the biggest thing that you do…is something I learned from WeWork which is like, [you] have to have a very clear image of the world you want to create.
An interview with Ahmed Elsamadisi, a co-founder and the CEO at, a data platform for making better decisions. Narrator raised $13.6M from Initialized Capital, Flybridge Capital Partners, Y Combinator, and others.
Peter Zhegin:
Hello, and welcome, everybody, wherever you might be. My name is Peter Zhegin, and I'm hosting datafounders, a series of interviews with entrepreneurs and investors who work on data science startups. Today is the interview number eight, and I'm talking to Ahmed Elsamadisi, who is a co-founder and the CEO of Narrator. Narrator builds a managed data system that allows you to have all data in one table. It's a pleasure to have you here, Ahmed.
Barr Moses:
Thank you. Very excited to be here.
Peter Zhegin:
Amazing. So to kick off, can you start from a brief introduction? I know that you've done a lot of interesting things starting from developing autonomous agents, many years ago.
Barr Moses:
Yeah, so I started my career in robotics. So I did autonomous cars, human-robot interaction, and kind of what a lot of people would consider to be AI today. I moved my way into AI for missile defence, where I was writing algorithms for tracking discrimination for the US government. So this is like, my fun fact is, if a nuclear bomb hits the US, today's current tactical tracking algorithms, four of them, I wrote them. So hopefully, they worked out well. And then I made my way into the startup world, actually joined WeWork to build WeWork's data team. And I built the data team and data infrastructure. I'm kind of got experienced the whole WeWork explosion, and the good explosion where we were growing really badly, not the later explosions. I had that, I left and started Narrator. So that's kind of my overall kind of background from start to Narrator.
Skills useful for a startup CEO
Peter Zhegin:
Awesome. And if you look back on it, what were the most important things that you went through in this background, what have you leanred that equipped you for a role or the CEO of a data science startup, the most important, things maybe?
Ahmed Elsamadisi:
When you start and you focus on human robot interaction, when you're doing robotics, you're really thinking about how do you make the best decision accurately and reliably.

  • So a lot of my early career was in making intelligent decisions. So there was a lot of frameworks and important building blocks that I needed to know on how to make the best decision, you'll see how that's very important in Narrator.
  • The second most important thing is how do you build and grow a team? That's something I had to learn with WeWork, which is - I was an individual contributor for most of my life. And then I became, how to grow teams, what do you motivate, align, inspire your team to move forward? How do you protect your team from all the clashes are coming at you. So you to learn how to grow and manage a team.
  • And the third thing is, you have to understand how startups kind of work. One of the benefits of being a part of WeWork was thar I was in touch with the CFO, and I was in touch with the CRO. And I got to experience raising money. And that was a really fast growing startup. So there's a lot of things I got to see that are really rare. And I got to use that knowledge when I went to go raise money, when I actually was growing narrator. And those are the key kind of learnings that I had to do. In the growth, and part of being in WeWork was experiencing the problem.
We'll talk a lot about the problem that Narrative solves. But the core thing that I think makes a huge difference was that I got to live and breathe the problem. Like I didn't actually want to be a founder when I like, didn't think I was going to be an entrepreneur. But I lived and breathed the problem for so long. That when I went to go look for a company that was solving the problem that I was seeing, I couldn't find anything. And I was like, okay, well, I have enough friends, let's go start building this company. And let me try to solve it. And I think that's still like, probably out of all my experience, knowing the problem so intimately. And having built and used, like, the current alternatives to solving the core problem, like, gave me so much. You're gonna hear a lot more about it, but it's what keeps you going. And it's what keeps you informed. And I think for any founder, seeing how large or how big company works, what small company works, how the best case scenario works, and what the worst case scenario is, like keeps you level headed when you're making your own decisions about your company.
Identifying a problem, historical perspective on tech
Peter Zhegin:
And let's talk about the problem. You encountered that problem and WeWork properly right, try to find the solution, there were none on the market. So let's talk about the problem. What was it about and still is probably
Ahmed Elsamadisi:
So the problem is how we answer ad hoc questions. So when you think about the world of data, data is built as a pipeline. So you have your raw data, your transform layer in the middle, and then your BI layer. And this transformation, we often call it dimensional modelling, which is you build your fact in dimensions, and you can put in a tool to visualise it and slice and dice it.

