How to Ignore Most Startup Advice and Build a Decent Software Business
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
Title |
| |
Title of Series | ||
Number of Parts | 132 | |
Author | ||
License | CC Attribution - NonCommercial - ShareAlike 3.0 Unported: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal and non-commercial purpose as long as the work is attributed to the author in the manner specified by the author or licensor and the work or content is shared also in adapted form only under the conditions of this | |
Identifiers | 10.5446/44943 (DOI) | |
Publisher | ||
Release Date | ||
Language |
Content Metadata
Subject Area | ||
Genre | ||
Abstract |
|
EuroPython 201862 / 132
2
3
7
8
10
14
15
19
22
27
29
30
31
34
35
41
44
54
55
56
58
59
61
66
74
77
78
80
81
85
87
91
93
96
98
103
104
105
109
110
111
113
115
116
118
120
121
122
123
125
127
128
129
130
131
132
00:00
SoftwareBuildingFormal languageNatural numberOpen setDigital signalOpen sourceProduct (business)Information technology consultingScale (map)Extension (kinesiology)Endliche ModelltheorieSoftwareExtension (kinesiology)CASE <Informatik>Physical systemLibrary (computing)Process (computing)Information technology consultingPublic domainSoftware developerService (economics)Loop (music)Term (mathematics)Open sourceData storage deviceCore dumpRoundness (object)TunisPosition operatorEndliche ModelltheorieClient (computing)CollaborationismDigital rights managementBitProjective planeCodeDecision theoryHypermediaPoint cloudNatural languageProduct (business)BootingWave packetExecution unitSpacetimeComputer animation
03:41
Computer networkScale (map)Operations researchEnterprise architectureFacebookOpen sourceMedianInsertion lossScaling (geometry)FacebookMultiplication signBusiness modelBuildingLocal ringService (economics)Product (business)Different (Kate Ryan album)Tournament (medieval)MereologySoftwareEnterprise architectureCASE <Informatik>Touch typingFraction (mathematics)Real numberMedianGame controllerLevel (video gaming)Type theoryInternet service providerOperator (mathematics)Process (computing)Factory (trading post)AreaSound effectAverageCellular automatonGame theoryRight angleComputer animation
11:15
BuildingNetwork topologyVulnerability (computing)Level (video gaming)Different (Kate Ryan album)Software developerBus (computing)Arithmetic meanPredictabilityNetwork topologyProfil (magazine)Branch (computer science)AuthorizationBuildingQuicksortOpen sourceDegree (graph theory)CASE <Informatik>WebsiteInsertion lossCore dumpCollaborationismType theoryDatabaseCodeLattice (order)DebuggerSoftware testingExpert systemSoftware1 (number)Projective planeDecision theoryComputer scienceTemplate (C++)Functional (mathematics)Stack (abstract data type)AreaFlow separationGoodness of fitShape (magazine)WritingArmMultiplication signEndliche ModelltheorieIntegrated development environmentEmailDirection (geometry)Solid geometryUniversal product codeFront and back endsHypermediaLine (geometry)Equaliser (mathematics)CodecProcess (computing)Self-organizationComputer animation
19:46
Software testingDecision theoryProduct (business)Inverse elementRandom numberNumberException handlingDecision theoryInverse elementMultiplication signMereologyEndliche ModelltheorieTournament (medieval)Line (geometry)Product (business)NumberObservational studyGoodness of fitMathematical analysisGodStatisticsCodeTwitterSign (mathematics)Computer-assisted translationAnalytic setDifferent (Kate Ryan album)Performance appraisalPosition operatorBusiness modelSoftware testingRight angleObject (grammar)Computer animation
25:54
Exception handlingDecision theoryPerfect groupMultiplication signComputer animation
26:44
Focus (optics)Product (business)SoftwareEstimationSoftware testingTime travelTournament (medieval)MereologyBuildingObject (grammar)FacebookInternetworkingTraffic reportingLine (geometry)Right angleInheritance (object-oriented programming)Product (business)Endliche ModelltheorieCartesian coordinate systemVector potentialInfinityGraph (mathematics)Category of beingAbstractionDifferent (Kate Ryan album)Interface (computing)Point (geometry)Software developerVirtual machineCASE <Informatik>Generic programmingBitGodComputer animation
32:37
TwitterMusical ensembleLevel of measurementDrill commandsBitMultiplication signComputer animationJSON
33:33
Projective planeBusiness modelOverhead (computing)Product (business)QuicksortBitVideo gameScaling (geometry)FamilyComplete metric spaceProcess (computing)Meeting/Interview
35:43
TwitterEntire functionLine (geometry)Programming paradigmBit
36:27
Inverter (logic gate)Latent heatType theoryMeeting/Interview
37:18
TwitterDifferent (Kate Ryan album)Deutscher FilmpreisSocial classMereologyFamilyElectric generatorSolid geometryVirtual machineInheritance (object-oriented programming)Computer animation
38:33
TwitterSoftware developerProcess (computing)BitSpacetimeIntegrated development environmentProduct (business)AreaDifferent (Kate Ryan album)Universe (mathematics)Meeting/Interview
39:55
TwitterLaptopBitSoftware developer
40:27
Goodness of fitMathematicsInsertion lossEndliche ModelltheorieDataflowFunctional (mathematics)Natural numberSolid geometry
