A Product Market Fit Show | Startup Podcast for Founders
00:19:05 · Arvind Jain, Pablo Srugo
[00:19:05] S0: Part of it is actually what is the goal that this founder has. Is the goal to actually build a successful company or is the goal to solve a problem, something that you're passionate about? For me, it was the latter. I was not trying to get a win. I was actually trying to solve that problem. That's why we made those decisions. We want to actually bring help to every single person. We want to make work more exciting. We want to make work less frustrating for every single person in the world. For a first-time founder, they're actually trying to – their goal is actually succeed as an entrepreneur. They want to actually learn that art. It's personal in that sense. For me, that was not what I was looking for. For me, my focus and to date now, all the decisions that we make even today are driven by that thing, which is we want to actually solve this particular problem for people. [00:20:02] S1: As you're doing this, what's the threshold where you do launch? What are you trying to get to where you feel like, okay, this product is ready to be sold?
00:20:02 · Pablo Srugo, Arvind Jain
[00:20:02] S1: As you're doing this, what's the threshold where you do launch? What are you trying to get to where you feel like, okay, this product is ready to be sold? [00:20:12] S0: Yeah, so as we said, for the first two years we didn't charge. We wanted to see engagement. We wanted to see how much people are using the product on a daily basis. When we saw that it crossed ten searches per day on average, that's when we knew that, well, look, now people are dependent on this product. They are actually getting value from it. That's when we felt we were ready to charge for this product. In our metrics, we're showing that both that they are actually finding things that they need and also they're actually using the product a lot. They're finding it useful to actually come and search here as opposed to using their old methods of finding. [00:20:54] S1: By the way, what was so hard about this product? Do you remember as you went through that development cycle parts that just were really hard to make to work properly or even places where you might've thought I wonder if we can actually get this to the vision that you have?
00:21:12 · Arvind Jain
[00:21:12] S0: Well, I mean, if you think about search, it is like magic. You can come in and ask any question and now we have answered that question for you, doesn't matter what question you came and asked. It's a difficult thing. You have to now think about this, just imagine the process of how are you going to actually solve this problem. First you have to actually connect Glean with all of your company information and data across hundreds of these systems that you have, all of them with clunky APIs or interfaces and data models. You have to bring information from all these different systems. You have to normalize it. What information is stale? What is out of date? When you answer questions for people, don't show them stuff that's no longer source of truth anymore. You have to think about how do you semantically understand data and knowledge because people are not going to be – you want people to be able to ask answer questions very freely in natural language. They can use terms that don't exist in the documents. There's a lot of technology that goes behind the scenes to actually build a product this. We've seen this. We've seen how complicated the Google Search stack was. Most of it actually did come from Google. We worked on search there. We understood the complexity of it. Then we had to think about the security aspects of it, which is that – imagine this, I go to Pepsi or some large company and tell them that, hey, look, we have this great search product. What we need is we need you to give all of your data to us. Then we'll actually make it searchable for you. They're going to wonder, okay, who are you? No brand company and you want all my access to all my data? That's an issue. If you start to think about the problem, although the problem is actually very meaningful, there are a lot of challenges.
00:23:08 · Pablo Srugo
[00:23:08] S1: How much did the advances in AI, like LLMs and all these sorts of things, did that play into what you were able to build at Glean? Was that a tailwind for you guys?
00:23:17 · Arvind Jain
[00:23:17] S0: Absolutely, and actually, it's always been part of Glean. Even on Day 1, our product, we're using LLMs. Now, of course, everybody talks about LLMs and we've all seen how powerful they are, but they were actually, initially, all of this research happened in Google and LLMs were actually, or these language models were actually built with one purpose, which is to make search better, to make Google Search better. When we started, we were actually able to use these BERT family of language models that Google had actually published in Open Domain. These were language models that were trained on all of the world's knowledge. They understood things like a user manual and a product guide are semantically equal and things. This model actually has all of that language understanding buried inside of it. We were able to use these models from Day 1 to build a really smart semantic search experience. We were doing vector embeddings and now everybody calls this vector search. There's no name for it like when we were using it initially. The journey has been a continuous one for us. It's not that four years later we realized that there's this massive new language model thing and we need to change our stack with it. We're progressively using it. These models just kept getting better and better. It allowed us to actually also make our product better. Initially, it looked more like the traditional Google where you come and ask questions and we'll surface that information back to you. The technology has been amazing enabler in terms of making a product better, but it's also amazing for us in terms of helping us push our business. This is where the transition happened from the vitamin to the painkiller. People saw that if only I had something like Chat GPT inside my company or all of my company knowledge, people realized that, whoa, no, this is something I absolutely have to have.
00:25:09 · Pablo Srugo, Arvind Jain
[00:25:09] S1: It makes sense. I think a lot of these products that are step changes you only realize later how much of a painkiller they are. I mean, Google itself, search itself is that. I'm sure before it, you ask people, hey, you could ask a million different questions a day, you’d be like, well, I don't have that many questions. Then you realize you do and you just constantly, everything that pops in your brain, right? Let me ask this final question. You take these two years. In 2021, you raised the $40 million series B. My question is, around that time frame, I mean, once you launch this product, how fast did revenue grow? How long did it take to get to $1 million in revenue or $5 million in revenue? Do you remember those days? [00:25:43] S0: We've been actually growing about 3x to 4x every year in terms of multiplying our revenues. We probably hit, right after we launched, because we did have a good set of design partners. There's a big pipeline that was already built for us. I think our first year of selling, we did $3 million, if I remember right. [00:26:06] S1: Perfect. Well, we'll stop it there. I appreciate you taking us through the journey. I'll just end with two questions that we always end with. The first one is, I mean, in your case specifically, you understood the problem deeply, but I'm still curious, when did you really know that you had true product market fit?