However, companies are changing. And they're not actually asking, everyone thinks they want plots, but they actually want answers to their questions. So you have stakeholders asking a question. And then that question has to go through your two week process to dimensionally model it, which is terrible, because you have to maintain this new model forever now. And then you have to build a dashboard that has a bunch of stuff that doesn't really fully answer, the question is not very clear. But then the stakeholder has to kind of interpret it.

And this is really fuzzy and really slow. So the problem that we wanted to solve, which is why do we need to build all these dimensional models to answer questions? Why can't there be a world where you have a question, answer instantly available? Like you have to answer questions really fast.

But I think the best experience, I always try to explain when you asked me like, what was the moment that you were like, aha, this I want to build like, this is the Narrator idea. And I always talk about the moment when our chief revenue officer WeWork was asking me how many people that came to our site called us. And I was like, Okay, well, I'm gonna go back to my team, I will get back to you in a couple weeks. He's like, what do you mean, just like, everyone comes on site calls us within 10 minutes? How hard could this be? And I'm like, well, like... So here's the problem, like in data, you have these different tables, and then there's no foreign key that ties a session to a call. And then you there's different identifiers, a cookie versus a phone number, and bla bla bla bla bla, bla bla, and here's all the issues. And so that simple question is going to take two weeks of work, and then we can answer it. And if you ask me, you dare ask a follow up question, that's another two weeks of work. And like, I remember myself being so embarrassed trying to explain the reality of data, which I knew was the best case scenario, like that's every company in the world. But why isn't it so simple? Like, why isn't it just look at every session, look at it if person callstus in 10 minutes, give me the first one. And now I know the conversion rate. And that's kind of what brought this idea of like, I want to build Narrator. Because I wanted to solve that problem to make answering questions really easy. And that's how you have to do it. It's the way that people stakeholders ask questions.

And if you actually study the history, which is like, that's kind of how I learned, I go down the history chain, I noticed that the reason we built dimensional modelling was for dashboards, like we had like, we had a lot of data, we couldn't process it. So we were like the only way to process a lot of data because storage and compute power is really expensive, wants to build plots, and we had all these BI tools like Tableau. So we're like, great, how do we build a plot? Well plot the x and y axes? Why don't we build a table that's designed for x's and y's. And then we can put it in here and make it work. And in when this whole approach was designed and built the way that we currently do it. It was built in a world where people were like, just seeing the data was a very big deal. Like, you just wanted to get a dashboard. You had any idea what was going on your business. This wasn't built to answer questions in minutes, like you would if you talk to any database expert, they say two weeks to get a data model into production is a great timeline. But you can't wait two weeks to answer questions.

And that's the problem is that we built a system to build dashboards while people needed to answer questions. And that's what the fundamental difference you needed to create. So that's the problem is there was no system built to answer questions. And there's a reason for it, which is like a little technical reason, which I can dive into, maybe toward the end of like. I'll just say it - is the foreign key, and you can't really join without foreign keys, and foreign keys don't exist. It's like a fundamental low problem. But if you actually look above that, and you see the core entity, it's that there's no system built to answer ad hoc questions. And that's what Narrator did.
Startup vs. corporate pet project
Peter Zhegin:
We'll talk about the way how you solve it in a moment. But a very quick question about the transition from being an employee at We Work to entrepreneurship. Was there a particular moment when you realised that you wanted to build it outside. Because what I hear sometimes, from my good friend recently, he said - Okay, let me try to do something cool within a corporation. He's he works with one of the best tech corporations, and he thinks about his own pet project. So it's a brilliant idea, a big problem, why not to solve it internally. What do you think about that?
Ahmed Elsamadisi:
So that's really great, because so I think it's two parts, right? When I actually started building the first prototype of Narrator, I like the idea of activity stream in WeWork. When I realised that I wanted to actually make this a company, we work CEO said - build it as a product line. And they're like, listen, we'll give you resources, money teams built as a product line within WeWork. And there's two problems that I saw.