41:40
Twitter
Transcript: English(auto-generated)
00:03
Yeah, yeah, thank you. It's really nice to be here. I love this venue with those like turning auditoriums. It's really cool. So Yeah, I'm gonna tell you something about building a software business and Maybe just yeah to start with some background about myself. You might know me from spacey
00:22
Which is an open source library for natural language processing in Python So if you have lots of text you want to analyze something in text You want to train a system to do something really custom on your text, then you might be using spacey and So in 2016, we founded our first company together with my co-founder Matt who also just gave a talk earlier over there and
00:44
We really decided to bootstrap the company be self-funded and we did what we recall. We raised the client round So we did some consulting to start the company to get some money in and then eventually in December last year we launched our first commercial tool prodigy and
01:01
Yeah, it's a it's an annotation tool for data scientists so basically it helps you label your data try out new approaches quickly and Really? Yeah iterate quicker on your data on your code with a model in the loop and that's you know Downloadable tool we're selling and we're offering and that puts us here So that's that's where we are right now and prodigy since the launch has been going very well
01:23
We have a few thousand users now from fortune 500 companies freelancers data scientists researchers and In fact, it's been going so well that we're currently working on an extension product, which will have a small service layer To really help people scale up their projects get more out of their data and manage annotators
01:43
So that's what we're currently doing and in a long term We're also planning a data store where you can then get pre-trained models for custom domains Say you're doing finance in Chinese you're doing social media in Spanish You'll be able to download a model Tune it tweak it add some more examples and really build a system. That's super super custom to your own use case and
02:05
Yeah, when we started a company we really Yeah, we made the conscious conscious decision to stay independent To be self-funded and that's worked very well. We still own 100% of the company We were able, you know, it gives us a lot of flexibility
02:21
We're much better off in our positions than we were if we just had jobs So it was a since this was such a conscious decision It also meant that we basically had to ignore a lot of the typical things people tell you about how to start up a company so For example, people often are very confused or surprised that it's basically it's possible to make money and it's possible to be profitable or
02:43
We're still a very small team. It's mostly me and my co-founder. We're working with one developer on the prodigy extension we have some cool collaborations coming up, but really the core of it is a very very small team and It also, you know, we can people are sometimes surprised that you know We really put prodigy on an online shop for sale and you can buy it and it's yours
03:00
You download it like a bit like pre cloud Adobe and you have it and it's just a simple tool that people like that You can use it doesn't do anything sneaky in it doesn't even go online it's just a thing and that's kind of the core of our company and We think that actually, you know a lot of so this talk, you know, I'm not gonna tell you how to run a business
03:21
I'm not gonna tell you to follow your dreams and do it exactly the way we did but I do want to share some some of the lessons we learned some of the insights we got and Just yeah some of the stuff we picked up along the way including or actually most importantly some of the misconceptions that maybe some people have about running a business or actually even selling software and I
03:43
Feel like it starts with this one, which to me is actually a very important one that we often hear It's basically you need to if you want to run a business you need to run it a lot So essentially you need to spend more money than you make or you even need to take money. That's not yours and then take yeah spend it spend it spend it until you can eventually make a profit and
04:04
The thing is this there's a reason why? This is a popular concept and why this works for some company. For example, I'm not saying you shouldn't be doing that There are plenty of business models where this actually really makes a lot of sense So peer just well here's some reasons and some company logos that are probably very recognizable
04:22
So we have one thing. Okay, if you need the network effect, so Facebook, there's not really a Place for a really small social network with 100 happy customers like you need. Yeah Facebook works because everyone's on Facebook Similarly Dropbox or slack only work if everyone's using it and if you really have to viral adoption of the product
04:42
