How to double profit margins for the most competitive industry in the world with software

Meet Ram Jayaraman. He is co-founder of Plate IQ. He’s helping restaurants actually understand how much money they’re spending on their food bills. This is an $860 billion market that has no pricing transparency until now. Plate IQ is used by 3% of restaurants in major cities today, and just getting started. 

A lot of companies claim to use AI to automate manual tasks, but Plate IQ has done it. They process now over 1M invoices per month. They took an unprofitable process and scaled it with pure software, and in the process are building something that gives pricing transparency that never existed before. 

You can watch the full video here: 

Here's the transcript of our conversation

Think of a business that's the hardest business in the world. Restaurants. It's only a three to 5% profit margin. What would happen if you could double that profit margin, just using software? That's exactly where Plate IQ plays. They're helping restaurants like Eleven Madison Park, all the way down to your neighborhood Jamba Juice, actually understand how much they're spending on food. It's an $860 billion market, every single year. And using software, they're able to let restaurants that you and I love stay in business, in the most competitive industry in the world. 

Garry: Ram, thank you so much for hanging out with me, man.

Ram: Yep, same here.

Garry: You've built something that is actually remaking every restaurant. And the crazy thing is, everyone eats.

Ram: Well, we started four years ago, really trying to help restaurants understand the money they are spending. And that information is kind of hidden in their invoices. On paper, and PDF, and so on, to kind of, really help them get all of the data that is there, the invoices.

Garry: This isn't the front of the house, this is the back of the house.

Ram: These are the un-sexy back of the restaurants, which you probably don't see, and a lot of them are a really tiny room with three tables in them, with bad wifi. There's a show in the front, then you walk in and see that. You see this nice food, you see all this decor, and all those things. And, how does it really get there? What really happens from a business standpoint as restaurant, how does the whole thing run?

Garry: And restaurants are incredibly hard to run, period. It's down to the penny, honestly.

Ram: Oh, yeah, the margins are incredibly thin. It's very common to have a small, single digit percentage as a margin, for restaurants.

Garry: And so, if you can move that margin, even a little bit, you're doubling their profit. You're taking restaurants that could go out of business, and turning them into, actually profitable.

Ram: Exactly. And with the advent of deliveries and all of the newer trends in the industry happening, the lot of the focus for these restaurants, we believe, should really be about how they run the business. And a lot of restaurants, historically, tend to be not the best at running as a business, and these are really creative folks who are getting into the business with a passion of it all. And so we really trying to help them squeeze out all the little margin they have from the back.

Garry: Well this sounds like something that software can help a businessperson who's pulling their hair out, they're trying to make a great experience, but also trying to make it a good business. And to date, before Plate IQ, they just didn't know whether they were getting ripped off or not.

Ram: Exactly, the data that we are getting from these invoices, it's really hidden because it's all on paper. And just ask yourself that, if you want to start a restaurant in San Francisco, and you wanna buy chicken at wholesale, where would you start? You can't just Google for a wholesale chicken, you might get a couple of hits. There's no to compare prices right? And that was one of the key things we saw earlier on, and in order to get there, you gotta bunker all this data which was hidden in paper. You really have to go on and like work a paper, to then slowly move them to a digital world and get all this data out of the paper.

How far along is Plate IQ?

Garry: What are some of the restaurants that you love the most, that use Plate IQ?

Ram: Well you probably may know The French Laundry, which is one of the most popular restaurants in the country. Tender Greens which is in the city, Mixt Greens which is also really good, and on the east coast, we again have a whole slew of things, everybody from the Halal carts guys, to like the Jamba Juice in the west coast. So it's a pretty wide spectrum, and I don't know if I have like, the absolute favorite at this point, I love them all equally so.

Garry: And so you're touching restaurants that are now nation-wide?

Ram: Yeah we are in all 50 states, at this point we have more than 10000 restaurants, and if you just look at the major cities, we're probably in the, I would say 2%, 3%, or even higher in some cities. So yeah, it's really getting there.

Garry: I mean 3% of any large business is pretty crazy, and the flip side to that is, you're just getting started actually.

Ram: Oh yeah, we're barely scratching the surface on this, and when we say restaurant, we don't just mean the restaurants that you see, this could also include cafes, and hospital cafeteria and so on. GAP for example uses us in their corporate cafeteria right? So there's a, anything that touches food and makes food, I think in some capacity, could benefit from Plate IQ.

Garry: Everyone eats. If you look at how much money is spent on GDP, food is one of the largest.

Ram: I think nation-wide as a country, we're spending like about $750 billion a year, which is pretty large. And restaurants themselves are spending somewhere in the tune of $250 billion a year in procuring all the products they're buying, to make the food for you. It's a pretty large market, and we're barely getting started.

Garry: What about the technical side? First off, how does it work for someone who signs up? And then what magic do you have to do on the back end?

