The Iconiq State of Software 2025 Report

Live from Schenectady, New York, it's SAS Talk with the Metrics Brothers, Growth and CAC.

And I'm Growth, better known as Ray Reich, founder and CEO of Benchmarkit.

And I'm CAC, better known as Dave Kellogg, EIR at Balderson Capital, independent consultant, and the author of Kellogg.

And together we are the Metrics Brothers.

And we go together like ICONIQ and benchmark reports.

Oh, brother CAC, I think you were supposed to say benchmark it and market research.

Well, well, Ray, I hate to break this to you, but other firms do produce benchmark reports and ICONIQ is one of them.

Yeah, I know you're right, Dave. And they do fine work.

They do, Ray. And, you know, look, I think there's something we should commend them upfront for, which is we've struggled with definitions and methodologies in some of the recent reports we reviewed. And ICONIQ does a great job at being descriptive and transparent, both about what they're calculating and how they calculate it.

Yes, Dave, they're exemplary. Oh, excuse me. Just as we are here at Benchmarkit.

Exemplary. Easy for you to say, right? They are, and I know you guys are too. But before we dive into today's topic, which is ICONIQ's new State of Software 2025 report, Let's hear a word from our sponsor.

The Metrics Brothers is proudly sponsored by Benchmarkit, the B2B AI and software industry's largest, broadest, and most contextualized benchmarking index. Dave Kellogg and Ray Groth Wright have put the Metrics Brothers podcast on hiatus for the remainder of 2026. All episodes we have ever recorded are still available on your favorite podcasting app. You can reach Dave at Kellogg and Ray on LinkedIn at Ray Reich. Thank you so much for listening to the Metrics Brothers podcast.

So Ray, this is a new report. It was published in September. It's called State of Software 2025. It's 73 pages long. And as we go through the episode, we will mention what slide we're looking at as we go through. Here on slide 4, 127 companies are included. Iconic companies are included as well as 10 public companies who are not iconics, 25 who are, and ergo 92 private companies. Some of the names in the report, people like Airtable, Dataiku, Flowcast, Vic.ai. Some of the publics, Alteryx, DocuSign, Snowflake, Zoom. Other publics include Atlassian, Cloudflare, HubSpot, and Monday. So Ray, this is gonna be a tough set of comparables.

It is. These are the Ivy League of SaaS companies, but let's go.

Oh, okay. Let's look at more specifically the firmographics on this thing. So of the 127 companies, there's a healthy distribution across ARR size range, across industry sector. About 75% are private, 25% roughly public. Most of the companies are selling to mid-market to enterprise targets, 43% in the Bay Area, basically about 50% in the rest of America and 10% in Asia-Pac and Europe. Anything else on the firmographics, Ray?

You know, the only thing that jumped out at me was they did it by AI product maturity. And they say that 59% of these companies are non-AI. And I think it's important that they define that as SaaS companies without AI products or features. So that jumped out at me.

Yeah, that's a surprising thing to have there, Ray. Not entirely sure what to make of it, but we'll talk about it as we go through the report. Now I'm going to hit the tops of waves on this, uh, slide 8 of their report. But look, public company valuations remain below historical peaks. We know that Rule of 40 has emerged as the strongest predictor of valuation. We'll cover that in an upcoming slide. Beat rates are recovering. Interesting, because they're talking about public companies beating earnings. So that's a good sign. AI is a pervasive theme. This was a fun one, Ray, where they looked at earnings call transcripts and they found that 94% of them mentioned AI. Thoughts on that one? And how is that consistent with the other one?

Well, if 94% of the market's talking about AI and only 59% of SaaS companies have an AI product, I think there's a market opportunity or issue there.

Yeah, yeah, there's something going on we don't totally understand. Uh, look, just skimming tops of waves in this report on slide 12, they talk about revenue multiples, which is a forward-looking enterprise value to revenue multiple for public SaaS companies. The median is stable at around 5. The high-growth companies, quote unquote, those growing greater than 20%, it's at 12. But we should mention here, Ray, that one of the few criticisms we have of their market research is that they don't— they flip a lot between looking at the median and looking at the top quartile. Did you notice that?

I do. And that's why as the readers go through this report, look at the top of every one and see if it says top quartile or if it says median in the footnotes.

