Numeral raises $35M Series B at $350M valuation to automate global sales tax compliance

Thanks so much for sharing your merch. Oh, fantastic. We love hats here. I'm throwing it on. We got our second guest. Uh, introduce yourself for the stream for those who don't know. $350 million man. Hello. How's it going? I'm Sam, co-founder, CEO of Numemeral, which you know, of course, because which you know and love.

We have been partnered with Numel for a number of months now. Uh, kick us off with the news of the day. The news of the day is we've raised 35 million on a 350 post. Congratulations. Take this. Hit that gong as hard as you want. Wait. Subtle but firm. Not too hard. Uh coming off series A from Benchmark as well.

Was that early? Exactly. We did a series A from Benchmark earlier this year and so just following on there, you know, capitalize on all the momentum. Who did the double down? We did it with uh Mayfield. Mayfield. So classic fun. Got a classic fun. Got to keep it old school.

Been in the business for generations for for decades. Uh, we got to update Jord's ad read for the probably the first 50 ad reads he was benchmark series A. We had a lot of fun. Don't forget Don't forget YC as well. YC. Love it. When did you do YC? We did YC beginning of 2023. Okay.

Um, actually working with my old boss at Airbnb, Gustav. So, it was fun to get to work with him again. No way. What uh when did you know you wanted to automate sales tax? Oh man. Uh, since the womb actually. I've been dreaming about it, you know.

No, for me, you know, I was similarly uh as you guys like actually running e-commerce businesses, still own a vitamin gummy shop doing 10 plus million a year. Amazing. And so why what is it about gummies that just sell better than any product on people love gummies?

They don't want to It's the best e-commerce business because, you know, you get that repeatability. It's easy to it's easy to predict. You take two a day. And I think people don't like pills. They love Americans. We love sweets. We love candies. So little treat. Um that's great.

when you discover the problem through that business. Exactly.

So my co uh you know my my co-founder there and I of the gummy business we were just dealing with um sales tax compliance for all we had to register in 40 plus states and so I was dealing with you know local parishes in Louisiana and like on the phone with some women in West Virginia and I got like an argument about penalty waiverss and I'm like this is insane that I'm trying to spend my time on growth and marketing or product and things like that and instead I was wasting like hours just dealing with you know the IRS compared to these states is like working with Apple or like Google or some like amazing company like Department of Revenue Alabama like give them a call and it it's insane.

And so I think the thesis was you have you've had these incumbent tools that have been around for 20 plus years. Um the most famous one is this one called Avalara. And the problem with them is it's a it's like kind of this this very narrow tool and you have to spend hours and a month dealing with them.

So you get physical mail from the states and then you have to figure out what to do with it, how to call the state and they won't even pick up the they're owned by private equity so they don't even pick up the phone when you have an issue.

Um, so for us it's, you know, we're not really trying, you know, it's software, but we really, if you're a customer, we really want it to feel like service more than software because, you know, with AI, since we invented AGI, you know, we're using we're using sales tax AGI. Sales tax AGI, right?

And so, um, no, but yeah, with with AI, you can deliver what feels more like a service. Um, but, you know, fundamentally, it's still a software business. What was the fundraising uh process like? I want I want to know uh just take the temperature of like you just went out and did a fundraising round.

You talked to a lot of VCs. Were people trying to get the story out of you? Was it Vibe round, Excel round? Like what is the focus of Silicon Valley VCs right now? Um since AI has become a tool that pretty much every company's pulling off the shelf um just what was the what was the vibe on the road show?

That's a good question. I think yeah, you have to AI has to be, you know, front and center. Uh any friend that's, you know, is raising and and not putting their front and center, it's hard to actually get a meeting or or be taken seriously.

My sense is, you know, it's starting to get maybe a little too far over the edge. I've seen some insane uh some insane markups.

Like we intentionally um actually took like a lower markup than we could have just because um you know, I just tend to run these things a little bit more incrementally and and and don't want to you know um overextend yourself.