  1. One problem was, let's say it took me three, four or five years, like the best case scenario, to build a billion dollar company. At that point, it would be 1% of the WeWwork. I don't want to be 1% of a company, like that's disposable. It's not like, that's the first challenge when you work with it, build something within a big company.
  2. Second thing, you have a huge bias, you're not building it for your customer, you're building it for WeWork hoping that it can be used for your customers. It's a different thing. If I build something for WeWork, I am scopescopingd into the WeWork problem. When I build something for external customers, you are seeing a whole lot of different problems. A lot of our early customers were like from healthcare, and like online, offline, and brands and e-commerce, and like schools, and it was like all these different... you have to build a way to answer any question for any of these kind of companies. That's the stress that you need.
And last thing was really the community. I don't mean many, like in the founder communities that I talked and I talked a lot of founders, because we all need each other to stay sane. You got to be like, okay, what's going on with you this week? You don't really involve yourself in that world when you're in part of a company. I personally have not interacted with like, the entrepreneurs building side projects in Netflix. Like, it's really, really, you need a community, you need that level. And you need that commitment, right? You want to solve a problem, build something like you're going to be told that it's stupid for like a long time. So you need that feeling and drive to do it. And being in a big company where like, you're kind of too comfortable, just doesn't it doesn't get the same.

Peter Zhegin:
Yeah, thank you for providing this framework, I guess it's very interesting angle: first, how significant it is 1% of the WeWork or you build a separate company, then really for whom do you build it for a customer, or for somebody who sells that to a customer? And then the community. I guess it's a very useful framework to think about building an independent company. Because right now I see a lot of persuasion coming from corporations who say, stay with us, we have innovation units, we have corporate venture arms, whatever. I'm not saying that's bad. What I'm saying is that anyone needs to have a kind of a framework in his or her head to decide what setups to choose. So now we see how you made this transition. And if I'm correct, you've met your co founders that we work, right?
Ahmed Elsamadisi:
Yeah. So the most of it was from WeWork.
What Narrator is
Peter Zhegin:
Let's let's talk about the solution. Because my understanding, correct me if I'm wrong, that you really try to change a lot of things at the fundamental, very low layer. So they dealing with data, kind of a data schema layer alone, maybe even deeper. Tell about that, please.
Ahmed Elsamadisi:
Yeah, so the core solution, and we can talk about the core issue, let's come up with the core innovation and build our way bottoms up. I'm just gonna wait right, actually, to mine.

So the idea behind Narrator was, we wanted to build something that made a difference, not like made you feel like you're making a difference. We wanted to build something... we want to change the world in the way that instead of looking at dashboards, and like, kind of guessing, we wanted to give you a way to answer questions. And we wanted to judge ourselves based on your ability to make the right decision.

So to do that, we have to solve a lot of problems down the chain, right? The first thing is you need to data, what data is available, we needed to make sure that all your data was available in one place. And that's how we built Narrator on top of a warehouse. We need it to be simple. So you have your mental model of how you think about your business, instead of having the data change your mental model to match the data, what if the data matched your mental model. So that's how we build a single table simplistic can be, 11 columns. So it's not like no JSON blobs, no abstractions, 11 columns, all centred around a customer. So we broke it down to like three core entities, customer, time, and action. Last, why customer time and action - because that's how we talk about our business. We're looking to change customer action in time, and therefore that data represented that way. You can start asking questions the way you naturally do.

The third, that's very probably the hardest oneis, is now that you have this really long table. When I asked my questions, I want to relate the havior in nuanced ways, and create why tables that can actually now slice and dice in a plot it? Well, how do I do that? People really consider this activity stream model. If you ask somebody why they won't use it, they'll say it's not credible. Like it can only answer five questions. So how do you get this model to answer any question?

That nuance was probably two years of work, that's like the core of what took the hardest part is getting a single table to be able to assemble any possible table you need. And then you have to build the world around it. We talked about this earlier, too, which is like the idea that you need to build an ecosystem around it right? Because answering your question isn't as simple as getting a table. No, I need to get the table, I need to be able to aggregate and plot and visualise and slice and dice. And I need to be able to look at a single customer and see their entire history of everything they've ever done in our company. So I can be inspired to ask more questions, I need to go with that follow up questions and make changes and edits and add layers to my data and answer that instantly.

And then I need to be able to take that whole thing and use it, whether it's taking a list of people and putting it in a Google Sheet so that me and my team can like fix them, or build a table putting it in materialised view and putting in your dashboard, or even sending it back to the product with a web hook. So we can make like live, we have a five minute recommendation engine, you build the data set quickly moving to a product, or you just tell the story of what your brain is going through to make those decisions and all the questions you're asking in order.