So that means okay You first need to spend before you can make money or probably Facebook. Nobody will ever pay for a Facebook account and that's fine that's just the type of business model similarly you if you need to scale you really need to Invest a lot up front in your infrastructure like a service like Twilio needs to serve their API's
05:01
Amazon Facebook a lot of these actually Dropbox a lot of these are very resource intensive and you need to set that up Once you get your users Your users need to be able to upload at one terabyte of data and you need to be able to handle that so that that costs money and that's an upfront cost and actually yeah, I included this startup Casper in there, which is the company that sells mattresses and
05:22
They also they took some funding They built their mattresses and then in the first I think in the first month of operations or sales They actually got almost half of that back because people bought these mattresses, but they still they needed some money to Engineer those and to produce them and Another one is well the competing on price super aggressively like uber and amazon are the prime examples of that
05:44
You need money and you need to subsidize the difference Otherwise, yeah, nobody's going to use your service and it really just works if you can Really compete and that needs money or if you as another alternative if you're going for enterprise sales
06:00
Yes, you need to invest a lot because you can before you can really get those really large enterprise deals like Salesforce or even Zendesk that's what is you know, they're really going for and here, you know It needs it needs the upfront investment You have to accept that maybe in the beginning you're not necessarily going to make a loss Make a profit and that you're mostly going to be making a loss until eventually like Amazon
06:21
you'll be very successful and making tons of money and your CEO is going to be the richest person in the world and but the thing is, you know the bigger just a bigger company in general isn't necessarily better because software is actually something that's actually kind of interesting here because I'm like a lot of other Traditional manufacturing products. It actually gets more expensive to build at scale not less. So
06:44
You know if you're building car parts cool You can start building a few but that's actually much more expensive than you buy wholesale. You have your own manufacturing pipeline build a huge in just like build a huge company build a huge like Factory where you assemble those and then you'll be able to compete on price if you
07:02
If you just build a few Artisanal car parts, you're probably going to be out competed by someone else that's not the case for software actually software usually gets much more expensive once you scale up and make it be and Another thing we've seen in the companies. I showed you earlier Most of these are actually in businesses where the there's only one winner
07:22
There's only one Amazon Amazon can only work if there's one Amazon Uber Yeah, they still have competition. There's local taxi companies There's lyft but the endgame is certainly we want to be the only uber and we want to be the only private modern taxi provider in any large city, so Those are all these really winner-takes-all industries
07:41
But actually most businesses are not like that in most industries. You can perfectly co exist and man There's not necessarily one winner and actually being in this winner-takes-all market really sucks because yeah, you can be super successful You can make lots of money. You can do everything right? and if someone else is like one tiny fraction better than you you're gonna lose and you're gonna lose everything because you're not gonna win and
08:04
Often even maybe for reasons that are absolutely beyond your control Someone's gonna win everyone else is gonna lose. So this is actually this is not not even that much fun And so you probably you probably all see. Yeah, some of you might know this this comic or this cartoon It's basically illustrating the concept of survivorship bias
08:22
The person on stage with lots of money says oh never stop buying lottery tickets No matter what anyone tells you I failed again and again But I never gave up I took extra jobs and put the money into tickets and here I am proof that if you put in time It pays off Well, yeah It's of course
08:42
It's true But you're obviously not seeing all the other people who did exactly the same thing and failed and it also shows that well what you see as best practices That you've learned that, you know, we may be adopting from companies and businesses in those winner-takes-all market all things we kind of have to take with a grain of salt because
09:00
Do we really know what caused them to win? We don't really it's just like, you know, we just have one winner. So this is You know, this is something we have to consider and but it also means you know If someone somehow as humans or a lot of people are very much drawn to these Tournaments and winner-takes-all markets and it's kind of ironic because if you think about the tech industry
09:21
