Why credit-based pricing can break trust (and how to fix it)

Software as a Service, for ages, you pay when you don't use it. But something about credits, like, it's maybe more obvious that exactly what you didn't use and what you left on the table. You need to make it really clear what the value of your credit is. What do you get from it? What is a successful use of a credit? And then also, what is the cost of an action? Will my credits roll over? Are there overages? What happens there? These are all things that have to be really, really clear. I'm one of the co-founders of Schematic. I think the reason we started this company is we found it really hard to evolve pricing in our previous roles. So before this, I led product at a company that I was kind of at from day zero to Series B. I saw a lot of growth. It was logistics payments. And when we wanted to do an add-on, it was months of engineering work. When we wanted to try usage-based, it was months of engineering work. And so, that's why we started Schematic, and that's kind of why we're obsessed with pricing. And today I want to talk really about credit burndown. Before I start, who is using credit burndown or is thinking about credit burndown? Anyone here? Yeah. Why is that? Are you using it or thinking about it? Thinking about it. Mostly because we have, like, a flat fee right now, and —abuse is the reason why. Okay, why not just usage-based if that's the only reason? I think if you do usage-based, people get scared, like psychologically. Like, oh, I'm always worried about my tier or whatever, versus credits, it's like you have to be full. Got it, control. Okay, so I'm gonna jump in. I want you guys to interrupt me with questions, but I'm not gonna spend a lot of time on this. I'm actually quoting Kyle's work here, but As you heard, like seat-based pricing is really dropping. You're seeing hybrid pricing surging. I'll show you some examples of that. You're also seeing gross margins going lower for AI companies versus traditional SaaS. You've got reporting that AI companies reporting that infrastructure costs are going up for them and also customers of AI saying that they're having trouble predicting their costs. I wanna talk a little bit more about, a little more like tactically the how, like what does good credit burndown look like? With AI, the cost is driven differently than SaaS. Like we talked about, you have real infrastructure costs. Usage-based models promote fairness, and then hybrid and credit models really allow you to have more predictable revenue and still have flexibility. Why credit burndown? You have this credibility of cash flow, and that's on both sides. Like you can, as a company selling your product, you can know that upfront you're selling a bunch of credits. You know that that cash flow is something you can count on. As the customer, similarly, you get that upfront and you burn it down. That prepaid element is really differentiating about credits, both at the high, like a downmarket customer where they don't wanna maybe commit to subscription, and even at the enterprise level where maybe they can buy a bunch of credits for a year. You also abstract some complexity. So you mentioned an example of if you have a complex product where maybe you have different actions that burn tokens at different rates,— it's kind of easier to put that behind your own currency, and you can say this action costs this much, and you don't have to explain all the math behind that or all the complexity behind that. Similar with variable consumption rates. Maybe you have parts of your product that are cheap for you to support internally and others that are expensive. You give the user a lot of control. They can top up their credits. They can do overages, or maybe they can just have a plan that has credits included. And it's really adaptable to hybrid. Like, that's maybe undersold, as you can have credits alongside a seat-based model, alongside an even a usage-based model that's not a credit. So it really plugs in well with other pricing models. If we look at ones that are done well, one of my favorites is Clay. There's a couple reasons they do this well, in my opinion. First, they're very good at making the definition of a credit really clear. They have a very good part of their website where they explain what a Clay credit is. It is their currency. They have this experience where you can tell, you can say like, how many companies are you targeting? How many contacts are you sending to? How do you want to use the data we bring you? And you basically just say like how you're going to use a product and it tries to tell you how many credits you'll need. So that you have that upfront definition and that upfront expectation setting, which is really powerful and kind of rare really. I think most companies do that more as you go. You just buy some credits, you kind of trust them, and you see how much you use over time. They do a really good job. Inside the product, they're also very good at showing you how much you've used, how much you have available, how the rollover is gonna work. So they have very thoughtful rollover policies, and they make that very obvious. And then finally, they have this, which you would think all companies that have credits do this, but actually very few in my experience too, which is they show you all of the actions you're taking in their product and how that actually has burned down credits from your balance. I don't know if any of you use like a Curse or a Windsurf, like that's not easy to find slash it's not there. And I think this is where you can really create a lot of trust. You can pull this data out, figure out how your usage is impacting your balance. So Clay, really good example of credits done well. Another one is v0. So they don't do everything well here. I don't know if Shahar is like listening yet that we have a guy coming. He's gonna speak from Vercel, but they do a bunch of things I like and some things I don't. They peg it to the dollar value. They haven't elected to have their own currency with its own cost basis that you then have to understand. You just say you buy credits at the dollar value. I think they do that well. That's pretty useful for more consumer-based products. I think they have a good auto top-up or recharge setting. They have different types of credits like what you get with your plan versus promotional versus à la carte. One thing I don't think they do well, which you see with a lot of the coding tools, is you don't really know how many credits you're going to burn any time you put a prompt in. Maybe they'll tell you like a certain model costs more, but you might do something and it might create a bunch of errors and then I have to fix those errors and you'll just burn a lot of credits and you'll be kind of bummed that you did that and that's a difficult problem to solve and so there's a little bit of transparency stuff I think they can do better. Maybe before I keep going, when you guys use credit-based products, what have you noticed? Like, are there ones that stuck out as being really good and others that have been really not so good? Yeah, you did.