Um I've seen I've been in startups where we went a little too crazy and and like you know it's easy it's easy to kind of just you have too much money, you spend too much.

So, um, but but yeah, it's if you if you have I think too if you have like some good AI logos of companies you're working with, we're seeing I've been seeing some like 100x 200x mult, you know, multiples out there, it feels a little bit backward just because they have they're working with like a lab or something like that.

If you Yeah, exactly. If you can close a lab, a couple labs, like you're going crazy. I mean, in some cases it's worth it, right? Look at Merkore, right? They're growing so fast and and that's their core business. That makes sense with Merkore.

Um, but yeah, if you're more of like a software tool that could be used by any company, not not directly indexed revenue-wise to the lab revenue, which is growing very fast. Yeah, it's a little bit of a just like it's a good good, you know, they're obviously cutting edge businesses.

They probably select the best software. So, but it's it's definitely starting to get a little frothy, you know, not putting my I'm not putting a ton of my own money into uh Yeah. Do you think that AI is changing any of like the fundamental physics of like the the business models?

like we we we we covered the story with notion. Uh notion said they normally have 90% gross margins, now they have 80% gross margins, but they're growing and so it's it's mathing out.

I feel like with as we enter like the SAS apocalypse where the legacy SAS players aren't moving fast enough and we see a new generation of of companies come up, I'm I'm wondering how much people should update on the actual like foundational like is there something changing about the monopoly thesis, how companies acrew value, the margins, what the actual long-term cash flows will be.

It feels very much like like it's a big opportunity to build something new, get new customers on board, grow a big business, but it doesn't necessarily change the actual structure of the business. But what's your take? Yeah, I think each business is different.

Um, so yeah, with the with the gross margins, even we've seen some compression because we're spending so much on running agents and um yeah, we have, you know, if you have a state portal, so many people aren't willing to admit that.

as they're like and it's like it's actually a good thing if you're spending a lot on inference because it means you're actually using you're not just saying you're well a lot of people were also on credits for a long time and so they were doing and or the credits were really discounted but now we're in more of a stable inference cost $4 per million tokens if you're doing something serious yeah our AWS bill I think doubled um like relative to our revenue right so it's like our gross margins have actually compressed but I think our product has gotten a lot better and you The hope is eventually over time, you know, these the inference costs go down, the models get more efficient.

And so I think to me that's like a little neurotic to worry. So I mean it depends who you are, but to me that that's like a neurotic thing to worry about like the short-term gross margins.

I think for me, at least for my business, where I'm excited is, you know, the services economy is like at least an order of magnitude larger than the than the software economy.

And so a lot of what we're starting to replace now is not really there's these legacy software providers but again as you bigger companies usually they don't even work directly with them. They'll hire an accounting firm to manage Avalar or Vert. Ex or these old school solutions.

They're like this this software is so rough I don't even want let's hire somebody to that. Like I've literally been in that position trying to buy the competitor product years ago and uh and being like it's expensive and also I don't have the man power to do it.

So we'll just continue to just like file whenever we can and it's very rough. It's all done in a spreadsheet. Yeah. I mean we are the uh official I have the real one. Yes. I got the tin here. The official tax provider of Lucy. My DTOC nicotine. They're going to get it right. We are powered by numeral. It's fantastic.

Yeah. And so it's interesting right? So, for example, we started we filed taxes for someone this week in Tanzania and Kenya and like we're the only software provider on earth that's I think ever filed taxes in in Tanzania and and so I think uh let's go Tanzania. Hit the gong for Tanzania. There we go.

You can you can send that to the office there. Whoever chatting with um Yeah. Yeah.

what uh what what you guys are scaling quickly and it's and it's a it's a category that's like not it's an interesting one because it feels like uh like investing in ecommerce infrastructure was at its hottest like maybe like four or five years ago and so is do you like that it's like a bit overlooked as a category and like where is most of the grow is the most of the growth coming from like online retail companies or is for SAS and great question.

