AI Credits Roundtable

All right, awesome. We've already got some folks in here. Hey everybody. Welcome to the Pricing SAS round table on AI credit models. I'm gonna give it a minute and then figure we can kick things off with you guys and and let y'all introduce yourselves and get into the meat of the of the conversation. This is my first time doing a a LinkedIn live broadcast. So I'm I'm curious how that goes. But Looks like it's working out okay.

Awesome. All right. We've got some more people joining. we've now got thirteen people. Welcome everybody. It's great to have you here. all right, guys. I think we have a lot of questions that we can go over and I feel like there's gonna be a lot of conversation. So why don't we kick things off? before we actually get into the questions, I think it'd be really helpful for you guys to just introduce yourselves and kind of tell people where you're working now, where you've been. And just kind of like set the stage and set context. So Rita, do you want to kick it off?

Sure. thanks, Rob. Hey everybody. I'm Rita Sherman and I lead pricing and packaging at Gong. Prior to that I was at Brays, and before that, a whole slew of different roles within consulting and retail, et cetera. been here for a couple of years. Most recently, our our biggest shift has been a couple of things. last year we took a a real move into modularizing, a somewhat

wouldn't say, you know, an early set packaging structure. So it needed a lot of sophistication that we were building into it and and just general maturity of the packaging structure. And then this past year we rolled out our hybrid model and so we introduced Gong credits, which has been, you know, a voyage of its own. And so happy to to share some of my learnings and even things that we're still working through.

I cannot wait to hear about the voyage. awesome. Jeff, do you wanna jump in next and then we'll get to you, Mano?

Totally. Cool. Jeff Curran, head of pricing and packaging at Miro. I've been there for about a year. and have we're we're actually currently thinking through what our credit should model should look like and kind of the agentic AI world. so we're working through that right now. So very timely there. prior to that, I was at Zendesk for five years, you know, launched Zendesk Suite, launched outcome-based pricing as business lead there. And so kind of been through several usage based pricing launches and adaptions the past couple years.

Love it. Mono, do you want to close us out? Awesome.

Sure. hey everyone. I'm Manu Mehra, currently at Databricks. I am part of the strategic deal pricing and right now I lead the Mare Industries patch. prior to that spent five years at Google Cloud, very similar role was leading deal pricing for Silicon Valley and Western States. I would say in the last eight to ten years have really seen the space evolve. And like work with some of the most largest and the most complex customers we can think of. So, yeah, super, super excited to be here and just be part of this amazing conversation.

Awesome. I am I'm super excited to have you three. I've like met all three of you guys personally and like have heard you guys talk about pricing more broadly and credits specifically. And I feel like each of you is kind of like working on different things, which I think makes this really interesting and there's just gonna be a ton for people to learn and take away from this. so the w the way that this will work for everybody is w we have kind of like some questions preloaded that we're gonna just start going through and I'm gonna kinda kick it off and then let these guys kind of pass it around. And we'll work through it. So I think to start, I think the question that kind of kicks this whole thing off is like, what were you charging before credits and kind of like what broke that made you switch? I think everybody kind of like looks at AI as kind of like the primary reason for that. But I'm curious, like in each of your cases, was it kind of like a proactive move or more of like a reactive move towards credits and kind of like what you were charging before? And we can go in the same order that we started in there. So Rita, I'll let you kick that off and then we can go over to Jeff.

Yeah, for sure. so prior to the rollout or the shift to our hybrid model, we were very traditional platform seat based, still are, candidly. so a year ago we really started to modularize and introduce new packages because we fell prey to the whole, you know, overloaded core seat, where we were starting to break the integrity of our pricing because like essentially just relying on heavily discounts when customers were like, I don't need this and I don't need that. So we started to split things apart. that was a long time coming and I joined right around the tail end of that and the rollout of it. And then actually the shift to this credits model was largely driven or led by our CEO, which is the first time that's happened to me. where I know, right? So we we essentially we we had this meeting

probably two years ago now or a year and a half ago, where we walked through, okay, these are the changes that we're making, these are the the opportunities that we see. And our CEO was like, Okay, well what about, you know, all these different monetization conversations that are happening around usage based, consumption based, outcome based. Tell me more about that. What's happening in the industry? I want to be on top of it. I want to be ahead of the curve. Where's the opportunity, etc. So that was one of the large drivers of it. And then it was a real journey, I will say, for us to evaluate like where does this make sense for our business? Obviously costs are a large driver. But it took some time between my team, the process, Product team, our finance team. I mean, everybody knows that pricing is like a team sport, right? So just a lot of conversations around like where does this make sense for our business and how do we roll this out in a way that we're essentially not going to shoot ourselves in the foot or upset a bunch of customers and do it really cautiously because the industry is changing so heavily. So that's what drove like that that shift. and then I'm sure we'll get into more about that as the questions go on. Yeah.

yeah. No, that's perfect. And and love that love that it was kind of like CEO involvement. That's awesome. Jeff, how about you?