That's why we created narratives is actionable analyses and story like format. So you have to build this whole world around this kind of core innovation of the simplicity and the innovation of being able to assemble any table and the ecosystem around it. And what when you combine this entire experience together, what you get is this freedom from data team, you are in you and your team setting up, you're going to be able to ask questions, you know, you're always using the right data, you know, because always on top of the raw source of truth, you know, there's no way that numbers are going to mismatch, because it's literally one table. So everyone's using the same data, you're free to ask questions that bridge all the systems. So all the hard questions about behaviour you can use now, because of the innovation of bridging sources. You know, this can live on past you, whether it's your analysis, or your dashboard or your product, you can take this data and use it in more places. And you can continue to build on top of it and explore and explore and explore. And that's kind of the thing that narrative has done.

And it shifts what we saw a lot with customers is that because of the way that dashboards are built, people are used to asking questions like hmm, how does gender affect conversion rate? How does like industry affect conversion rate? And we call this a first order question, so you're just like trying to slice it by people. But those industry and gender might be very early indicators, but they're really the kind of worst indicators you can have. Better indicators is like how many times somebody view that how it works page, like that tells me way more than than likely to they're more interested in looking at behaviour, it tells me a lot more about the customer than looking at their like gender. And but looking at behavioural questions is really hard. And then so nobody's really asking those kind of questions, because they take way longer to answer and the data team hates answering those questions.

So a Narrator by building this ad hoc layer, Narrator frees people to ask all the questions they want to ask, answer them so quickly, and get that natural progression. Like we will see us in meetings going, hh, what about this, we just check for the answer. And we go oh, and then it builds up a second follow up question. And a follow up question. And because these follow up questions are happening in minutes, not weeks, you actually are learning a lot more and kind of iterating. That's the experience of having an ad hoc layer. And that's kind of solution that Narrator has created.
Helping users to change their behaviour
Peter Zhegin:
And it's very interesting. So you build that, let's call it full stack or end to end product. And also you change a behaviour for user, you teach users to ask questions differently, right? So what I see here is maybe two things that you need to overcome, right, you need to firstly you need to sell this full stack solution and then you need to persuade people to change the behaviour. How do you go around that? Typically software kind of uses existing behaviours, right? It goes viral, etc, etc. So in your case, you kind of ask you users to change. How does it work?
Ahmed Elsamadisi:
Yeah, blood, sweat and tears. So, like, you're 100% right If we're like building a better something, it's a lot easier, right? Because you're just like, hey, you use Mixpanel were 'better mixpanel' you use Looker, we're 'better Looker'.

If you're building something different, you actually don't have to convince people to use your product, you have to convince people to invest in a concept of an ad hoc questioning layer. And you have to convince people to ask questions in a new way, which is they're like, I know my SQL, I know how to slice and dice data. Like, why would I ask a question in new way? So there's a lot of roadblocks down that avenue. And especially if you have to sell to like stakeholders, we don't even know what you're talking about.

So the thing that keeps us kind of able to do that is the fact that we understand the problem. Like do we know the problem, data people are facing, and people are used to data are very well aware of why these questions are really hard and why this is such an exhausting process. And they know their problem in and out. So we try to resonate with those people. And say, like, just give us your time of day to show you that, like what we are doing help solve this problem. And by knowing that problems, so intimately helps us really resonate with those people. And then the second problem those people, especially with Narrator, single table answer any question, they go - bullshit. So they're like I get this will solve a lot of our problems, but I don't believe that you will do it.

And we spend a lot of time in is demoing, really giving people like, I give a demo on any one of your viewers, any one of the viewers can go online and book it. And part of demo is asking any question in the world? And let's see if I can answer in life. If I can't answer live, then you would. I've been doing this now for two years, in demoing this exact way, and hasn't been subject. So challenge accepted for everyon. Giving people that experience to like see, like how any possible question could be answered using the narrator approach is huge. I'm giving it making it free to start, like, narrator when it first started, we were consultancy. And we'll talk a little about kind of that up, we were $10,000 a month, now it's free to start. And the price is starts at $200 a month. Like that gives you a lot more like giving people the ability to kind of get involved.