What's that split? Yeah. So, we started the business fully focused on e-commerce. That's where I was coming from. I mean, I'd worked at software businesses before.

I was a PM at Airbnb, but um we started just in ecom and I think yeah, that market's been really dry for you know people uh servicing the e-commerce um space like the you know software the software provide providing for that space because the TAM is somewhat limited.

You know, e-commerce growth is something like 5% year-over-year. This is a problem. Yeah. you saturate and then you're growing at 5% and no VCs are giving. Exactly.

Most of the good businesses I see buil being built are either bootstrapped or vin or more like a venture light model which makes sense just because the TAM is going to be more cap more like an app in the Shopify app store plug and play do something minimal scale it really become like a public company. Exactly.

I think we were the first company benchmark funded that's actually been servicing uh you know e-commerce merchants. Right.

But again, and was that do you think they had a sense of like this numeral while it looks like a Shopify plugin or or just like an e-commerce infrastructure tool, it's really going to be eating into the services revenue from these legacy tax like prep.

I mean, it's so crazy with Lucy like we did not rip out and replace to get unnumemer. We went from nothing and this was like a seven eight-year-old business at the time. Yeah.

And so yeah, so like just to benchmark like our our partner Chathan, you know, when he was early in his career, he actually was did an analysis on Avalara, you know, they were started in 2006, I think, Avalara. And so they were just going public or something. He's like, "Wow, this business should be amazing.

" He checked back 10 years later and it was just being operated really poorly. That's why they had a you know, host private equity takeover because it wasn't they weren't operating well. Um so I think he he had seen this category and was like waiting for there to be someone to go and challenge.

Um you know, we started fully e-commerce. Now, you know, it's it's a closer split.

we have a lot more SAS and e-commerce businesses especially with our international pro um with our international service with the the only um player in the space that can actually go and register and file in 60 plus countries and so I feel like if you're if you're making some consumer AI app and you're selling to users that are in Tanzania or something exactly right so Lucy is probably not selling a lot in Tanzania but you know we have like these these hot AI companies that like the one in Tanzania they're they're like a year a year old but they have you know hund00 million of ARR and so you have customers around the world and you know there your stripe dashboard is blowing up with all these warnings and people don't you know they get freaked out but they don't deduce they just kind of kick the can for a long time but you know sometimes you can you know do that and not get in trouble but some places will just like like you know you don't want to mess with those countries right you don't want to mess with Mexico like the authorities there yeah it's pretty it is a pretty new phenomenon that uh I mean there were a lot of apps in the last boom of you know software that might internationally distributed because of AWS and and because of Google uh on day one, but they weren't necessarily taking money from individuals in that comp in that country on day one.

It's more like there's there's ads that are being shown there, but the companies that are paying are in the United States or they're doing something else and they're just like, "Oh, we'll monetize it later.

" No, these these companies that are doing uh new AI products or something like they're putting the credit card down in that local country on day one and and then you need to actually process that, right? This round this round got this round got leaked by R for Rock, right? Yes, they did. What was that?

What was Who do you think Who do you think are for Rock is? What was the impact? What was the impact from that? Like did it was it helpful at all? Hard to know closed when it The round the round had already closed by then, so it didn't really matter. But, you know, it was a little bit low on our revenue.

Um, so I did retweets because they clocked the revenue. They clocked the revenue. Not like crazy amount low, but you definitely uh seven figures low. And so yeah, I was like, come on.

Like we have the most AR of all the sales tax uh startups in the space and the most most customers, the most money raised and so you know, you always want to be a little bit cocky there because it's hard to know. Yeah. What do what do you like uh projecting out a few years like what do you think happens to the category?

Can the category support number of players? Do you think some of these venturebacked competitors will just end up being like a decent little business?