Yeah, I'll I'll actually talk about could maybe both so I I was at Zendesk a couple years ago. like I I remember the day that Chat GPT three point five launched, like over Thanksgiving weekend. and like and it's because it was like such a existential change and threat to customer support. And so like I think like that in terms of like what broke to a sleep-based model is like kind of significant risk to A seat-based model with LLM launch. And so that's like where Zendesk started the journey towards usage. certainly started kind of like with outcome-based, it it allowed to sit kind of alongside seat-based really, really cleanly and really nicely. so that's what we did there. At Miro, you know, also a seat-based model. I think like in terms of a little bit more proactive, I think, on on launching, like wanted to build. Capabilities that didn't scale with humans. And so knew that we needed to introduce kind of a more usage-based model. and in terms of like what we're seeing now, like in terms, I think the it's a really interesting question of kind of like what broke to force you to do that, like to kind of like cause this. And like, you know, we're seeing a lot of changes right now with just like the agentic nature that's really moving us from kind of like phase one of. AI credits to phase two AI credits. And so you know that's I think another big shift that's just happening right now. that's really causing a lot of a lot of updates and models really, really quickly.

for sure. Yeah, I I'm definitely seeing that in the data as well. just like credit models evolving and kind of getting a lot more complex. And Manu, would love to hear your perspective at Databricks. I I I believe they've had credits for a long time. So like much longer than I think like a lot of traditional SAS players. So yeah, I'd love to hear your perspective on that.

Yeah, like predominantly I come from like I would say in the last de good part of last decade I've spent like working with cloud companies. So predominantly cloud tends to be more usage based. So we don't do like that much seed based in the first place to start with. But again, like with credits that Even the usage base is kind of evolving. And and the big reason is like at least in my journey, Robert, I've seen is there is a if you just go like seed-based or even like usage based in that case, there's a big disconnect between like what you're charging the customer, what value you are driving for the customer, and what it costs you internally. So I think that's where credits really come handy because the simplest way I think of credits is it's it's more like a flexible currency, right? Which could be used towards anything. So it gives a lot of predictability. So I think that's a big change that I'm seeing. And Potentially where I see this space further evolving in the next I would say three to five years is like it won't be completely credits-based as well. I see this evolving into a hybrid architecture where there is some portion of outcome-based pricing because where we are standardizing our outcomes more and more. And at the same time, there is also these consumption-based credits, right? Like where you are basically driving consumption and offering credits to the customer. So I think that's where I would see this space evolve. and then yeah. The seat waste pricing i is is great, like I would say, back in the day, but like as things become more autonomous as we try to like build more workflows and autonomous agents, I think credits give that flexibility because end of the day, the same task.

could like use very different tokens, just the complexity of the task, right? So I think that's where credit gives that flexibility to the customers. So I think that's the big change that I'm seeing, Rob, in my space where customers need that predictability and end of the day, like seed-based or or if you talk about traditional software, it was getting access to the software versus what's changing is now software doing the work for you with AI. So like how do you how do you monetize that and how do you give predictability to the customers? And I think that's where credits is is the path forward.

Yeah. Well we can I I love your writing on that by the way, Mani. You've done a lot of writing around like agent P and L and kinda like how to think about the relationship between value and cost. And that's kind of where I wanted to go next is like I I think some companies like on the infrastructure side, like it actually makes sense for them to charge for tokens. And then you get like to the exact opposite of that, kinda like where you were before Miro, Jeff, Zendesk, where like you can actually charge for outcomes and like you can agree on what an outcome is with your customers and actually make that happen. I think like the vast majority of companies live somewhere between those two things where like they've got these underlying tokens and then they have these different jobs that they solve or kind of like tasks that they help with or outputs that they can help create. And the challenge is starting to figure out like what are those worth and like how do you ladder up the the the price there? so yeah, I'd be curious like Rita, we can start with you. What what do credits mean in terms of value in your product? And like I I'd love any of your thoughts just kind of like on that generally, like How you guys thought about this and and designing your credit model, and like deciding which features get credits, which ones don't consume credits, anything like in that vein, I think would be really helpful.

Yeah. so when we started thinking through this, I mean we knew that there was a a number of decisions that we had to make and they all felt very overwhelming at the time. Some of which were, okay, you know, looking at some of these agenda capabilities that were rolling out that are on our roadmap that exist today, like what makes sense. And we actually there was a little bit of trial and error. So we had just released as an example our new like role play trainer or AI roleplay trainer. And so we thought, you does this make sense? Like, should this be on a consumption model? And it was a little bit of looking at the use cases, talking to beta customers, understanding, like, does this make sense? Is this the right place? And we found or we thought, you know, no, this doesn't really make sense, and it doesn't match what we expect to see in terms of like perceived value and where you might see a customer willing to spend, right? Because Moving anything to a usage base or consumption layer is like a big flashlight on any flaws that you might have or a magnifying glass, right? Like it's it's now you're now banking on the fact that the product is good enough that people are going to want to see it, see the value in it, and continue using it. And so you really need to have some confidence there when you're shifting anything. And then the other thing we didn't want to do.

A couple things. One is we didn't want to have a laundry list of different agentic capabilities or features that had a different price point, like or different credit, you know, essentially become a restaurant menu. We were trying to avoid that as a whole and make it digestible for customers. And then the third piece is we were building out what we just released is gong enriched, so essentially enrichment tool. And we wanted to find a way in which we can use one currency for both AI capabilities and data enrichment.

And so a lot of the things that we were building at the time for something that was months away from release needed to speak to both of those. So it was a huge balancing act. As far as the agenda capabilities, where we landed was let's focus on our autonomous agents, right? As we release more, we're working towards it. Let's let that be the key driver. Anything that's autonomous, anything that essentially is heavy AI processing. And, you know, for those that are not familiar with Gong, like our our bread and butter or what a lot many people know us as is this call retorting recording conversation, you know, that's that's sort of The the baseline of what we do. We've light years evolved from that. but let's use that as sort of the the the measurement as we start out and we can always enhance it. And so thinking of things like a credit is a call when you're using an autonomous agent to build something against your calls or to analyze things in your calls. And so starting there, which is really familiar to our buyer, having the same credit breakdown across different features, they all come back to that like same calculation, made it very digestible and we knew like let's start here and evolve as we release new features that are gonna need more complexity or non deterministic or any you know any direction that we go. How do we grow from this rather than like shift it entirely because we're also trying to be mindful of, you know, falling prey to the quarterly pricing changes.