Now I'm going to look at slide 13 real quick, Ray, on Rule of 40. And I know there's some inconsistent data out there. So this can be confusing. Quick history on the Rule of 40. It was actually created on the theory that a metric that measured a balance of growth and profit would be a better predictor of revenue multiple than, say, growth alone or profit alone or some other metrics. So this thing was kind of invented to predict revenue. Um, years, years later, Bessemer came along with Rule of X and said, hey, we can do a 2. Roughly a 2.0, 2.3x growth-weighted Rule of 40, and that's going to predict revenue even better, right? And now people measure the R-squared between different metrics and how well they predict enterprise value to revenue multiple. And in their latest sample here in Q2 2025, the strongest predictor is Rule of 40. with a 34% R-squared to the EV-to-revenue multiple. Revenue alone looks to be, if I can read through their colors, 20%. So R40 is doing its job. It's predicting value better than revenue alone. What are your thoughts on this topic, Ray?

Well, I do find if you look at revenue growth, if you look at, you know, back in early 2021, the R-squared for revenue growth to the enterprise value-to-revenue multiples was 49%. And that's all the way down to 20% today. And Rule of 40 back then in the go-go growth days was only at 12%, and now it's up to 34%. So that's the one thing. And then, Dave, I have to go to just slide 14 real quick, that R-squared. What's really important is even though Rule of 40 is balancing growth and profitability, free cash flow margin, growth is still about 1.9 times more important to that multiple than free cash flow margin. So I think that's important to read.

Yeah, and that's pretty stable over time, right? That's the genesis of Rule of X. I think when they published that note, they felt it was 2.3x, if I remember the number. And Dave Kellogg said, I just use 2x. Like, to me, the best overall predictor is a 2x growth-weighted Rule of 40. And the funny thing, Ray, is I'd actually argue that you can use this relative importance As a bubble indicator. So when this metric's running around 2, we're looking at slide 14 here, that's normal times. Like growth has 2 times the impact of profit because it should, right? Because growth compounds, profit doesn't. But when that multiple gets up to 4, 5, 10, you can know that you're actually in a bubble. So I have a— I kind of use this metric backwards maybe to say, hey, if revenue growth alone is predicting revenue multiple better than R40, or if the relative impact of growth in R40 is like 5 to 10 times, uh, free cash flow, then, then hey folks, we're in a bubble.

Dave, the only thing I'll say is I'm thinking back to our State of AI episode where we talked about supernovas and shooting stars.

Yep.

I would love to see the balance of growth to free cash flow in those private AI companies.

Yeah, well, they got plenty of both, right? So hopefully it still offsets. That was the whole point of R40. So quick look at slide 17, just a lot of capital going into AI. The most interesting thing about this chart and why I call it out is it looks flat. This is a bad charting practice, folks, but they're showing— they're ending the chart with 1H 2025 comparing it all to all of 2024, and it makes it look like things are flat. But the reality is It's flat when you compare the first half of 2025 to all of 2024, i.e., it's on track to double. $377 billion in, in the first half of '25 compared to $363 in '24. Any thoughts on that one, Ray?

No, I totally agree with the chart. And the only thing I would say is I'd love to know if this is just VC and PE money or if it includes corporate investments also, because there's so many corporate investments in AI. companies today.

Yeah, I suspect it has to, but, but short answer is I'm not sure. I'm going to slip over now to slide 19 where they talk about some key top line and efficiency metrics and how they're trending longitudinally, i.e., over time. Uh, I'm just going to eyeball some of these charts. In my opinion, growth rates have basically stabilized. They're down from like the 2020 days, but growth by segment is roughly flat. There's a couple of small trends that I won't bring up here. One on slide 21 I will bring up is that NDR seems to have settled into a new range that we can now remember as a rule of thumb, which is what I like to do. And in the first half of 2025, for companies less than $100 million, the top quartile, not the median— there we go again, never quite sure what you're looking at at the iconic data, so read the fine print. is 121%, and for greater than $100 million, it's 110%, roughly. Thoughts on NDR and where it's stabilizing, Ray?

No, I think the only thing I'd take away from this slide beyond that it's stabilizing is the end count. The end counts at the bottom of each of these slides, and the end count was at, you know, like 233 for the less than $100 million in 2020, and now it's down to 41. Above $100 million, it's 33. So I'm just questioning why the N count's going down so much, but that's the benchmarker in me.