It's all like, you know it's often we're all kind of nerds who would have maybe in back in school have rolled their eyes at like the kid who wanted to become a pro football player because it's such a ridiculous goal and you know, of course, you're not gonna make it and He probably you're gonna get injured anyways, and you can't do it and it's such a yeah You should rather go into business and do something real and at the same time here
09:42
We are in tech and a lot of people, you know, then go aggressively for winner-takes-all markets and Yeah, the tournaments where you have to win or not The good news is if a lot of people are doing that that actually leaves many other high value opportunities and things untouched and You know, I mean actually high value like you have to consider if you if you in if you're taking part in a tournament
10:04
You need to you. Yeah, you can't deliver on a few million you deliver You have to deliver on the billion dollar promise and that means that you know There's so many sensible nice businesses you can pursue that are absolutely You know sane Very profitable, but that nobody else touches because they're just not part of the tournament. So
10:24
Yeah, this this opens up all kinds of opportunities in software and it also means that you know if in general you Go and optimize for the median outcome, which means and not the mean outcome So you don't look at like what's theoretically? the most The best possible outcome which is you're going to be the most successful company in the world and you're going to be the richest person
10:45
In the world and instead say okay, what's kind of the most reasonable? high value opportunity Where I could make the most money out of on average then the advantage is that actually you know you're gonna You know, you you will definitely end up with something at the end
11:02
Even if you don't succeed as wildly as maybe you dream, but of course it still means If you you know If you're going for these opportunities you somebody of course needs to do this and you often can't do it all by yourself Which leads me to the next misconception, which is you need to hire lots of people you start a company
11:21
someone needs to build it you just need lots of people and And the thing is as I said earlier, you know, people are often surprised that we still we have a very small team we're doing lots of stuff, but the team itself is very small and we once had a Customer and send us an email and inquire whether our company would pass the bus test because that's an important requirement
11:43
Otherwise, they would have to reconsider doing business with us And in case you haven't heard of the bus test is this kind of thought experiment of what would happen if four people in? Your company were run over by a bus tomorrow. Would you still be functioning? And of course in our case sounds is no We'd be like minus
12:00
Something like no, that's of course It's not gonna happen and also but he chose it's a big the customer ended up buying prodigy anyways And apparently, you know, we must have still been convincing enough even without the bus test but it still shows like the perceptions people have and also how people view a stable company and what really makes a product good and it's interesting because if you if you haven't thought of this before
12:23
You'd probably be very surprised how I'm small The teams are that develop the software you use every day from open source to close source to anything really Often the core developers and the core authors are one or two people Maybe you have up to five people but it's just because Google releases some new amazing technology
12:42
It doesn't mean that all of Google worked on that sure It was made possible by Google being a very large organization, but the teams themselves are very small Maybe you have five people who really did all the core work on a thing. So This also really shows that you know the difference there's authorship and there's just in general the company and authorship
13:00
Really needs a very very small team because it's like I mean try try co-writing a novel with 20 people That's a really interesting art project. But like that's not gonna be very efficient Just similarly try writing a function in Python with five people. How do you do that? Does everyone write a line? Do you like, you know talk sit together in a meeting talk about it? Oh, let's put like how are we calling this variable? Oh, no, I would call it something else
13:23
No, of course one person writes it other people can test it In essence, you need a very small Team making the decisions and really writing a thing and also as we all know, it's not that difficult Fundamentally to write Software it's you know, you can write a function and then the refinement process. That's you know, what takes maybe takes more time and
13:47
Similarly a lot of people argue. Oh well, but yeah sure Okay, if you know what you want to build small teams all make sense But you need to know what you want to build first, so you need to build lots of things So yeah, you need to if you have you can have ten teams and you will build something and then maybe you find the one
14:00