Customers are just not there yet and they're not that adaptable. I think they'll get there, but they're still living in a world where they don't wanna see a single pricing change until their renewal. So it's it's been a really interesting experience in that.

You mentioned not wanting to have kind of like a restaurant menu for your credit table, which I totally respect. And like I've I've seen some of those and like I've seen some companies start there and then kind of like punt on that because it's just like you literally have to update it like every week and it just seems like it becomes its own kind of monster. did you guys end up choosing like a different way to design it? Like do you have like ranges or do you just kind of like have like not even? No.

No, it's like a single so you know, we've now evolved into this like read work output framework that we're rolling out as we're as as we're moving into more of like agent building custom agents, which we've also recently announced. but it started with just this read component where it was, you know, depending on the type of conversation history you have, whether it be a call and it could be like a long call or a short call or email, like really where there's a perceived

difference in value and willingness to pay, it equals a credit. And that doesn't necessarily change. So there's consistency throughout whether you're using it for one feature or another. Of course we have like discount flexibility in terms of credits. We sell them in predefined packs. So we we just try to keep it clean. And I think one of the biggest takeaways for me is I was I was talking to a beta customer for one of our products and she's like, I manage so many different budgets and so many different contracts. Please just keep it simple for me. Like I know that this is the direction that everybody's moving, but I don't want to have a PhD just to understand my bill. So I thought that was really telling and I can appreciate being in that position and not wanting to have to like read through pages of documents to understand how pricing works.

Yeah, that's awesome. I'm I'm really excited to talk to you in a bit about like sales and like how your sales team deals with 'cause like I follow a bunch of gong salespeople on LinkedIn. I think if you're on LinkedIn and in SAS, you'd probably have come across some gong salespeople. so we can get to that in a bit. But Jeff, I'd love to hear your perspective, whether it's from Zendesk or from Miro, on kind of like getting to the value component of of the credit model.

Yeah. Yeah. I mean so like Miro, Miro we we price on artifacts, kind of like what AI helps you create. And so like on on a Miro board, it could be a document, it could be slides, it could be sticky notes, whatever. And so I think like it it aligns really well to you know tangible things that a human would have done that now is being done for you. Makes a ton of sense. And so that's kind of like where we sit now. Like the reason why we use credits is because you can do lots of different things. And so instead of pricing each of those, where like you have to pay for a slides like budget and a document budget, like you have a credit budget and you can use it however you want. And so I think obviously helpful there. you know, in terms of like kind of like where where we're looking, this actually sounds like it's kind of closer to where where Rita's going as well, is like You know, from a the thing that's that's really challenging with agentic is like you don't know how many steps it'll take before you ask, you don't know how big a board will be when you ask it to create it. And so like there's just a lot more uncertainty. And so we're building that in with a a similar framework and kind of like we think of it instead of forget exactly read, work, write, I think you said, like ours is like contact, like You have a bunch of context that you use to create the document. Like then there's thinking aspect, and then there's an output aspect. And so we have those kind of similar same three areas. And like and we would price according kind of like accordingly on each of those. but it's a lot less like prescriptive prior to the creation because like you don't know what's gonna get built necessarily until after the fact. And so We're moving to more of a like a work done concept instead of like pricing the output of what's actually created, essentially. Yep.

I like that. I you you said something that like really resonates. It's like you don't know what products are coming down the pipe, so it's really hard to create a credit model that's going to be like durable forever. Like it kind of has to evolve. Like that's just the nature of it. and I have more on that in the next section, just around balancing kind of like AI costs and margins and all that stuff, because all of that's obviously in flux as well. Manu, how about you? How about you in and Databricks or your experience at Google Cloud. like I'd be curious how you think about the value I've you if for those that are unfamiliar, Manu wrote an awesome article on like an agent P and L framework and kinda how to think about agent economics, which I highly encourage I'll share it with everybody. but yeah, I would love to hear kind of how you guys are doing this at Databricks.

Yeah, I would like in my experience, Rob, I think like when we talk about designing credits, and as the as the space is like so rapidly evolving, it becomes less of a function of okay, are we giving credits for GPU time or is it for tokens or is it for like some infrastructure consumption? It's I think it's more of a function of what value or what outcomes are we driving for the customer. And and the way I look at it in my space as like think of it like a three-layered process. One is the value creation layer, which is basically, okay, there is a customer problem that I'm trying to solve. And here is a pool of credits that helps you with that. So that's the that's the topmost layer in my mind, which is the value creation layer. The second layer I would say is the consumption layer. Okay. What actionable things that is the customer doing towards which these credits will get consumed. So that's your your consumption layer. And last but not least is your economic layer, which is like what is my true cost to do to serve these inference or orchestration or or infrastructure. So I think as long as the the design of the credits kind of revolves around these three layers, which is the value layer, consumption layer, and the economic layer, I think that's a successful designing of credits. And then as I I and I think I really Echo the thought that Rita brought about. Like, end of the day, it needs to be really simplified for the customers and for sales. They don't need a PhD to go and explain this bill to the customer, right? Like, so I think the more simpler it becomes, like customer-facing, of course, there would be complexities internally where we figure out what the true cost to do this serving because these in these tasks could vary a lot from one customer to another. So I think as long as we keep it simple on the customer side, but make sure we are following this three layer three layered structure internally, I think that sets you up for success in the long term.