Uh good question or uh do you think I mean there's a bunch of you know massive massive you know tax prep you know big big consulting so maybe it can support multiple billion dollar players.

So when you look in the public markets right you have companies like Alvalara who is take you know they're they're going to go public again next year. Um people are saying like 1. 1 1. 2 two billion of revenue. Um Thompson Reuters has a subdivision doing like indirect tax. Yeah.

Called One source secretly a billion dollars of revenue supposedly and like nobody it's like this thing nobody's looked at you know it's like a plug into SAP and Oracle and so the tool looks insane like it's built in the 1970s but you know if you're some big if you're IBM if you're IBM like or whoever um you might be using it and there's another one Vertex Inc.

founded in the 1970s, another public company, sales tax software. Um, you know, it's but companies like Door Dash still rely on Vertex and they racked physical service in San Francisco and to get the right tax rate. I think they updated with like a CDROM to make the rates get updated.

So, are you how much time are you spending how much time are you spending like basically doing these like much bigger highlevel partnership stuff where you're going to these sort of scaled software companies and saying like, "Look, the way you're doing this is insane.

it's going to take time to switch over, but like there's a better way. Is that where you're spending a lot of time? Yeah. So, you know, we started more in like the kind of SMB e-commerce space. And so, that's where a lot of our early traction's been.

Now, we're starting to have conversations and and starting to sign customers with hundreds of millions of revenue, work with a customer of you guys, um, eight sleep, right? So, that's another unicorn. And so, we're starting to work with bigger and bigger businesses.

still really focused though on software, you know, kind of startups and and and direct to consumer businesses. We haven't gone and started talking to with legacy businesses a lot.

That's, you know, working with accounting firms and things like and like the big four that, you know, this just takes a long time and we've obviously had some of those conversations, but it's just not our focus. What's the ladder for signing bigger and bigger customers?

Do you have a rule of thumb like never jump more than one order of magnitude?

Like if you if you have a company that's doing a hundred million in sales, you feel like, yeah, we could probably do a company with 800 million in sales, but if a company with eight billion showed up tomorrow, that's going to be after we sign the $800 million contract. What do you think? Yeah, good question.

I think yeah, order magnitude seems high. You know, you try to go a little bit conservatively, right? Because we are running their tax engine and you know, you want to make sure that doesn't go down and lots of money on the line.

So we try to have it um yeah like like four 4x roughly double 4x or you scale up with folks um we haven't we've never had like an issue with outages and stuff.

I think you know in 2025 it's not that hard to scale your your servers and these type of things but um you know yeah we do a lot of hype marketing but we're still kind of cons we're very conservative with the way we run our product just because it is you know a compliance product and so we'll make a lot of noise but when you actually comes down to the details you got to be pretty uptight you know about it tricky to use AI in a compliance context like what what kind of guardrails do you have to put on the the the LLMs that are doing work I can imagine a bunch of places where it's fine to have non-deterministic uh you know you know just data ingest or you know uh something where there's a human in the loop but but where have you where have you shied away from putting it in place?

Yeah, where we think about it is like there's still usually a human that does the last mile.

So I would give you a couple examples like one is when we we read for clients you you get physical mail from every state and country and you you'll get like a like a stack every month and it's like exhausting to have to read through all this physical mail.

So we get it, we scan it, and we have AI take a first pass reading it.

And most of the mail that AI, you know, it's like this is like if you can take physical mail that was going to somebody like the amount that I would I actually should find a personal service like I never want to get mail like at my house like I used a service years ago where you you send all your mail there, they scan it and then you get like an email inbox basically.

Um, obviously that's way better now with a with modern AI and I can imagine that be part of the workflow. Well, it's like you get a letter from California or somewhere and it's like, oh, sometimes you get a like I get a uh, you know, like like Joel of stress, right? Cortisol. That's the word I was looking for.

You read my mind. Yeah. You get a little cortisol from the IRS or whoever it is. Um, but you know, half the time it's like we got one last week for a client. It was, you know, breast pumps in California are tax exempt for the next 3 months and like Lucy probably got this letter. And so you got the stress for nothing.