Ram: The experience we wanted to give restaurants was something, to use your word magical, where like, they're getting all these like invoices in all the spend they're doing, which could be in paper, PDF, whatnot.

Whatever they use today, we don't wanna change behavior at all right? The one problem is that restaurants are not really that tech savvy pretty often, so we wanted to just have the least bit of intrusion of their daily life. So they just scan their paper, or like have a PDF, they can email the whole thing to us, and they kind forget about it at that point right? And that's really where the magic comes in where we're kinda using a combination of like, ML and humans to digitize all this data into a structured form. So that eventually we can answer questions like where is the cheapest chicken in San Francisco, right?

Garry: That's valuable.

How Plate IQ scaled the unscalable using software (yes, really, AI)

Ram: It's really valuable. It's a really hard problem to scale, and we didn't really know it, going in how hard it's gonna be, which is probably why we got into it. And when we started, obviously it was a lot of manual work. And it's kinda like one of those YC things that talks about, like do things that don't scale, and so we started something which doesn't scale at all--

Garry: Which is people.

Ram: Mostly people, they know very little technology, we spent more money on people than we were making at one point.

Garry: So this is either a Mechanical Turk, or what did you use?

Ram: We built our own version of Mechanical Turk, because the kind of precision and the data quantity that we had to provide for restaurants was pretty high, that it was a little bit hard to build a product on top of Turk itself. Now we kind of use a combination of things, we use Mechanical Turk, we have our own, in fact we have teams in three different countries which are a little bit more custom to us. And of course we use a lot of machine learning and AI to do this.

Garry: That's really sort of the big switch, you went from totally unscalable with people, to now more automated system using ML.

Ram: Exactly. And we're processing well over a million invoices a month.

Garry: Incredible.

Ram: Which is a lot of things to do. And again a lot of invoices are processed when we are doing it at the level we are doing where, picking out small details like, you bought like a bag of chicken which had 50 pounds in it, and you bought it for $30, and last week you bought it for $28 right? That level of precision I think, we're pretty proud of what we did.

Garry: It's really noteworthy for me, as a venture capitalist—we see a lot of companies that come in and say that they're going to use AI, but when you peel just a little bit back, you'll realize it's all people, and then you talk to the tech team and they actually have no possible way they could automate it the way you have. It's impressive to me, whenever we see a team that is capable of going from, unscalable to actually scalable, that's the whole magic right here. That's why a billion dollar business can exist that's pure software here.

Ram: We went from I would say somewhere call it like you said close to 0% automation, to now it's about 80% there. Of course we have plans to like, take it further as we go on.

How did Ram get into tech?

Garry: Ram how did you get into startups, how did you get into tech?

Ram: This kind of like, thanks my dad, where he had the foresight, of like, enrolling me in a computer class in my middle school, when I was like probably eight. Afterschool you had to go there, and if you want to really go to the computer lab to do any programming, you had to really choose between history and computers. And somehow he had the foresight to enroll me back in the days. And got a really early start there. And then, at the moment I laid my eyes on a computer I was like, oh my god, this thing is so amazing! And that's where I got, that's been the whole history of the tech itself. And as I like, started college and all these things, I really wanted to do something on my own, rather than just, it's great to work in technology, but I feel like you can have the largest impact if you actually did something by yourself, which is not working at a large company. That's why I was super keen on moving to the Valley, and once I got here, I joined as one of the founding teams at Automatic, which was part of YC before, where you made like a Fitbit for your car, that was my real first foray into like startup world, and then it was Plate IQ.

Garry: What advice do you have for people who are just starting out who are sort of the engineers like you and me, 22 years old, just graduated, or even, you know, 19 years old, maybe gonna skip college, like what do you even tell them if they want to be a co-founder and CTO of a company like Plate IQ someday?

Ram: While the technology part of it is obviously the attraction for all of people who are engineering oriented to get into tech and doing these things, really understanding that your tech is really there to help your customers, and looking at the customer side of the world, and the people side of like, who is using your product and that part, even if you're building something which is really high tech and it's not just like a bare product, I think truly understanding the fact that there are people on the other side, using technology, and they're serving a specific need, I think would really help you into building the right tech, and choosing the right technology stack. You wouldn't chase the hottest next stack that is coming on the market, you will actually pick one that's just safe and usable.

Garry: So if you're a technologist, don't focus on just the tech, think about how that tech can actually be put to good use by other people.

Ram: Exactly. And like, the next newest JavaScript framework is probably not something you need, you know, you can use something which seemingly is not cool, as long as it actually serves your purpose, which is ultimately keeping your customers happy and it's building something of value.

Garry: How did you end up focusing on restaurant tech?

Ram: Oh, food is a personal passion for me, like I watch every food show on Netflix, I'm like a hobby cook at home--

Garry: Right, me too!