I love that. I love that structure. yeah, that's a really, really helpful frame. And I think it dovetails really nicely into kind of like my next question for you guys. So AI costs are changing rapidly. both like the prices of like frontier lab models, they're launch they're launching new models all the time. Open weight models are launching. There's this idea of kind of like model routers that a lot of companies are using. Like I think Notion just like boosted their model router and they're really it seems like really kind of trying to drive people towards that. you know, open router just got bought by Stripe, so I think they're pretty bullish on on that functionality. And then you have like other inference clouds and companies that are allowing you to kind of like post train your own model, right? So like Harvey did that. I'm pretty sure Fin AI did that for customer success. so they're kind of like all these different ways I feel like that you can engage with AI models under the hood of your product. so I'm curious, like, you don't have to go into the weeds of like what exactly you guys are doing, but like I'd be c I'd be curious to hear kind of like how you guys are thinking about AI costs and how that ladders up to what you're actually charging. I think like a general thought that I've I've had like kind of like a Just like a general framing is like if costs for tokens are generally going down, then inevitably like somebody's gonna capture that margin over the course of a contract, right? And it's like, is the vendor like going to give more value away to the customer or is the vendor gonna capture that margin? And I'm just I'm I'm I'm just generally curious like how how people are thinking about this stuff. So yeah, we can we can start with you again, Rita, and I'd I'd love to hear your thoughts on on the cost component.

Yeah, I mean it's a conversation that we have too. And you know, I was at the the monetize conference recently, and I think it was somebody from Replit who was saying that they like pass pass through any of the savings, right? They're really transparent. I'm not sure we're it's almost like a risk tolerance there, candidly. and I don't I don't I think we're a little bit more averse. And so the way that we've designed it to protect ourselves from the sort of like these eroding margins is we price against an exchange rate we can adjust.

versus like fixed feature pricing, right? So as I mentioned, it's a credit per call. And we've even like w we started, I'd say, on the more conservative side. Like we had some data and usage to to run it up against and and just making some hypotheses around like how usage was gonna drive credit consumption and credit costs and then balancing it out as a credit model because obviously it varies so much from customer to customer. and so we started a little bit more conservatively with some room to change things for the better of a customer. So instead of charging something for a half credit, now it's point three credits, right? in terms of volume and and we will communicate that. But then I think there's still this level of like, okay, let's let's be mindful depending on And you know, what we know contracts that we've signed in terms of like our own infrastructure costs, how we balance like this this whole AI credit versus a data credit with our enrichment product, right? Those are different margins. So how we balance the two of those and and where they can sort of be a safety net for one another has been really helpful. And then I think by leaving it more on the conservative side and try to capture some of that margin, but allowing room for discounts of credits or issuance of credits for you know concessions, escalations, et cetera. Like we've kept ourselves in a good place thus far. but it's, I mean, we're looking at these numbers daily and just making sure that there's not bleeding happening that we need to somehow avoid. I will also say where we are being a little bit more aggressive is Knowing that like if we're going to, you know, issue promo credits or something like that for a new feature, we're going to do it in a way that we want our customers using it. because for us, adoption is number one, right? If they're sticky products, we all know as AI users that like once you start, you can't stop, right? It's it's really hard to pull back. so we would rather be a little bit more generous upfront to start to make sure our customers are seeing that value and using it, and then you know.

recoup if if need be. But we haven't gotten to the point that we're in a bad place.

Awesome. Is the so I'm curious too, like from an org structure, like on the cost, like is that your product team that's like managing AI costs? Or like do you guys have like a billing engineering team or it like I'm I'm just curious in general like how that structure works. 'Cause it bec it like it Yeah, I was gonna say it yeah. Sorry, go ahead.

Yeah. We're very very cross functional. We I mean like we we have constant conversations between our our product and product analytics side and our strategic finance and myself. And so it's like a a very

it's it's very balanced, right? We're not solely looking at product for like cost analysis, like we'll get the data from them. And then our strategic finance or data analytics team will do a like a whole bunch of analysis and and just different runs to understand like what those implications are gonna be to the business.

That's awesome. Yeah, it it I asked because it feels like those costs are becoming super important to pricing more broadly and it it does seem like something that's inherently going to live in the engineering

And so I feel like part of figuring out pricing is like connecting like building a bridge to connect those two like very closely. so very cool that you guys are doing that and and already have that that structure in place. Jeff, how about you? How are you guys thinking about like AI costs at Miro?

Yeah, I mean I think so I mean we like we as much as possible we try to like separate the cost component from like the value component. And so like when we think about obvious thing for a price person say, but like when when we think about pricing, like we we are primarily talking about the value the incremental value that Miro is adding. And so, like, we will certainly pass through the cost plus the incremental value. But like what it ends up doing is like, I guess, a couple of things to kind of manage it. Like, number one, defaults are super powerful with this sort of thing. And so if you have a default model, like vast majority of people will have the cost structure associated with that model. So you can design your margin comfort around it pretty clearly. And then, like, but by allowing kind of like these two cost and value. like things to move separately. Like you can hold value and if a cost decreases, like you can pass that cost decrease on to the customer and it will lower their cost, it'll lower revenue, but it'll increase margin for for Miro essentially. And so like it's it aligns the margin for a company and the cost for a customer pretty well. And so like there's flexibility or like fluctuations in revenue, perhaps. but again, like defaults are are really, really strong with this sort of thing. And so you can design it around that pretty confidently, more or less.