Then the kind of demoing game. And then really just trying to bridge the gap. Like I was on the other side, I was a data person having a data team. And all these companies were selling me like, oh, we're gonna solve all your problems. And it's all just bullshit marketing. And when people realise that Narrator is not just bullshit marketing, but it's an actual innovation, and something different people give us giving you time of day. And that's how you actually get to change the minds of people and really start changing that behaviour. And those are the people that people who experience it love it, they tell their friends and you kind of kind of to go that way. So it's really just building that one at a time powering through it, to try to get everyone to change. And oh, are the customers who experience it really are excited. So it keeps you going to like find the next one.
How to keep motivated after being rejected
Peter Zhegin:
Cool. Two important things is I've heard from you. So domain expertise and deep understanding of what you're actually doing. And then a lot of demos, a lot of talks with customers. I wanted all to just to make a step back. So there is another important bunch of people whom you need to persuade. Its investors, and they come typically early as and customers. So I assume with that idea you was rejected not once, right. So I'm curious about two things. First, what was your approach to pitch quite a complex and innovative thing? And second, what did keep you going? Was it some customer development? Was it some customers who gave you interesting answers? Because when you hear a lot of 'No', from investors, what did keep you going?
Ahmed Elsamadisi:
So I've heard a lot of noes in my life, I've been laughed out of a room, I remember one person said, if you solve this problem, then you must be your company would be worth a billion dollars, there's no way you solved it. So like leave and I was like, what? But we have, I have it like I can show you my theoretical solution. They're like Ha ha, funny leaves like this, you don't have anything. And it's really hard. When you're building something different. Like you're, you don't get traction. Like you're like, you have to spend two years innovating and building it to get it to actually work. And then you can actually start selling but it's still like, there's a lot of product and stuff to build.So I've been, I think daily, I was hearing somebody telling me to change my product change idea. Even my team was like, let's build something different. Um, so really bringing it all together is very important. So the two things that you will talk about right now is like you said, How do you kind of stay motivated? And how do you pitch investors? Um, and there is, I think it's kind of the same to me. Um, the way like, the way you inspire your team and the way you inspire your investor and the way you inspire your customer is all kind of the scene.

  1. And the first one is, like you said, right, it's having a deep understanding of the problem. Like every time when I tell the story of narrator I talk about the the world that exists today. Why it needs to change? Why people can't answer questions that they need. Like, at one point an executive come to us and tell us like, hey, we need data to help with bonuses. And we're like trying to answer data. And like, because of the way that data works and like the nature, and we had a really intense time crunch, like building five dimension modelling, cause issues and bugs and things happened It ended up going with wrong bonuses being given out to every single person. Like, if you're doing like value, but like based on your performance, and your ability to use data is like trying to maintain a huge system, it just doesn't work. So I know the impact. I've seen people get fired with bad data, I've seen companies like lose their mind, because they saw a plot and they're like, month over month, were down 50%. And like the whole company explodes, like. So having that problem be constantly, like lingering over me as the thing that I want to change. The thing that I know that when those things happen, the data team like myself, hated our lives, we would be yelled at things like this. And I knew the person I wanted to impact and I knew that the problem that I want to solve. So reminding everyone, whether it's your investor, customer, there's that that person that you're fighting for, is key.

  2. The second thing that kept us going was having another data person, so you will talk a little bit is a data people are so important to Narrator And the data team like really kept us kept us and we're only two people as me and Britney. And we would look at each other and be like constant reminding each other of 'remember, when we had to do this without narrator, this would have been a whole month. And like now it's here' Like, literally, we would talk about like if this doesn't exis, how would we go back to like, going back. It's like being in a Tesla and then having to go back ride a horse? It's like, what, are you talking about? I can't go do that. So painting that picture of that efficiency and that experience is huge for our customers and for our investors. I used to always demo for investors, I would say ask me any question.

    Before we even had a product, I'd say ask me any question. And I can show you how theoretically I can answer it with narrator really quickly. Or I would when we raised our series A I had like the investor sitting next to me as I was like, ask me life and let me show you the experience as a data person, as a data team, what I would have done what I could do now.

  3. And the last thing, which I think is the biggest thing that you do, and this is something I learned from WeWork which is like, we have to have a very clear image of the world you want to create. The first pitch deck ever made for Narrator was about two parts. It was about building a standardised layer where all data can be standardised in a single table. And it was called the universal data model. And now we call it activity stream. And it became about kind of allowing people imagining a library or catalogue of questions. And you can say because of standardisation, you can share and reuse. So you can imagine like what is my CAC? And you can run that entire analysis that tells a story. So someone can know, how do you think of a CAC. How do you answer it using the company's data? And why is that so important?