But then sometimes it is maybe a letter, hey, there's a lean on your business. There's a big issue here.

And so anything that like requires human intervention, we have like a a mail team of six people that literally will go they open up the the they'll read they'll read the letters and they'll go and like actually handle the the last mile compliance.

Similarly on the tax research side, you know, how do you know that Juno I'm making this up Juno Alaska has a 4 cent soda can tax law? Often it gets passed in some city council notes. And so the way these legacy providers do it is they have hundreds of lawyers and and researchers just reading this stuff by hand.

And like we asked we hired some we have a couple lawyers on staff. We asked him like how would you figure out the taxability of I think it was like uh bagels in in California. And he's like going on the California like controlf bread bagel like it didn't work. Okay, baked goods finally worked.

And so this is like such an obvious thing where AI is able to go and do that first pass of like we put in bagel as a category. Okay, it finds all the baked goods. It makes a burst best first guess, but any single piece of tax research, it's not even that we have like one human lawyer look after.

We actually have to have two legal opinions on any single fact. And but it but it really makes them like 100x lawyers. So all the busy work they were wasting their time doing, they don't have to spend time.

And then the things that are really complex that requires going through, you know, huge adjudications and things like this, they can really put use their legal minds for.

Are you rating scrapers yet to scrape every local law that's passed and then at least pipeline that in to say, "Hey, maybe we still need a lawyer to look at this, but like we noticed that something changed. " Exact. That's 100% what we're doing, right?

So, yeah, we have a lot of like this we have a tax research group and yeah, that's like where a lot of the fun work is getting done. We're doing the real AI research that that that we need. This is this is important. How are you thinking of scaling headcount?

This feels like a business where obviously you you probably will grow the team quite a bit post B, but it feels like the kind of business that you should be like, you know, not like running in the red for 12 years in a row. Like you can probably start to make like real profit pretty hopefully. Yeah.

So, I mean, even like our raise, right, I kept 10% dilution.

I like I we could have raised like a bigger louder round but I think for us you know I've seen excess before and like excess hiring so you know we don't have so we started the year 25 folks we're up to 65 um but you know we but we but even that like everyone it always feels like tight you know there's no office manager nothing like that I was plunging the toys yesterday we'll post some footage someone got of me like they broke so I think yeah so there we go uh so I think no the point is though I think we always try to run these things as as as lean as possible.

I think still the hardest area to scale exponentially or like that tends to scale more linearly is is your go to market sales team. It's just um you know we we have some self-service stuff we're building out but I just think in general um AI has not replaced sales people.

It's still a very human endeavor and so that's probably we and that and obviously AI can eat steak play a round of golf. Not yet going to stay that way. We're getting there but um you should cap off the show with us. Let's let's let's uh check the timeline and see uh if there's any breaking news before the weekend.

Usually these companies wait until the second they see we go off the air. Yeah. And then they they I mean the the big news that we haven't covered yet is Nvidia did one of those zombie acquisitions. They bought this company in Fabrica and they paid $900 million.

They hired the CF the CEO and they licensed the startup technology. The CEO is Roshan Sankar, the guy behind Broadcom's AVGO, Tomahawk switches. So Nvidia is is now getting into the game. Some clear chat. Yes. Yes. Yes. Nvidia isn't just buying talent. They're locking down the AI data centers networking backbone.

Well, congrats Dylan Patel in Somebody should make it. Somebody should make an AI text blocker for your browser so that it detects like chat GBT generating. does feel like it's pretty detectable at this point, but for some reason, uh, it's not. Um, what else is is going on in here?