Yeah. And we're going to see this. Look at slide 23, Ray. You'll see, I'm going to go to the N count or the N size. And basically at the bottom of the slide, and this slide is looking at CAC payback period, and it goes from 142 in their pre-2020 benchmark to 21 today. So, and when these numbers get that small, you got to be careful of what you're looking at. Even last year in 2024, It was up in the 90s. So I don't know why you have a sample of 127 companies and only 21 are included in this question. I'm not sure why, but, but keep an eye on the N. I think that's the rule.

Yeah. The other thing, since you're on slide 23 here, which is CAC payback period, which is at 15.9 months, the only thing I would recommend on this one also is that if this is only private companies, because they calculate CAC payback period on new ARR, But public companies calculate CAC payback period based upon implied net new ARR. So there is a difference between private and public.

Yeah. And knowing that, Ray, and also knowing Iconiq, 2 things. One, I bet if we went footnote diving, we could find the answer. And 2, the obvious correction to that is you do the privates on net as well, even though they don't typically do it, right? Because you can't do the other correction because you don't have the data. But in any case, the trend here is that CAC payback periods according to them, again, top quartile, is 16 months, which frankly I think is a real good, particularly in, you know, mid-market enterprise sales, that in my mind, that's a real good CAC payback period. And as mentioned, we've already talked about just how difficult these comps can be. Um, the, the next slide I think we should talk about is Rule of 40. I want to go to slide 25, Ray. And the issue I have with these charts is they never show the overall average. They're just showing the segmented values. So you can't look at this chart and go, oh, the median R40 for the whole set was blah. What they give you is the top quartile segmented, right? So it's a little tricky to read, but presumably if you work at a portfolio company, if you're listening to our podcast, you can go find the line that correlates to you. And if you're a $50 to $100 million company, you would find that the top quartile R40 is 36. Which strikes me as a— for that size range, a pretty good R40 score. Any thoughts on this chart, Ray?

No, the only thing that jumps out at me is number one, the fact that $200 million is at 57%. That's really, really good. And if, if the top quartile is 36% or 31%, even though it's good for the smaller companies, doesn't sound like a rule. Sounds more like the goal.

Yeah, yeah. So, okay, in any case, net magic number, they do. I agree with you, Ray, by the way. Um, on net magic number, it's the next chart there on slide 26. I don't love magic number. I view it as an inverted CAC ratio, right? People know that. Also, same issue Ray raised earlier, privates versus publics. Privates tend to do it on new ARR, publics do it on net new or net new implied, really. Um, so what they're showing, it's around 1, which is fortunate because when you invert 1, you get 1. Um, and, uh, So, and that's a really good— when you look at net magic number, I think of it as an extraction ratio, Ray. I think of CAC as how much you spend to get a dollar of new ARR. I think of net magic number as how much ARR you extract from a dollar of sales and marketing spend, right? Because I needed some visceral way to think of it. And look, it's around 1 if you want to be top quartile. So roughly consistent with the CAC payback period. ARR per FTE, Ray, on slide 27. Again, these numbers strike me as high, right? Their smallest company segment is nearly $200K median. Now we flip it to median in ARR per FTE and their big companies is $330K. Any thoughts on this data or this metric? And again, look at the end, Ray. You're segmenting 63 people into 4 buckets.

Yeah. Yeah, very small. I guess what I would say is in the research we've been doing lately, we do see ARR per FTE has ticked up anywhere from 10 to 20% per segment over the last couple years. And this looks pretty consistent, right? If I go from $308 to $331 for the $200 million plus, yeah, that's up, what, 23? So that's about 8, 9%. So this, this fits with what I'm seeing out there, Dave.

Okay. Um, slide 32, they've got burn multiple. I don't think it's that useful with this coarse segmentation. They're basically saying below $100 million and above $100 million. Well, above $100 million, burn multiple in my mind starts to be a little bit— well, it's not— it's never irrelevant, but it's not as important a metric. And below $100 million, it actually changes quite a bit with scale. Like, there's a difference between burn multiple to $5 million and burn multiple $50 million or burn multiple at $90 million. Here they're showing the median. So now we're back to medians. Below $100 million is 1.7x. This is generally consistent with my rule of thumb, which is you want your burn multiple to run around 2 or less. But if you're getting up into 3, 4, 5, 6 range, it's getting awfully, awfully high. Any thoughts on burn, Ray?