    The world today, if you think about like front end, all the software we use, there's so much open source in front end software. In terms of algorithms, there's so much open source algorithms, but answering questions with data -there's no open source analyses, because the input is so different, that you can't really build an analysis to be used.

    So there's people who have built the same analysis for 30 companies. And as they moved over, and today, I bet there's another hundred people answering the question of why you should model should I use. So by standardising data, you can actually share and build this library of analyses. And that was always the vision, I always tell them Imagine if you're a series A company, and on day one, you can set up Narrator in one day, and you have access to 1000 questions, questions you want to answer questions you didn't even know you wanted. And whenever you had a new question, you're able to insert a new question instantly by yourself. And if you thought through a built in analysis, you can share that with the rest of the world continually growing as a community together, getting better and better and better.

    That vision is the vision of Narrator And that vision actually just became possible. We just released templates two weeks ago, we have five templates now for like CAC, attribution and things. And we have planned to build another hundred. And now it's possible to actually like go into narrator oh, I want my CAC,k click run and get that on top of your data. So those three things really experiencing the problem, always reminding who you're fighting for, having a team that you can actually bounce ideas and stay motivated, and having that vision of the world that you want to create. You can paint that picture really, really well. And the right investor, the right customer and the right team and the right like partner or employee or person to join your team will value those three. And that's the people you want to be in your circle.
What to look at when the market is not here yet
Peter Zhegin:
Amazing. If you allow, I would love to maybe push you a bit here a bit. What you described are wonderful ways to keep someone going. But I would say they're quite internalised. So you have this vision, right you and you stick to this vision, your or you talk with your teammates, and you reinforce each other. What I'm curious, would it be helpful? Or maybe was it helpful for you to look at some external data? Let's say you talk to the customer, you did some maybe research or you get some numbers, that kind of evidences, or they really didn't matter much?
Ahmed Elsamadisi:
Yeah, so we, we avoided that for a bunch of reasons. So like, we're doing something different. So like, what is the total addressable market? Like, I don't know, a trillion dollars, it's all data. Like, I don't know. It's you talked about like, oh, like, the CDP are growing, and they're like a billion dollar industry and like, people are needing a solution. Like, that's if you're building something a little better. And you're saying like, like a lot of pitches, or like Looker got $3 billion by Google, we are building the next generation of Looker, we believe we can be that big also. That's not what Narrator is.

I used to tell my investors that and this is something that I highly don't recommend. But I would tell my investors that they're not gonna make money off Narrator. I told every investor, I said, Narrator is not for you. Narrator is for your kids. Like, the thing that I want to change is how every single person the world makes a decision. It's going to be a long journey. And I'm stubborn, and I'm going to go fight for that journey. So chances are, you're going to die before we reach our goal, to build something so impactful for everyone. And that's why you have to be willing to come on the journey with us, for your kids for the next generation.

I benefit from Narrator because I'd like use it, and love it. But the real impact is in in like, five, six years and 10-20 years, when no one is thinking about data modelling to answer questions.

Like when that doesn't exist in their head. It's like, there's like, when you watch videos online, you don't think about like the JavaScript plugin that you have to instal every time you try to watch the video, like that's been solved. That's the world that I want to create. And I find that it's really hard. It's already hard enough to explain what you're trying to do. Because the person doesn't understand the problem, like your investors, like I don't get it. Doesn't Looker help you find insights? You're like, no, they're a dashboarding tool. They're like, well, why don't you use Watson, IBM Watson promises to solve and you're like, No, it's just kind of marketing bullshit. And you're like, so you can't really use it.

It's really weird because a data company and I'm saying you can't really use data, when you're doing something new and different. The error and the bounds are really high. So you really got to just trust people that feel your passion, understand what you're trying to do, and trust that you're the right person to go for that journey. And I think that's the people who followed us, that's the people who joined us.
From a consultancy to a product company
Peter Zhegin:
I guess that's, that's an interesting approach for really something new. And also interesting to me to understand to understand how someone can do customer development with a very new product. When you build something clearly you, you are the one who tells customers what they need. Because they might really don't know that they need it. So how did you find the balance between on the one hand, you need to really listen to these guys and understand what they really want? And at the same time, you had as you have this vision, so how do you balance that out? In the early customer development process?
Ahmed Elsamadisi:
So Narrator has done a couple of things that are very unique. Narrator didn't have any product people, no product managers, no product people, and no designers. And we operated initially as a consultancy, so how do you build a product that solves this core problem?