Uh, oh, uh, JP Morgan put out a report on Nvidia. Nvidia's investment in Intel. Nvidia put five billion in, marked up the US taxpayer, marked up Leopold Ashen Brener. Let's hear it. Let's hear it for Leopold and the US taxpayer. Uh, fantastic.

uh but uh JP Morgan's analysis said more value will be is seen acrewing to Nvidia in the Intel collaboration. There's still no commitment to Intel foundry. So Intel's still going around trying to get enough of those customers to justify their next big fab uh Nvidia is uh is is going to benefit a bunch from this.

So even though uh it's just interesting because uh there was one framing of this story which was like Nvidia is is making this investment as like an olive branch that they just have 60 billion sitting around. They're going to throw five billion into Intel.

Uh maybe just because they think the stock will go up or maybe that that they'll they'll they'll they'll like basically give five billion away as an olive branch to support this American chip manufacturer and then they'll they'll trade something and they'll be able to sell into China or something like that.

Well, now JB Morgan's here saying that uh yeah, Nvidia is actually going to get a lot of value out of this deal, which is kind of Yeah, they just right day after the announcement, they just nuke the chart and mark itself 5 million. Uh we got a post here from Nat Elias, who I saw in the chat earlier.

I don't know if they're still here. Says, "The reason you can so easily pull allnighters in college is because you're supposed to be having kids at that age. " 25,000 likes. That is a true true gigabanger. Giga banger. You don't see that very very often.

Uh uh Tyler, do you think that uh DeepMind solved Navier Stro Stokes? Did you see this post from Google DeepMind? We're announcing a major advance in the studies in the study of fluid dynamics with AI. Uh and then they put the water emoji in a joint paper with researchers from Brown University, NYU, and Stanford.

They they're they're they're announcing an announcement. Uh they haven't actually announced this. What do you think? Are you familiar with this problem? Can you explain the the significance of Navier Stokes? Because Sam here says, "Don't tell me they figured out Navier Stokes.

" Um, I don't know that I can do a really good explanation, but I did see in the comments it's like I'm pretty like 90% sure it's not it's like some random math problem. Okay. Well, not that interesting. So, well, still, let's let's hear it for random math problems. We want to solve all the math problems.

Come on, let's do it. Uh, we also have some news in the chat. Netflix is trying to buy Warner Brothers Discovery. Uh, we thought uh Ellison was bidding and Netflix might be trying to bid. This is from Puck News. Thank you to Reggie for uh sharing that news.

Uh hopefully you guys don't figure out that we'll read basically anything you put in the chat because you could put some crazy fake news in there if you if you wanted to. Uh we we talked about stuff. Paramount's up another almost 6% today. 40% over the last month. So we have one last ad readbasel. com.

Your bezel concier is available now to source you any watch on the planet. Seriously, any watch. You got a series B, you got to pick something up and get bezel. Yeah, you deserve a little treat. And we'll close with this question. Uh, how long is uh this guy messaged me to quote, "Hop on a quick call.

" How long are those supposed to typically last? Somebody asked you for a quick call. How long do you think that lasts? What's I mean, it's supposed to be 15, but it always ends up being 30. 15 to 30. Okay. Well, Chad GPT says, "Quick call can last anywhere from 5 minutes to 2 to four months.

Some quick calls have even been going on since the early 1900s with ancestors continuing the conversation to loop in key stakeholders and ensure crossgenerational synergies for the warm handoff. But this would be a lot easier to explain over voice mode. Do you have five minutes to hop on a call?

This is uh clearly photoshopped, but uh absolutely hilarious. Hilarious. I I wasn't I wasn't familiar. He has access to Is the chat GPT5 instant a real feature? I don't have that. Mine just says chat GPT5 usually when I go to chat auto who's getting access to in this seems like some new feature.

I don't know if it's all Photoshop but uh clearly a very clever Photoshop. You guys should hop on a quick call with the weekend with the weekend. Thank you for hanging out with us. Thank you for joining Sam. It's been an honor to have you. Uh and uh the progress is tremendous.

And uh everybody in the chat, everybody listening, have a fantastic weekend. Thank you for tuning