thing that's good and But essentially, you know if you if you are building the right stuff that obviously matters a lot more than like building lots So yeah, if you know what you want to build if you really go for that, that's much more valuable than Scaling it up before you even have the first thing that works. But now the question is well
14:22
How does that how does that work in practice? How do people assemble their teams and here I'm these illustrations show two very common Methods that people use to build their software team. So the one more traditional one is a specialist So you have several people everyone has a deeper skill You have someone who does front end you have someone who does back end you have someone who does DevOps
14:41
And then you have and and those people all work together And form the team so you have all the things you need covering and this works great but of course, there's some weaknesses which Is that a the people don't really understand? Necessarily understand what the other person is working on So you actually have to invest a lot in making the collaboration work
15:00
You have to have lots of meetings. Everyone has to always share what they're working on so you can move forward Similarly, if you have one person who does DevOps and that person is sick. Well, then you don't have DevOps. So that's So there's always there's several problems here So a lot of companies even larger ones have moved over to and more modern approach, which is the generalist
15:20
So here you have people and they all have skills in Different areas and they all kind of have the same skill. So this is very similar to Full stack developer even though we still not fully sure what that even means but you can at least define some stack and say I want developers who can do that stack and here you have the advantage that everyone can work together much better because they understand the three levels if you have
15:42
someone who can do database stuff and back end and they everyone knows that you can you know collaborate much better and Similarly you can if one person gets sick leaves great You can you can replace them and you can also hire much better. You have your Template of what's what engineer do I need? What person do I need?
16:01
You can have your book of interview questions and you know if they pass that that's a full stack developer And that's a person I want to hire But one thing that's missing here again is well, you don't really have any deep specialization You don't have any experts. Everyone's kind of doing the same thing. And so a third You know alternative model is this idea of the complementary skills. So
16:23
Everyone here you see has like some deep foundation of skills, but they also have other ones they have kind of this very unique personal blend of skills and Yeah, that's something that's actually incredibly difficult to achieve for larger teams and it's not that it's like impractical
16:40
You see companies trying to do a lot with like setting up their labs Trying to hire like really high-profile developers who have that sort of profile to really get their thing moving forward, but it's it's difficult because You know, you don't have the same like predictability you have with a generalist But this is a really really unique opportunity You can use if you're a small company because you can just have a few of a few of those complementary people
17:05
With very different skills different specializations and and you know You have some overlap in them and that can really move you forward and that's something you can't necessarily do if you just Take advice from how very large companies are doing it And really what these these type of skills come come down to is this idea of what usually is called t-shaped skills
17:24
Because so you have like this kind of base and then you have like the arms for the t-shirt but I like to think of it more as tree shaped skills because Tree is really you have a solid stem and then you have you know, a few different branches of different different thicknesses and different You know, you you can also have like two trees that kind of overlap and the other nice thing is a t-shirt is pretty static
17:46
If anything it can only like I don't get worse a tree is trees alive and a tree you know lives the tree can grow so you can even if needed you can grow more skills and You know really develop in your career and actually yeah, I think we quite a good example of this
18:03
I would say because yeah in my case I've yeah, I'm a software developer I've always I've always kind of made websites as a teenager. I also happen to yeah, my degree includes my includes media, but also linguistics, so I've worked in marketing and sales, but I've kind of come back to
18:20
software development and Similarly, I'm a co-founder Matt. He's He used to be a researcher an academic mostly specializing in computer science. So that's a good stem there, but then he also happened to Become an expert in writing super-fast Python code and recently we had kind of had a gap in defobs And so he kind of happened to grow some skills in that direction as well, which we can use
