have to look in the supermarket for this. You can look back over 100 years to the Sears and Roebuck catalog of 1903 where there's a horizontal assortment of house dresses and wrappers. I don't know what a wrapper is but I think that it is a piece of clothing that is wrapped, and I think some of these women are wearing it. This is another example of apples, oranges and pears.
The point of going back in time and looking at these old fashioned catalogs is to show you that this stuff is deeply, deeply embedded inside the operating system of humans. This isn't a new thing that Basecamp invented when they discriminated pricing across a product here on a webpage. This is deeply, deeply embedded inside people's brains and how they make decisions.
Another example, and this could only come from one country on the planet, the whole family gets a gun. Now there's the junior gun, the middle gun, and the varsity level gun. This is another example of horizontal assortment for the purposes of selling projectile weapons to a family, who otherwise looks very happy. This is old, old stuff. This stuff is powerful old ideas.
Vertical assortment, now that we've talked about it, I hope you see it everywhere. You see it in the product line of Boeing. You see it in the product line of BMW. You see it in the product line of Basecamp.
Good, better, best, that discriminating assortment that starts with something basic and nudges that demand curve with something a little bit better that goes way up to the top and says look how awesome we are because we can build things like this; and maybe even comes back down and says here's something that will make you think that this one's an even better value.
This is like old, old tactics that can be very, very helpful to you as you build your own assortment.
Vertical assortment is not a new idea. It's an old, old idea. Teddy bears, this is from 1909. Again, sorry, it's from America but you could find examples like this I'm sure anywhere in the world. These bears are the most sensible and serviceable. They offer different sizes, different levels of stuffing. You can get glass eyes or wood eyes. This is good, better, best at work. They didn't invent it, and neither did I. It's deeply, deeply embedded inside people. Vertical assortment in talking machines. You can get the wax cylinder one or the vinyl disc version. Although I don't know if they were vinyl, they were probably resin back then.
Last point on assortments is building assortments through optional add-ons. I'll tell you a quick eBay story. I worked there for a long time, and one of my jobs was to come up with features that would help people sell more stuff because that's how we made money. I should say, because that's how people ended up happy and that's how we made money. But one of the features that we worked on was this idea of putting a thumbnail picture into every listing, and you could do that for 25 cents.
Now, I'll tell you a secret. It doesn't actually cost 25 cents for eBay to put that picture in every listing. But it did have the nice advantage of letting those merchants who wanted pictures in their listings to express that desire by customizing their feature set themselves. They could turn it on or turn it off. They could discriminate themselves what features and functions they wanted to turn on. They could discriminate how much they were willing to pay by individual product. This is what Dolan, if he were here, would call price customization.
This feature started out, we charged 25 cents for picture, probably 10% of people on the site used it. After about three years of product marketing, it was used by 70% of people and it accounted for more than 100% of eBay's operating profit from this one dumb feature. That's not to say that money is the goal because what that money funded was the creative energy of thousands of people, or millions of people really.
It's about how do you make a machine with your product that actually becomes a self-funding machine, because you can't raise outside money forever.
I've kind of distilled those four things we talked about: the substitutes and complements, the intuitive and the rational considerations, the value-based prototyping and the assortment, the horizontal and vertical assortment into this little worksheet. This comes with the slide, so you'll see it.
I promise you, I will bet you $1,000 that if you spend an hour with this worksheet just putting down the thoughts and ideas that we've covered for your product, you will end that hour significantly better thinking and better judgment on pricing than you did at the beginning. I have used this with software, hardware, consumer, small business, enterprise, robots, toys, fashion apparel. I have used this in hundreds of settings and it works. If it saves you time and pain, because dealing with pricing at eBay is what made me gray â€” if it saves you that pain, then it's worth it.
The Dualit toaster, I don't want to go into too much depth on this, but I do want to say it's worth looking at this example of substitutes and complements. I should have put this right up front actually when we were doing it because I don't think I did a great job at explaining it.
The substitutes for that toaster: Do not eat toast by a toaster at Target and hide it in your closet so that people can't see that you didn't step up to the Dualit. You can toast in the oven. Did you know that ovens can toast using the broil feature? You can delegate this decision to the cook. You can keep using the old toaster (that's the one you are supposed to hide in the cupboard). The complements: What do you imagine is in this kitchen of the target user? It probably looks a lot like that one. It probably has that award-winning concrete and steel, with that music that you guys like to play here. There's that happening for sure in this house. You get the idea though.
You can actually go through and do this in a worksheet and at the end of it you have a lot more insight into your customer than you did at the beginning.
We talked about value-based prototyping. I think the interesting thing here is if you were to sit and unpack what goes into the perceived value of this toaster, if you were to unpack the world view of the target user of this toaster, you would get a wealth of insight about how blended this intuitive and rational analytic brain is. You can imagine in the perceived value of this toaster, a lot of self-esteem issues, a lot of economic security and wellbeing issues. People who buy this toaster aren't evil, they're just people. You can go unpack those issues in your target customer as well.
Building assortment. They are masters of this at Dualit. They have the original toaster, they have three other types of toasters and you can get all the way up to a six-slicer. They know what they're doing these people with respect to toasters.
So really, those are the four ideas I wanted to cover and I wanted to also just say that if you have specific questions or ideas or push back on this stuff, I'd love to hear it. The only way it gets better is if you push at it and poke at it. Anyway, who has a thought?