Ram: I love garnishing my food and so on. And I kind of happened to stumble up this with, like Bhavuk who's my co-founder, where like, he was already working with restaurants and bringing something from email and all this, we were both at Blackberry before, we had an email background there. And we happened to talk about this problem, and we kind of realized that like, the problem of like, business expenses, and dealing with that, like how did it for consumers, was really acute for restaurants, and because of the scale of the different number suppliers they're dealing with, and all those different problems. So it kind of naturally occurred of like, a passion, deep technology need, because of the need for solving this hard problem, and creating something of value which can be used at scale, so kind of like a perfect combination of all three.

Garry: We're seeing this at, basically every type of business, if there's no pricing transparency, you don't know how much it costs, but the data is clearly available, it's clearly there, it's just never been structured in this way, we really love those types of businesses because, especially businesses that are as large as food, because everyone eats. I mean, it's housing, transportation, food. That's all, I mean what do human beings really need?

Ram: Yeah exactly. And you can see this time and again, like I said, like Flexport got its start at really by like, I think it was Import Genius I believe, that really unlocked the data from import and all these things. Zillow, which is creating real estate, is all about like showing you what the data is, and rather like MLS right? And so yeah, we believe that unlocking data, which is typically present in like, non-digital fashions, and like finding those niches, I think are a great place to find a great business.

Garry: There are just so many businesses out there, that you walk in, you talk to the business owner, you ask them, do you know how much you're paying? And they don't know.

Ram: We were at a bar trying to sell the product, and we were trying to show the bar owner the value of the analytics that we're providing, and we show them, see your vodka pricing has gone up by like 8% or something, he was like, that's impossible, I have like a contract for my vodka, the price can never go up. We had to pull up the actual invoice from which the price changed, he left the meeting, shot inside, and started like calling his supplier, calling the general manager, and started--

Garry: "Let me fix this right now."

Ram: Right now, because that's not possible! It was not just the fact that they didn't understand the change, but the fact that, he didn't even think it was possible to happen until we kind of showed them the value there. That is one of the first anecdotal examples where we realized that like, oh my god, we have something of value here. It's not just something they're thinking about.

Garry: Where do you wanna take Plate IQ from here? You started the company 2014 right?

Ram: Late 2014, 2015.

Garry: That's right. This is a great example of it takes years to actually build something. It was just you and your co-founder in 2014, walk me through sort of the, it's been five years then? Almost five years.

Ram: Just crossing five years, so yeah, we really started in earnest in November of 2014, so started 2015. We got into YC into 2015 summer. And when we were in YC, we probably had a dozen restaurants. The first year we grew, but not like some of the crazy, like 8000% growth or whatever. And so here we are four, almost five years into it, and like I mentioned we're at 10000 restaurants. And like, when I look ahead, like we talked about earlier, it's a huge market, we're barely getting started. And with the advent of like, the new trends like ghost kitchens, and like delivery apps and all those things, we believe there's gonna be an even bigger need for back office software like Plate IQ. We're really putting ourselves in the shoes of the restaurant as a customer, and not as a seller of products. And we can keep doing this for years to come. Any young person getting into tech, you should really think long term, we are five years into it, and we are, to repeat the word again, we're barely getting started. It takes a long time to build anything of real value.

Garry: This is the land year of the ten year overnight success.

Ram: Yes! Building a company takes time, and if you're really doing something of value, especially when you see all this news in the tech, in TechCrunch, oh this person did this, and that person did, oftentimes we don't realize how often it took to get there. Even the story recently about Honey and like if you look at the founders it took them a long time to get where Honey was, and then suddenly you see Honey in the news, you don't see all the things going on in the back end.

Garry: But it goes to show, if you make something that people really really want, then the success more or less becomes inevitable. And then the trick is actually still, how do you hire? How you do build that software? How do you go from unscalable to scalable? That's the struggle day to day, it's hard.

Ram: But ultimately, that's what's gonna get you the most success, you know? Because most startups tend to fail. And I think the number one reason I believe startups may fail is, probably because they aren't focusing on the right problem. Even within restaurants, we had so many people talk to us about, oh you could do this, you could do that, you know like--

Garry: You've heard it all in restaurant tech I'm sure at this point.

Ram: Oh yeah like, one of the most common questions we get is like, oh so you guys do inventory management? No we don't do inventory management. We work with other partners who do, because the problem that we're working in itself may seem a little bit of a niche problem from the outside, but it's like such a massive problem once you really dig into it right? And that's all there is to it, is like at it every single day and making like small improvements every single day to your product.

Garry: Awesome man. Well we're catching you halfway through to the ten year overnight success, and so we'll check back soon! But you're definitely on the right track, I'm so excited to be an investor in Plate IQ.

Ram: And thank you for all the support, it's great that you actually understand the value of how to do certain things like this. And not just chase like the next round, or the next overnight thing to be doing. And being able to afford to take the long term view I think it's a really great thing to have people like you onboard.

Garry: Thanks man! Thanks for hanging out!

Ram: Thank you so much.

Learn more about Plate IQ at