That's awesome. Manu, how about you? you're muted, I think.

sorry. Yeah, I think in my experience, I think since like the AI costs are so dynamic in nature because there are model improvements that are coming out like by the week, if not by the day, and there is like so many co so much incoming from the competition. I think when I think of managing AI costs and profitability, I think it becomes less of doing it at the token or the model level. It becomes more at the workflow level. Because again, it goes back to the point I made on the previous discussion was like

How do we like build these outcomes or these like the value for the for the customers? And I think that's where managing it at the workflow level goes a long way. Now it's a hard problem to solve because the workflows could look very different from one customer to another, right? So it's very hard to standardize your cost based on that. so I think like the the the easiest way to do this is like if we can like create a certain set of workflows and customer segments and establish contribution margins based on those customer segments and workflows, I think that kind of gives you a more realistic profitability view versus like doing it by tokens or by models. and again, as I said, like this is a hard problem to solve. So I I think this is a little more theoretical in in my view, but as long as we get more data and more customers start using our products, I think that's where the industry could evolve. And end of the day There isn't like one size fits all, right? Because with credits, we still want to give them flexibility for the underlying technology. But at the same time, we don't want to give them like a big rate card of credits as well. We don't give wanna give them like 20 different kinds of credits or 30 different kinds of credits because then we go back to the product pricing complexities, right? So I think just solving that predictability at the workflow level could could help us, is is how I would see this. this space evolving in the in the next years to come.

And so the the that that makes a ton of sense. So like the the idea there would be that like for different types of workflows you kind of have like a target margin that you're that you're looking to hit and like understanding that like there's gonna be a range. Got it. Okay.

Exactly. And again like Those workflows, now I think the there's one complexity there, Rob, is like how many workflows do you define? Because if you're serving like thousands of customers, like you can't have thousands of workflows. And so I think standardizing that is the tricky part. Maybe like there has to be a set number which could evolve. And I also echoed the thought that that Jeff mentioned that if you improve cost, you basically pass it on to the customer. and of course you you get improved margins, but if your unit economics are improving, more likely than not,

the volume consumption will also go up from the customer side. So it'll kind of, even though your costs are going up, the volume will kind of solve for it and you'll still see the revenue grow in the long term. So I think that transparency and that flexibility and then that passing on to the customers, I think builds a long-term relationship. So I totally echo that thought as well.

Love that. yeah, to to your point, like defining every workflow, I think that can that can be hard. And I I I think like especially for some of these more like broader horizontal SaaS companies where it's it's almost like they're they're better off charging for tokens than trying to define too many workflows. awesome. Okay, so the next next section that I wanted to talk about here is kind of like credit operations and management. I feel like one

big area like people talk about credit models and like kind of how to design your credit model, but then I think usually the next question that I see is like how do we actually handle rollovers, top-ups, and like flexibility around our credits. And so yeah, I'm curious, like a few layers here, but like how did you guys decide how many credits to include in each plan? whether or not to do rollovers, top-ups, and kind of like the rationale behind behind those decisions. So Rito, we can kick it back to you on

Yeah, so I don't know where everybody's at in terms of their AI, you know, whether it's some sort of new monetization journey, but we had a ton of these questions as we were designing this. And so we did a three-day three-day offsite. Cross-functionally. So we had folks from finance, from accounting, from product, from RevOps, from product marketing, from enablement, and we all just got in a room and like had different owners for each of these areas think through recommendations and then would present to the group as to like how our approach is. So that's how we kicked off a lot of these discussions in in true. candor. I tried to do this on my own and it was it's just it's so much guys. And so like you you definitely need to lean on a team here. so how we came to a lot of these decisions, particularly around like what's the right default number of credits, you know, whether we do rollover. So a lot of those were real like cross-functional decisions that had kind of like pros and cons of, you know, if we do it this way, what complexities does it introduce? How do customers react to it versus if we do it that way, can we even do it one way or the other? So for example, default credits was we knew we were going to do default credits. Like that was just an industry norm and we wanted to go forward with that, Jeff. So I'm like aligned with you. It's really powerful. and finding that right number was just a lot of different like modeling in terms of What products do we think are going to be on credits? What are the costs of those? Like what are the margin implications for each of those? I think we were looking at like four or five at the time. What's the right amount? You know, and then even mapping out what percentage of our customers would probably blow through those default credits versus what wouldn't was a big tell as to what's the right number here, at least to start. And so we we landed somewhere and we knew we could always shift it, always issue more. So it, you know, you're not locked in here. roll over you know.

I I don't know if we if we would have done things differently because we've somewhat made it a little bit more complicated. much because we don't want to necessarily sandbag and like give our customers all these credits and then only until, you know, 10 months down the line will you ever see a purchase of credits. And so our default credits expire at the end of the like the year within a contract. So if you're in a multi year contract, you you get issued every year within that contract, versus if you purchase any credits, you have until the end of your contract, I'll you know, whether it's one year or three years. that's introduced

Complexity for sure. I'm still supportive of that decision. I think it was the right way to go. We kind of tie the default credits to the seats. So like the logic is there, which I think is half the battle is just making sure your customers understand why you've done it a certain way.

Sales team can explain it. Yeah.

Yes, yes you can explain it. Those are that's that's the biggest sell to start. It's not your customers, it's your sales team. Like they really have to buy into this and that it's in itself is a journey. and then, you know, Gong's one of Gong's principles is create raving fans. And it is not I mean like it is legitimate. That is constantly something that we're looking at. And so we wanted to make sure like we don't have overages, we don't, you know, send invoices. One

you know, you need to have the infrastructure to do things like that. But two, I think even if we even if we did or were prepared to take that on, we didn't want to go in that direction. And so I mean we were down to like, does a customer go into a negative balance? What happens? So it's a lot of like what ifs that you have to map out and then essentially design what the path is gonna be. And things are continuously popping up that we mean, I I hate to say this, but our operational team meets daily still for 30 minutes to talk through all

of the things that are happening and there's always something popping up that we are working through. So it's it's a lot of like it's like death by a thousand cuts a little bit when it comes to figuring all this out.