  1. First, you better tested it yourself. So we operated as a consultancy, where I was supporting eight companies. And the rule was they can ask me unlimited questions over Slack. I don't have you know, it's like to have eight companies bombarding you with questions. You'll be able to answer a questions like instantly, you need to make sure your product is handles all the little edge cases, all the problems. We used to call it death by 1000 cuts. We don't want that we don't want a bunch of features that are just like, extra. We want to make sure that those there's zero cuts as you're going through their core process. So we iterated for a whole year we were answering questions and really doing trying to answer questions in less than 10 minutes. That's kind of how we iterate through a product. And so we like liveed it, breathe it. And that's kind of when we tell customers what the experience is like we're speaking out of our own experience. Like, okay, this is the questions you're getting.

  2. Second thing we did was when we started building the product, the way that we control the kind of what gets built, and what how do we iterate is all by data. So like, data is one doing customer support. So if you submit a ticket today, me or Brintey could be the ones who pick up a ticket and talk to you about it, we understand your problem. We're using a product every single day. And we're saying here are things that we need to do to make the experience a bit better. So our customers do submit a lot of recommendations and say, like, oh, this is really kind of annoying. And we take that, because it's all filtered through data, we understand so deeply what the customer is experiencing. We know what they're trying to do. We know what the right answer that we're trying to get them to do is, and we know how that what they know. We can slowly built things in a way to guide them and guide direction.

    So by being such an ingrained core customer, and being so intimate with all the customers, and having battle tested in the worst conditions, we're actually able to make decisions. And I'll move the product forward.

    Every day a person who uses our product will tell you that all the little things are thought about, like down to the naming of how we auto name columns, like down to how you swap and add and build. The fact that you can have multiple aggregations on the same parent data set are all things that if you work in data, you are all the frustrations that we have now solved by iterating on it, the fact that you sometimes want to just right click and say give me everything this customer did and see their story without leaving your page. If you want a Jupiter notebook like but high quality presentations, all these little things that you will experience when you go use the product is because data people have been ingrained in it.

    And we ensured that we weren't building a product without direction. We weren't like, tell me what you want, we'll build it. We were like, we're gonna build a product that solves the problem. And the problem, like you remember, wasn't visualisation wasn't data, it was better decisions. So like, if, for any reason, a customer made the wrong decision, because they misread the plot, that's Narratives fault. Plot should not be able to be misread. So we need to make sure that that's solved.

    And that's kind of how you build stuff for a customer. So it ends up not being a bunch of like extraneous features, like Narrator doesn't have a trillion shiny little gadgets that do really nothing. We have a core tool to be able to ask and answer questions, and we build that super well. Data people will know that difference.
Peter Zhegin:
It's a very striking example. Not very frequent, unfortunately, when you really transitioned successfully from a consultancy to a product company, right? What I would be curious about is do you think that this approach will work for any type of product, or that was a particular type of product? Let's call it data science product, right? That allowed you to transition from a consultancy, to a product company? Do you think it would work let's say if you are building a customer service automation tool, or something like that? What was your thoughts on that?
Ahmed Elsamadisi:
Yeah, so the transition between consulting to product is really depends on how you were consulting. So let's say like, for example, for us, we had a bunch of rules while we were consulting. Like, if you're consulting, you're trying to solve a problem for our customer. But we had rules that we set very core, you could not use anything outside of data model and you could not use anything outside of Narrator. Okay, so we were really not consulting, we were really using our internal product for the customer. It is very different. It's very easy when you're building like an automation tool to have a bunch of edge cases, or some custom code, build some custom things, change it have these little snippets you built in for the customer. And what ends up happening is you have this thing that works for some customers, but it's like now clunky, it has a lot of extra things in it. And now it's when you go to sell a product and you're like, I can't use this. Like, it was so hard for us when we were doing like a lot of our early like some of our consulting were like e-commerce and marketing. And people were like, oh, why don't just make it more marketing toward? And we had rules that was like no, we cannot do that. The goal of Narrator was to answer any question, we can't build anything custom for any company, for any industry, which is really rare.