18:45
so now we have that covered as well and It really I think this type of personality profile. That's not this is not I don't think I don't believe this is a personality trait I think most of us actually have this and Those types of people it's just that in a lot of a traditional environments
19:02
This is selected against and this is also not really encouraged because it doesn't really make you fit Well into any of those predefined roles, but once you once you leave that idea and once you You know move out of that perception. You actually You can you can build a really? Good team this way and you can really have a good
19:22
Yeah, you can cover everything you need with very few people and also without requiring Running at a loss because yeah, you need such a large team And the only thing is okay. So now you have you have your team of Tree shape people you've you have something you want to build you like all I don't need to run at a loss
19:44
I actually yeah, I can sell something But how do I make any decisions really someone needs to make decisions and that leads me to the next misconception Which is you can't make good decisions without testing all of your assumptions And I would say, you know, if you look at this first
20:01
it does submit sentence itself does sound kind of reasonable like it doesn't it doesn't sound as like outlandish as maybe some of the other misconceptions But I do think there are some problems with this kind of idea and I'll show you what I mean by that and what one good way of looking at this is Looking at how people assess failure when like a startup fails
20:23
So here's a typical thing of someone and that someone would say after yeah, they think didn't work out Which is turned out nobody wanted our product. I wish we'd spent more time validating Ideas next time I'll definitely you know, do the whole thing fully data-driven. I'll collect evidence on everything so I'm not making that mistake again and
20:43
If we think back to that little comic about the survivorship bias This is actually it's an interesting inverse of that kind of concept because it also says well We didn't do X and we failed and therefore we conclude that X would have saved us and that's That's a very yeah, that's a very interesting assessment because in reality well, do we really do we really know that and
21:06
Here's here's because here's some stats based on This is great website called an autopsy. I owe and they collect startup autopsies. So basically if a company fails They'll collect what went wrong and They also compile like interesting analytics
21:22
So this one is reason startup failed based on 300 of those autopsies and it might be important to note that these are actually these They only collect autopsies of startups that actually failed as in they closed their doors shut down stopped It doesn't include any startups who went on the amazing journey, which is also not necessarily a sign of great success
21:42
But so this is really just the actual failures and you see what people are saying Because another thing important to know this is based on people's assessment like the people running the companies. This is not like some Objective analysis of what happened so people say well not the right team wrong business model product not a hit no market need and out competed and
22:04
This to me this kind of shows that actually people don't really have a good sense for what's going wrong or how to even Assess that or analyze that I don't even really fully get the difference of wrong business model product not a hit No market need kind of all the same thing. Not the right team Maybe you thought you needed another engineer
22:22
Maybe you were actually an idiot and all your team hated you and that was the problem like who knows this is yeah It's really not that conclusive and but still what people Take away from this is like oh we need to just test more. We need to validate more Even if we think something is right, what if we're wrong? Let's a be test all the things and of course
22:43
I'm not saying that testing things is bad. Like we publish spacey We believe that it's absolutely important to do a rigorous evaluation of how our stuff performs How our stuff works and how accurate it is. We're not just saying oh, well, we put in a lot of effort when you wrote the code Looks good feels good. So must be gone
23:02
We don't need any numbers Of course not but you can easily fall into this trap where you're really trying to validate even things you already know Because maybe it could be that well not what if you're missing something what if you think the sky is blue But actually it's green like this could be you know groundbreaking So and a lot of this comes from the tournament mentality if you're taking part in a tournament
23:21
Well, what you need is you need that one thing. Nobody else has thought of because that's gonna make you win So it makes sense, but if you don't take it in part in the tournament, do you really need that? But not sure. So here's here's a more real-life Example of this so we have employee cat Looked at a company Twitter cringed was like, oh my god, really embarrassing. We look like we look super clueless