No, just exactly what you said. If you're a $20 million versus $80 million company and you're VC-backed, it's going to be very different. Too coarse for me too.

Yep. So, uh, on slide 33, they have some interesting kind of survey questions here, and this is just about, uh, priorities. I'm gonna hit it real fast, but in top— in order, the top priorities are GTM acceleration, distantly followed by efficiency and NRR, right? At least in terms of what's our top priority. Um, And then AI commercialization. So, and those top 5 are really, I guess they're top 4. I don't know how they ranked them, but those are the top 4 priorities that people are focusing on. Only an N size of 58. I just think it's interesting when they throw kind of non-quant questions in. Any thought on this one, Ray, before we go over to look about AI?

Yeah, I know we were talking about how great they are at defining things. I didn't see how they define GTM acceleration. Is that growth acceleration? Is it growth efficiency? I don't know what it is.

Yeah, really good point, Ray. Really good point. And it's kind of a weird presentation of it too, but, uh, good point for a company that usually defines things really well. That's missing. All right, let's skip on to the AI section, the age of AI. They do some comparison contrast of SaaS and AI companies, or traditional SaaS and AI companies. Um, there's a little bit of kind of philosophy on slide 35, which isn't bad. They then talk about some of the new metrics people are tracking. Nothing that exciting here, Ray. They basically say, hey, metrics are in 3 buckets. Adoption, people are looking at stuff like user count, net new ARR, free-to-paid conversion. People have always looked at that stuff. Engagement metrics, wow-to-mao ratio. Okay, pretty normal. AI mentions, renewals. Okay, I didn't see anything particularly new here. They look at business outcomes. You know, number of resolutions handled by AI, cost savings, CSAT, number of agents deployed. Anything strike you on slide 36, Ray?

I was at an event last week and we asked how many people measure the business outcomes of their customers, and it was less than 20%. It's like SaaS software was supposed to improve business outcomes, right? Increase revenue, decrease expenses, increase productivity. I'm glad to see that we actually think we're doing it now with AI.

Yeah, yeah, yeah, true. I mean, we're always getting closer. It's kind of like a slow asymptotic convergence on business outcomes. I mean, even back in the day when we sold relational databases, the idea was that would let me, you know, you could flexibly analyze your data to make better decisions to run your company better, right? And that dates back to my first day in the software industry. So we've been going after this for a long time and delivering along the way. But there's still plenty of work to do. They have a section on the age of AI. They talk about something we covered a few episodes ago, Ray, which is, well, we call it months to 100 million. They, and I'm going to give them a little dig again, sometimes they use years, sometimes they use months, sometimes they use quarters. So they got to standardize on a metric here. I think we should standardize on months personally. And they talk about ElevenLabs that took 2 years to 100 and got there at 150 people. Cursor, 1 year to 100, got there at 20 people, which still blows my mind. Glean took 3 years to get to 100, got there with 1,000 people. And then Lovable, who took 8 months to get to 100 and got there at 45 people. So we've talked about this metric. Any thoughts on this, Ray?

Well, they did something very good on page 38. On the x-axis, it says quarter since $1 million ARR. They told us the starting point.

Good point. Because I was giving them a dig for switching to quarters from years or months. But, but they did something highly commendable, which is they told you the starting point, which is nobody ever does. They just go months to $100 million. And like, from when? From when the company was founded? From when they sold their first license? Or from when they had their first trial? So good for them. There's a nice chart there. We've talked about the crazy growth rates of AI. On slide 40, they talk about some what I call hockey stick companies, which companies where growth decelerated and then took off. They talk about Palantir, which is obviously a super hot and well-followed company right now. My favorite example is Bolt, which I think had existed for, I don't know, 5 years. And then they hit the hockey stick and went in less than a year. I think, I can't remember the exact number, But so it wasn't from inception because it would have been like 60 or 80 months, but from hockey stick, they did it in like 12 months from effectively, you know, $1 to $100 million in ARR.

Hey, Dave, I can't believe it. We've only got a couple minutes left. What else do you want to cover here?