When you think about all consulting that is usually vertical specific, industry specific and we're like no, no, no, no. The ability to answer any question. So that's kind of the core thinking that you're starting. And I tell this to any company that started as a consulting, starting as a consulting is really good to get customer data, it's really good to experience the problem, it's really good to make battle-test your product and make some revenue.

But you're not a consultant, you're using consulting to learn - big difference. So if you're using consulting to learn, don't forget, you have a product and your product has rules and guidelines that you need to maintain. If we use if you're just a consultant, then you can do whatever you want. And that's really, really bad. Because then you're gonna go, when you go to product, it's not going to fit, you're gonna be like, oh, wait, this doesn't actually work. Because your customers not gonna be able to do all the hacks that you have to do to do it. So it needs to be built in, it needs to be seamless. And it's really, really tricky. So I would say, anyone who's going into consulting, make sure you're using consulting to help inform your product, not being a consultant.
How the data science community could become better
Peter Zhegin:
Thank you for that it was a very great turn to discuss a consultancy, because lots of data scientists really do consulting, and then try to switch to product. And just to be cautious of your time, maybe, the last bit. We've discussed quite a lot today. Is there any theme, or topic, or question that we didn't cover, but you believe deserves to be covered?
Ahmed Elsamadisi:
Yeah, I think the last thing that I think is super important is the concept of a community. Like, when you are building a company, you need to get your first people to give you a shot. And you're not competing against, like, in data, you're not competing against like a random archaic industry that uses paper. You're competing against Google, Amazon, and a lot of leading companies, like throwing billions of dollars at the problem, and having a huge engineering teams.

When you're in your tiny little startup, like us seven people trying to go against these huge behemoth telling you what to do, the only thing you actually can rely on is like we said, let's what in tears. But you hope that the community is willing to give you a shot. And that's really, really hard, right? Because I know that I know, being a data person, you're in inundated with a bunch of bullshit products, trying to like, get take your money. And it's really, really hard to be something different. And stick out and try to get people to believe you and listen to you.

It's it's a tricky business, right? There's 1000 tools to help you plot data, like literally 1000 tools. And everyone's like, oh, we'll help you find insights. And you're like no use visualise data, you don't help you find insights. And to try to get people to like, say like, we help you answer ad hoc questions, the data community needs to be a little bit tighter, the front end community is so much better at that. So when we talk about like, software engineers, there are so much more willing to try new tools, there's so much more willing to be like, oh, this is a new company, they look at the founder, this person understands the problem, let's try their product. We use like a graph database from a company the experience that we use, like our logging from Honeycomb, like we have a lot of tools that we've added to our system, because we found people that like the community of software engineers, were like, oh, this person knows what's up. Like, let's try their product. And they just try it.

Data is not like that. It is very anti trying, there are a lot more slow moving. And it because of the way that the data community is, we've made it really hard for companies to innovate, as a company innovating, like coming into the data is really tricky. Because everyone's like I already paid a million dollars for Looker and Tableau and I have all these tools. I don't want to change and you're like, but you're like, not happy. And you're and they're like, yeah, I know. But like, there's no other way to do it. I'm like, give me a chance. And they're like, I've heard that a million times, I don't want to talk to you. And for us, it's like getting that first demo to like trick change someone's mind is so important right? Before a demo, it's all words that you kind of like are snapping.

So that's the change I would love to see. And I always talk about this in anywhere I go, because there are like minded data people who want to build a data community where we're able to support each other and allow something to really help. But there is a lot of noise in data. And it makes it really hard to give people a shot. I'm sure there's some brilliant people out there trying to figure out how do you find the right signal between all this noise and enable the data community and people are willing to give each other a chanc. I think that would be like huge thing for all of us integrating community trying to fight this battle.
Peter Zhegin:
Yeah, and I guess it will be better over time as the community mature. And then I'm sure there are people who try to innovate. So people like you also will make the community more responsive. The more you innovate, I guess, the more response you get at certain point. So I'm sure we will hear a lot of interesting news from you and from the Narrator. Thank you very much Ahmed for talking with us today. Really appreciate that.
Ahmed Elsamadisi:
Thank you so much. I really appreciate your time as well. It was really fun.