23:44
We just like we treating random crap five times a day. God Can we stop this and business cats like well, do you have any numbers to back that up? Why do you think that and well, no But you know without numbers, how do you know you're right? Well by thinking, you know, it looks it looks really bad
24:02
And you know, I'm sure you could pull up for both positions You could pull up some study that showed oh you have to have high engagement on Twitter because although you know You need to tweet five times a day Retweets are especially good to drive engagement and sales and on the other hand Well, of course, you can look at something and be like God this looks embarrassing. This is really bad. We probably
24:20
should stop this and What now what we can take take away from this is well There is always here There's always a value in like just actually reasoning about things and the decisions we make are based on reason And if you want to be if we run a business We need to be making lots of decisions all the time all day and it's not just one decision
24:41
We constantly need to decide and we win if we're right and we lose if we're wrong it's yeah, it sounds simple, but it well it actually it is like if you you know, if you if you Need to make you know, you need to decide something and you need to be right and the more right you are the Like faster, you gotta move forward and that's that's also why there's nothing wrong with building things that you think are good
25:02
It sounds yeah, it sounds like such terrible such a terrible idea. What if I think it's good and everyone else hates it. Well You didn't there likely be a lot of other problems down the line that will also not help you succeed So that's yeah You can't really make up for a lack of insight and reasoning by a be testing things to death and especially assuming that well
25:22
You're you're running a company, you know, you know things about what you're doing You're not going to go there and say well, I have no idea what I'm talking about. I'm absolutely clueless I don't know. I'm just gonna try everything and wait till something sticks you could but it's probably it probably be a very bad idea So, you know if you really say well, that's something I think is good and you go for that. There's really
25:42
You know if that succeeds that's a good sign if that already fails. Well likely you would have probably made a lot More decisions down the line that were equally as bad That would have still caused you to fail and so Actually, let's so basically, okay. So now we assume we have our company we're like, oh that's like not waste time on
26:03
let's always time on a be testing everything to death and Optimizing for some perfect outcome of something we don't even have We have a team of tree-shaped skills We're ready to make a profit. And so what are we actually? Making money off and what are we actually selling? And so here this is something that's especially in our industry. This is
26:26
Really one of the to me one of the absolute biggest Misconceptions that are really puzzling and I really lead to very very weird outcomes Which is people seem to believe well the true value whatever you're building the true value lies in your users data and the true value Data is the new oil whoever came up with that bullshit. That's kind of what drives that and
26:45
As an example, he's like just a little screenshot from our tool that we're selling prodigy and basically You know the idea is you buy it you pay us once you download it you get a lifetime license and it's yours it doesn't connect to the internet just runs on your machine and You know developers developers really like that people are like great. It doesn't try to trick me
27:04
I can just use it and it just works But when we started this a lot of people were like, oh my god, you people are so naive and stupid You're leaving so much money on the table like imagine if you could just track whatever you users are doing Why do you not give them an interface to upload their data and then you can keep it and then you can have it Everything they label you could have them train one AI model that can rule them all and
27:26
Because you have to consider what people are doing. There is for example a financial company wants to analyze Company reports better so they'll label whatever occurs there But the real the absolute highest value companies get out of the technology is the really really deep specialization
27:41
So yeah, you can have you can train something generic that just finds people in text But what really makes a difference is if you can make it find a very exact categories in your super specific text So what people if people are successful what they be doing here is so specific that it's absolutely Useless to us and in general people tend to very much overestimate
28:02
The value may of label data maybe because label data is often very painful and in reality Okay, if we were if we were going to do this and if we needed labeled data Well, we could just do it ourselves, right? We we happen to have this super efficient annotation tool, which also we happen to have built And so, you know