Yeah. That's awesome context. Yeah. one thing that we've definitely found is like the the longer a company offers credits, the more likely they are to offer rollovers and just like offer general flexibility, which I don't know if it's if that's like just hearing complaints from customers and like needing to do that kind of to like meet their standards or meet whatever they want. or if it's more of like an operational thing. Like first you like launch the credits and then you kind of like build that kind of functionality out. But that's definitely something that we've seen.

Yeah, I think there's also like, you know, what percentage your business is coming from these credits, right? How much of a risk to the business, right? For right for us it's pretty small right now. It's nominal. But as that changes, are we going to draw a firmer line in the sand? Are we looking at these on an exception basis? So a lot of these are decisions that you will probably to your point change as things evolve.

Yeah. Cool. Jeff, what about you at Miro and or Zendesk?

Yeah. Yeah. I mean I guess like maybe I'll I'll take it like maybe a little bit differently. Like if I were doing it now, right? Like there I think a lot of like everybody on this call I think has worked through it in you know, in in the last couple of years and like there's more uncertainty then. And so I think like one thing we're starting to see, especially with credit like rollovers and expiry and stuff like that, is like standard market standards are solidifying. within things like that. And so like I would start with like what the market standard is and like, and then try if you want to add a level of complexity, like justify why that complexity is worth it. And so like that's what I would take if I were doing it now, where I would kind of start. and that might be worth it. And like Zendesk certainly we had some that was worth it because we wanted to handle seasonality, which is really critical for customer support. And like Nero definitely has some because we wanted to, you know, build for PLG and like totally. so I think like there are reasons to do all of them. I think and then to your question of like how do you decide how many in each plan, you know, the really hard part is like in generally speaking, you're like launching these credit models alongside whatever features you're launching with, and so you don't actually know usage prior to trying to like build the model, like that's the really challenging part. So like I would just start conservative and like you can you can change it later as long as you're giving people more. And like if you're trying to take some away, it's really painful. And so like don't do that. so like that would be my like to like how I would do it now if if you guys are starting essentially.

love what you said about first of all defaults earlier, but then also kind of like the standard and like what people kind of expect. And I I think that's actually like really underrated. Like I think that's why there's like good, better, best. Like still, like I think that's why like three plans is so popular, is because like people are used to seeing it. There are gonna be a lot of differences between your prop pricing model and like your product, but like at least people have like a familiar structure that they're looking at, right? And so like

Yeah. And it's like when you when you want to use it as like differentiation, like you can differentiate against whoever your customer like competitors are because we do this thing better and like that's totally worth it. but there's lots of times when you should just do whatever the market's doing.

Yeah. Do you think that like I'd be curious just real quick before we move on to Manu, d do you see any standardization ish things happening with credits, that you think are like starting to become kind of like the norm? And if not, that's okay. But I I'm just curious. Yep.

I mean month monthly allotments with no rollover is where I would I would start. I would start there. Yeah.

The default setting. Yeah. Yeah. Awesome. Manu, how about you?

Yeah, I think like in my experience, Rob, I when we talk about like setting up these credits or or giving these allowances, I think these need to be tied to like the workload frequency or like depending upon what's your adoption, what kind of customer what's the demand plan driving the adoption or even like where your customer is on the maturity phase. So I think that should basically really define the allowances for your for your credits. And and we should also give them like initially what I see is like when the customers are trying to run these POCs or do some experimentation, you need to have some form of promotional credits. And if I were to define the allowances within these credits, I would do it in three buckets. One would be: okay, these are your promotional credits because you're trying to run a POC or you're trying to do some experimentation on my platform. The second one would be the committed credits, where okay, now you are ready to commit with us, and here is a pool of credits that you get. And then finally is the top. Pop-up credits, which is basically this is your buffer, which could be like 10%, 20%, however you want to define, but this is above and beyond, which you could purchase at a discounted price because you've gone and already used all your committed credits. And around all these three buckets that we talked about, I think the way to like kind of allocate these credits should be like at a at the overall consumption level versus like making it very rigid on a certain product level. Because then if you're then the more narrow you scope these credits. the h there are more chances that they will go unused. And if you're like giving this as a pool of credits, the idea is that more likely than not the customer will be able to consume these credits. And the more broader we scope it, there is less chances of like us getting into rollovers or shortfall or underconsumption of those kind of things. So I think like as long as we can be flexible with the allocation, as long as the customer's workload mix or the demand plan is driving where these credits will get allocated. And and and used, I think that gives customers a lot of flexibility. and then I would say what I really see is the last bucket. I want to touch upon like the top up credits, which is basically you got these credits as part of a committed customer, but you already burned through it or you already exhausted these credits. So, and I would say like 80 to 90 percent of the customers easily burn through them. So having those predefined TNCs on customers' commitment, saying that okay, you

have this option to top up, let's say during a stipulated period of time, gives customers a safety net to say even if I burn through these, I have this. When it comes to rollover, I think there is always good to have like a certain amount of rollover, especially for the enterprise customers, because that just builds trust, right? Like something happens with the economy and things go south and then they don't the the workloads never actually spun up, which we expect it to. So having that certain portion of rollover, like certain amount of rollover really helps the customer. Now, do we want to give hundred percent rollover? I don't think so. It's like it's more like, okay, can we do a certain percentage rollover? Because again, At some point of time at in time these credits need to expire. And that's what I think Rita, you also touched upon that like there has to be a certain expiration date with certain portion that you can roll over. So I think like that's how I would see this space is and as said, like those three buckets really come handy, like the promotional credits, the committed credits, and then the top-up credits if they burn through the committed credits. and and that kind of gives customers a lot of flexibility in the long run.