I think I actually canceled subscriptions. Okay. Because there are certain months where I'm using it, or I'm not using it, and I think that's a bit of a challenge on the value dynamic. And I think you sort of marked this as common, saying like, you have an annual deal, you should get to use all their credits, not have them go away if you don't use it a month. And that's kind of something I struggle with, but even like MinSurf or Cursor or whatever, they have their thresholds, but I need to get to— Yeah. Keep them, but I'm paying for it. So it's a question of like, are you in kind of the gaming dynamic where you're making a lot of money from people who are, or gross profit from people who are not actually paying or not using your product and delivering value? Is that at risk in the future? And how you answer those questions in this sort of new world of like mapping like value for customers.

The comment there was, he's used tools where he didn't use all his credits, and that ended up making him cancel because maybe not getting all that value was just super clear. Like, software as a service for ages, you pay when you don't use it. But something about credits, like, it's maybe more obvious that exactly what you didn't use and what you left on the table. I have seen companies do rollovers really well. Some, like, Clay actually, it rolls over, but you have to use that rollover in that month. You have other companies that do like a partial rollover. That can be a good way to handle that feeling of like, I didn't use it. Do I get to keep it?

Yeah. Something else I've seen as a pitfall at Klei, actually, where I've done some advisory work over the last 6 months, is in the onboarding, customers blow through their credit allocation in like minutes. Yeah. 'cause they're like, I want to enrich this list of 20,000 emails, go. And then they just destroy all their credits and they're like, wait, what happened? They're still like figuring out the product. And so something about giving the customer a seatbelt when you give them credits in the onboarding, especially if it's a free trial, which Clay has, it's a usage-based free trial. So when you're done with your credits, you have to pay. And then there's also this feeling of in the product, It's like you're at a casino and you've been given like a finite amount of money and everything costs money and it's super not fun. You're not getting any money back, right? You're at a casino. It's like this, but there's like little coins everywhere. Like it's not coins you get, it's coins you spend. Yeah, yeah, yeah. So like these two problems and I'll candidly say like these are not yet solved. We identified them as issues that like lots of first-time customers experience. We focus on other things to improve, like activation auditing, but there's still kind of like problems out there in the wild. Yeah, a lot of it's about framing, messaging, you know, safety, safety mechanisms for the customer. But those are two things that like still to this day are issues that the team is thinking about.

Those are great stories. I mean, it leads me into some of these pitfalls, which is— and this has been covered a bit already, but you need to make it really clear what the value of your credit is. What do you get from it? What is a successful use of a credit? And then also, what is the cost of an action? I showed you like, actually Clay does a pretty good job in context showing you what will cost, what a certain action will cost. Not all companies do that. And then we talked about limits. Like, will my credits roll over? Are there overages? What happens there? These are all things that have to be really, really clear. And one maybe a little bit more of a boring topic but worth mentioning quickly is revenue recognition is a little trickier with credits. There's something called performance obligations, which means you can't actually recognize revenue on unused credits, at least not right away, at least not until the performance obligation is fulfilled, which means you get offered some credits. Once those credits expire, if they're not unused, you can recognize that revenue. I'm not going to go deep into this unless you guys want to, but there are— this is something you have to do well.

Is this a US-based thing or everywhere?