Yeah, I think we could wrap this section up. I want to do a couple other comments. They do some comparisons between AI-native and non-AI companies. And as Ray pointed out at the beginning, we're a little worried about their definition of non-AI because they're saying that 60% of the companies have no AI in them. And that's just so not consistent with anything we read or see. So you should view these comparisons with 2 grains of salt. One, that we're a little worried about how they define that. And 2, they don't actually show any material differences. And maybe this is why, like, I don't see a lot going, wow, when I look at slide 43. They do an interesting idea, which is look at how headcount is distributed. But once again, I don't see any major wow. And maybe the major wow might have to do with the fact that there's something wrong in the taxonomy. Not sure. So they looked at funnel, then they— oh, the last thing we've got to talk about is here. They talked about forward-deployed engineers, and this was great. So they looked for how many people were hiring them. And forward-deployed engineers is In short, a super hot hiring trend. And at first glance, it appears to be the same thing as a new name for post-sales consulting, but it's not. And we may have to do an episode on this, Ray, to talk about what's a forward-deployed engineer versus what's a normal post-sales consultant. And I think I, and I'm pretty cynical or skeptical, think there is one. But this is just data saying in February '24, The number of vacancies were 30 for FDEs, forward deployed engineers, and now it's up to 375. So, so this trend has caught fire. Thoughts on this one, Ray?

Now, just on the same chart, they say that FDE, the forward deployed engineers, should be able to realize at least a 5 to 10x return on that investment. Number one, I'd really love to know how they measure that, because to me, the reason you do forward deployed engineers is to make sure that you get them up and running quickly, that you get the maximum adoption and you have the most revenue over 12, 24 months. So I just wanna know how they're gonna measure that.

Got it. So I think we'll wrap up, Ray. They do a slide on burn multiple, which we won't talk about again. I mean, short answer is it looks like it's around, around 2 is the right answer. The only interesting thing on the chart is they say that for AI-native companies, It does drop a lot above $100 million, which it never really did for some of the non-AI companies in this sample, right? It was pretty consistent between below and above. So that was interesting. The headcount split doesn't make that big a difference. So I'll wrap up by saying, look, they have what they call the iconic Enterprise 5, which I think are 5 great metrics, and they show them split in significant detail here. And those metrics are ARR growth, Net dollar retention, rule of 40, net magic number, and ARR per FTE. And if I were on an executive team of a startup, I would definitely be looking at these 5 metrics. And the report then goes through and shows you a pretty detailed segmentation for each of them. So this is just useful benchmark data, in my opinion. And then the report wraps up. Any overall thoughts on this, Ray, or thoughts on the tail of the report?

No, I would just say, you know, those enterprise 5 starting on page 52 are really good and I like it. And what's good, I love what Iconiq does, but at the same time, every time I read the report, I guess I have such high expectations, I want even more from them. So, but good on them for doing it.

Yeah, and I think, look, the only thing I'd say on this is I love their research too. I think they are by far the most rigorous of the kind of VCs who publish research. My net is always, this is like VCs wanting the best thing for their children. They're always setting like a really high bar, right? So we've got a very tough set of companies that we're comparing to. And then, for example, on slide 52, what are we looking at? Are we looking at the median? No, we're looking at the top quartile of the kids who got into Harvard, not just the median kid that got into Harvard, right? And that's always the thing you need to bear in mind with iconic research is I do think You're gonna feel bad when you look at it. And you need to look at it and go, wait a minute. I mean, first, if you're trying to go to Harvard, if you're trying to go public, this is the bar you got to hit. Very useful for that. But, but if you have a different aspiration for your company, this, this research can be depressing because you're looking at literally the top of the top when you look at these numbers.

Hey, Brother Dave, thank you for this one. It was another good and fun one.

Thanks, Ray. The Metrics Brothers AI, SaaS, and More is a production of the Metrics Brothers Growth and CAC and a member of the Benchmarkit Podcast Network. By accessing this podcast, you acknowledge that the Metrics Brothers make no warranty, guarantee, or representation as to the accuracy or sufficiency of the information provided or the humor content of the jokes told. Ray, information, opinions, and recommendations presented are, according to our spouses, probably wrong and provided for information purposes only. This podcast should not be considered professional or for that matter, unprofessional advice. We disclaim any and all liability for any direct, indirect, indirect, misdirect, incidental, special, ordinary, consequential, inconsequential, or other damages arising out of any use of, or God help you, reliance upon the information presented here. Ray Grothreich is based in New York City and available on X @RayReich. Dave Kellogg is based in Silicon Valley and available on X @DaveKellogg. Kellogg. Schedectity, which is French for unspellable, is not our actual production location. Thanks for listening.