Yeah, that's a super helpful structure. and the the what you were saying around kind of like allotments made me think about something. Like I honestly wonder how many companies like really run into situations where like rollovers are happening a lot. Like I th I think I'm sure they're like and like I'd be curious to get your answers to that, but I feel like there are some apps where that's like way more likely. Like if you look at obviously like a cloud code or a lovable, like back when we were using Lovable heavily, I feel like my partner at Pricing Sass John was like doing top ups like every day. It was like kind of dangerous. But then you get to like clay. Like I remember when Clay relaunched their pricing, like they they basically set the action threshold. So they have two metrics like data credits and then actions. And they set the actions threshold, I think, so that like ninety nine percent of their customers would not trip the actions limit, which I feel like almost makes it almost more like a like fair use policy than like a true like kind of credit limit. So to your point, I think like setting those limits really high, like also is like a nice way to kind of transition people slowly from license to usage. It's like most people aren't actually gonna have to deal with the dynamics of usage based pricing.

Cool. All right. That was a fun section. okay, I know we only have 13 minutes left. People are asking awesome questions. I want to do like one more question for you guys, and then we can turn it over and do like some rapid fire for questions that have come in from the audience. I guess just like I was debating between like surprises, like looking ahead. I guess like I'd be curious just to hear like if you guys have like any surprises about like launching credits or like maybe if there's like one thing that you think people should like think more about if they are like thinking about launching a credit model that you think is super important.

I think just happened. I thought so. Go ahead.

And we yeah, sorry, Rita, we can sorry.

the one y I've I've had this question asked a couple of times, kind of like a learning, right? the one thing that we went one direction in and I certainly would have done something differently is a lot of our credit consuming features are only run by our business or tech admins at Gong, like if you were if you're a customer, right? So it's not like everyday consumers. That's since changed, but that was at least how we launched it. So we thought, okay, there's some controls there. We've got different levels of like governance and whatnot. We'll inform them, we'll write help center articles, we'll do all the things. we had quite a few execs give us a fun little call to say, I own the purse strings, I carry the budget. Why would you not have reached out to me directly on this type of change, especially because this is like a mid-contract change, even though I'm not, you know, have to I don't have to pay anything if I don't want to, right? Like we we weren't necessarily forcing anyone into anything. the fact that we opted, you know, we tried to do this a little stealthier and we opted out of communicating with our exec stakeholders or economic buyers, I think was

something that I would have done very differently. So your comms are crucial. There is a lot of sensitivity in the market right now, with any type of move to credits. And so like be prepared to defend why you're moving in this direction, how you're going to allow customers to forecast, predict, control, see, you know, all the things before you get into it. We did a lot of that, which helped, but that was the

Yes. Okay. That's I feel like that's such a good such good advice. Like I like not to bring it back to Clay again, but like th that memo that they wrote about the pricing change. I was talking to somebody about how it was almost like the the rap at the end of Eight Mile when Eminem's character basically lays out all the things that people are gonna say against him before they can actually do it. And it's like I I do feel like if you're like launching something like this, it that's actually like the exercise that you should do is like just go through every objection and just stay ready.

It's risky. I mean, like I say that clay are like pricing influencers, right? So they sort of lead a lot of what happens. You could so I mean, for all I know, had we gone the other route, it could have been way worse, right? We would have just essentially brought more attention to ourselves versus like the the few escalations we had. So perhaps I'm wrong, but I think there's probably a delicate balance in which you can

do both of those without without making a s like, sure, you can make a big splashy announcement or you can make sure that your exec stakeholder is reaching out to the accounts exec stakeholder and and do this in a a softer way as well.

Yes. That's awesome. Jeff, how about you?

I would echo that. I mean like that we we had the same learning. and like I think I think it's even more common with like, you know, the more your like DNA is like PLG focused versus like enterprise focused, like it it was a you know, communicate early often. 'cause like it takes time. Like people want time sales teams want time to absorb it and ask questions, etc. customers need that as well. And then like the other thing on our side is just like the the granularity of controls required was like far greater than than we knew at at launch. And so like that is the other thing, like you cannot launch it, at least to enterprises, without really granular controls on models and people and teams and and it like it's really a lot. And so that's the other clean from our side.

I one hundred percent agree with that. You only have one chance to launch something really well. And if they're not going to use it because you don't have the controls, you've lost your momentum. yeah, big learning there too.

Yeah, I think like I totally echo what Rita and Jeff mentioned. I think like when we are launching these credits model, I think the the biggest and the foremost thing is having the cross functional alignment. Like just working with your sales, your finance, like what What moves the needle for each one of those teams, right? Even like product and engineering for that matter, like what moves the needle for them. So I think just having that cross-functional alignment is the most important thing for me. And I think the other thing that I've seen, like at least some of the challenges that I've seen, is like sometimes we have these misaligned incentives, right? Like we are driving these credits to drive consumption, sales could be gone for something else, product is count for something else. So I think just having a very good alignment on what these credits mean for each of those internal stakeholders. That helps to go a long way because if everybody's trying to move their needle in a different direction, then we are not solving a collective problem. so I think just aligning those incentives across internal teams and then also having that commercial transparency, like what it means for the customer and what it means for a vendor internally is very important, and that needs to translate very clearly on your order forms or your commercial constructs. So I think those are some of the learnings I've had in my My experiences like aligning incentives internally and then just the commercial transparency that you pass it on to your customers.

well put. awesome. Well thank you guys for sharing so many awesome findings and takeaways. I want to turn it over to some of the questions we've gotten from the audience. So this section we can just like just jump in if you have something to say. Like it's okay. We we don't have to we don't have to go go in order or anything like that. First question comes from Radika who runs monetization at LinkedIn. What level of detail do your order forms slash legal terms carry around X outcome costs Y credits? For example, some sort of credit rate card based on type of outcome or whatever your usage metric is. Is that ever part of OFs or legal terms? Anyone have any thoughts on that?