I know about the US accounting kind of code on this. I assume there's everywhere some aspect of this. Even included credits have to—

do we need to worry about performance? I mean, overages make sense, but—

That's a good question. The question was, do you have to recognize performance obligation on included credits? You do. It's a bit easier though, because when the plan period elapses, it's expired. So, you can recognize that revenue. But anything that you pay upfront, that makes sense.

That journey recognition cannot hit on into performances.

And that's why you want to make sure those standalone bundles that you offer do have an expiration date. Otherwise, you may never be able to recognize that. There's even more detail around here. If you offer included and standalone credits, you have to separate the revenue on that included credits from what you get in that plan. Again, we can get into that if you guys care about the accounting stuff. I've had to learn it because of what we're building. But it's something to keep in mind. I wanted to show you really quickly— Let's put refunds in there as well. Oh, refunds. That's right.— that's an aspect of it. And also promotional credits. These all have strong accounting repercussions. So let me bring this over. You can set up credits with Stripe. You can set up credits with Metronome, with Orb. We have a way to do that relatively easily. So this is a little test app where— called Price Spy that I built that just allows you to monitor your competitor's pricing page. And you, it takes a snapshot of their pricing page and it uses an AI to analyze it, figure out what is their pricing, how good is their UX even, all of that. And then it does that on a cadence. And I've set up credits with Schematic so that I have this, this is like my customer dashboard that shows me, you know, what plan I'm on, what credits are included in my plan. So every time it goes out and does a snapshot, that costs 1 SpyBuck. You can see here, like, what my credit grants. I have some that were included in my plan, and I was given promotional credits. I can buy more credits relatively just by clicking this button. I can change my plan. I can unsubscribe. So that's us trying to make it a really clear experience to the end user about, you know, what you've bought. How to top it up. I don't know if anyone's an Arc user, but Arc is, like, really starting to die on me. But this is the app itself. Again, you get these snapshots. And one thing I want to improve here is, like, this snapshot costed X bucks. And I can do that with the schematic platform. And inside here, like, when you're actually managing this, like, the way it works a little bit more. You have an event that comes in, that gets enriched. It's associated with a company, with a feature. You can define here what it costs. So, I've got the snapshot feature. I've got it entitled to different plans. I've got it set up so it consumes one credit. I can also go in and manage companies, directly. So I can go here, find that company. I can also decide to see their ledger. So I see their ledger of how many credits they have. It scrolls for me. I can expire grants if I need to. I can, I can do promotional credits from here. I can do exceptions from here., and then also I can manage the actual customer portal part with our drop-in component. And that's just a quick example of how that actually can be implemented. Again, you can do this with other tools, too. I think what you want to end up thinking about with credits is what is your value metric. But you don't want to just be a cost-plus with this. I mean, you can. I think that's one way people are doing it is, like, think about What is your internal cost and make sure you're not losing money on it. But if you can also create a value story around it, like what is the outcome of that credit? What is the job that it's actually allowing for? Especially with AI, maybe it's replacing human labor. You can build that into the credit and that becomes a lot stronger than just cost plus. Another kind of tactical thing is start simple, track everything, make sure you're sending in events for everything, make sure— you're understanding what your inference cost is, even if you're starting with just one credit type. Design for transparency, that means, like, have that portal that shows the credit balance, that shows what your usage, how that usage ends up impacting that credit balance. Make sure you're set up to iterate and experiment. If you're spending a lot of time, like, on billing code, that can get expensive. So, do the upfront work to make sure you can have that agility. And finally, expect that if you're growing, there's going to be complexity. You won't be able to do just the simple pricing. You'll want to. Your customers want simple pricing. But as you launch more things, you're going to need to monetize those things. And you're going to need those things to work well together. Any more questions on credit burndown or generally on how to set up your pricing? Yes.

Maybe it would pick up in the latter sessions. But in this slide, at the coffee where you're deciding, the number of credits to get to be included in free trial or whatever. How do you decide what kind of like pricing studies or interviews or like research would you do?

I think the way I think about it is what do you want them to do in a trial that gets them to a value moment, like an aha moment, and make sure that the credits you provide allow them to do that enough times where they get hooked or see the value. So work backwards from what do you want them to accomplish, and then how many credits does it cost to get to that, and then do you want them to do it once or twice or three times before they feel like they need to pay you. Thank you.