I guess. Yeah. So we we did a whole table and I I worked with our product technical team, our legal team, our deal desk that owns much of our order forms on like where does like it's essentially a table of like where do things live? Do they live in Ts and C's, do they live in order form or do they live in help center? By and large, things live in help center. Like if you can avoid I mean nothing lives in Ts and C's really, but if you can minimize what you put in order forms, it's better. put it in your help center. You will certainly have whether it's like ELAs or just enterprise customers who are asking to put something in an order form. So like I would ensure that you build something scalable, which we have where you know if somebody is requesting put in the number default credits per seat in my order form, we can do that, but it's by request rather than like as a default. So less is more.

Love that. Anything to add, Jeff Romano?

I think like yeah, I was just about to say like I would like take a step back and like I think like at least for the way I see in my world is like when we talk about application of these credits, right? Like at least my baseline is to like do a very open pool of credits versus tying it to certain products and and the reason being

If you understand the workload mix and if you understand the if you have a very good line of sight of how these workloads of the customers would evolve in the next three to five years, then it makes sense to go narrower scope. But if your demand plan is not very well supported, or if you don't understand like where this this spend would happen, then it's always advisable to go more wider net. And you could do that on your commercial construct, right? So you could like very clearly call out on your order form to say, okay, is this going to be like a narrow scope or it'll be like widely scoped across the overall consumption. So I think just taking a step back and having a very good understanding of how this workload for a given customer would evolve kind of gives you the answer of how you would set this up in your in in your commercials.

Awesome. yeah, that's great feedback. Okay. Next question. Maggie, who runs pricing at Harvey. and I think actually recently started at Harvey. Congrats, Maggie. okay, few questions here. I'm interested in the following topics from the group. One, decision on fungibility of credits across products. I'm actually also super interested in your answers to this, slash product suites and pros and cons there. Two w why don't we just start there? 'Cause she she has a few really good questions, but I think fun fungibility of credits across products and product suites is pretty hot right now. so I'd be curious what what you guys think about that.

One pool. Well like like justify why you would want more. I think it's customers will hate it.

That's part of the value problem, right? Why also are you rolling out credits? speed to market, right? Being able to adjust whatever your goals are and what features you wanna use. So they we we made our credits fully fungible.

Nice. okay, cool. Second question here aligning credits to tasks. yes. She she says yes, so I think she agrees with like that general idea. But what about general purpose horizontal parts of the product? How do you have customers' budget predictably here when tasks aren't as clearly defined?

I mean, I think that's the challenge, right? It's it's it's really difficult.

Yeah, that's the hard part. Yeah.

I don't have a good answer for that. Like we have estimators and tools and whatnot that we're using because there are different ways in which you can use credits. And of course, it's like the number one thing particularly large enterprises asking is how many credits will I need? Even if I give you an estimate, if you change your strategy as to how you use and leverage Gong tomorrow, it will completely change. So we've actually pivoted a little bit of our narrative. I think there's a few questions in there around like controls. So this came late for us. Like we had some very basic level controls and governance. We've since rolled out credit profiles where you can lock in by user by feature how many credits per month they use that's what you should be budgeting against not how many you're gonna need start there we'll give you use cases case studies examples of how you can use trackers or mcp or enrich or any of our tools or products and and But like it varies so much by customer and it's not like large customers use a lot. For example, we have we have these smart trackers and a customer can use 10,000 credits for a smart tracker per day or 200, right? It depends on how you build it. So it's like my least favorite question to answer, but it's just giving those tools and making sure that there's something that gives them that confidence of how much to spend or how much they will spend versus keeping it open ended.

Yeah, like Miro Miro's Miro's pretty horizontal as well. And so I think like we see a lot of this. Like I would I guess my the one thing I would add is like you can customers, generally speaking, want closest to outcome as possible. And so when you can do that, it's usually advantageous, but you don't have to do it for everything. And so like if you have things where you can say, like this discrete thing costs this many credits, like great.

categorize those and talk about those. And then like if there are things that are, you know, really challenging to talk about in that way, like categorize those and talk about those differently. And like that's reasonable and like it helps customers as much as you're able. It's it's as much transparency and clarity as you can give them. And like that's all you can do.

I love that. Manu, any thoughts on this one?

Yeah, I think like this goes back to like anytime like you're giving these credits, right? Like whether it's horizontal or vertical, I think The main idea is if you have a good understanding of the workload mix and how this the future looks like, at least in the next, I would say two to three years, that kind of falls into that second bucket that I talked about, the committed credits. And then there is always this option of like some portion being unknown, let's say 10 to 20% flexibility or buffer, which kind of comes from your top up credits. So I think that's how I would see this. Like, yes, there needs to be a good understanding with the ability that there is flexibility for the customers if they go above and beyond. so yeah, I think that's that's how I I see this.

Awesome. I just realized we're already over time. So we already hit it. this was awesome. I I'm seeing some of the questions in chat. so I'm gonna I'm gonna log those and I'll start a thread in the pricing's asked community. If people are not in there, join in there and we can get Rita, Jeff, and and Manu in there as well if if you guys want to chime in. this was awesome. I'm really excited to do more of these. This was really, really fun. I really appreciate and thank you guys for coming and sharing all of your experience. and I'll share the recording with everybody. We'll definitely be doing more of these. So thanks, guys.