28:20
I think we'd be quite effective at just doing it ourselves instead of trying to engineer product in a way that basically yeah takes our users data so we can do something with it in the future and Again, this all really comes back down to that idea of taking part in a tournament because you know If you if you need have a lot of upfront cost you need to raise money
28:41
You obviously don't have a product yet, and that's fine We've seen plenty of companies where you know that has to be the case like you're building a social network Of course, you don't yet have a social network. Otherwise, you wouldn't be raising money to build it So you'd need to sell the potential of your application you just sell something that's in the future and you need to present that And that's more important in the reality obviously because you need to convince someone to give you money and that's fine
29:02
But all of these changes if you're not actually taking part in the tournament if you don't care about this Well, you know, you actually only have to think about what you can really charge people for money money for right now And that's clearly not the data because nobody yet cares about data You're hoping data is the new oil and at some point people will but actually in reality
29:23
It's also it's only it's only useful in cases Like for example Facebook Facebook's data is nice because they have everyone on the thing And for you a random application not not necessarily So, you know this it's really not what like the idea of adding other random objectives
29:42
To really make essentially make your product worse is really Yeah kind of defeats the purpose and it's kind of it's a bit like, you know these these superhero comics where they you have this super villain who engineers a teleporter or time travel machine and then uses that to rob banks and you're like
30:02
You could just imagine selling that your time machine how how much money could you make or Charge people money to use your time machine. That's like so great No, you do something really stupid with it And that's that's really that's I would say that's actually quite a similar idea You have a great product instead of selling it You're like, oh no I'll give it away for free so that at some point I can get maybe get something out of it that I can then
30:22
sell and so Instead I think what really know what really makes the whole thing You know work is what people people like stuff that works people like value and people Contrary to what people are going to tell you people companies don't mind paying for things
30:41
In fact, like they actually like to pay for stuff that works. They like to pay for stuff. That's theirs and that they can Keep and own and so instead of monetizing something abstract that you came up with here. There's this money right there There's a product right there. There's really something there that you can monetize and Also your users they're not your users are not
31:02
Interchangeable test subjects that you can like a B test stuff on there like actual the actual people So it's like, you know, if you're running a restaurant cool You can test on your friends at home and you have crazy ideas but you have to you have to be able to cook if you're the chef and you have to Know what people want to eat like you can't say hey, let's once people come in You can't be like, oh, let's let's test and sell everything for super cheap
31:23
And now next week say let's try how high we can go with our price before people leave And like how spicy can we make this dish before someone complains? It's like by after the third week you won't have any people anymore. And if you're actually selling something You know people This there's not an infinite supply of people you can AB test your software on and they reopen and they have money
31:46
and I want to buy things and you can just give them things and The bottom line is you know, people have come up with so many ways of measuring how successful your engagement like Okay, we've seen revenue is really good If you optimizing for that first and you have all these great graphs and you still don't know what you're doing
32:02
But actually there's a really really simple Really straightforward way that you can measure how successful you are and that's called profit And if you make profit is very it can be harsh because it's very difficult to lie to yourself If you're looking at your profit because you're not making profit you're not making profit But it also means that you know the enough like the bottom line is that you really okay you're making profit
32:24
You're looking at the profit. You could tell how well you're doing and one big advantage profit has over engagement over revenue over Anything else? It's yours. It's money. It pays your bills and you can keep it afterwards Thank you
32:54
Yeah, thank you very much. Enos. I think that was a really great keynote was it
33:05
And you as audience, you know the drill there are two microphones you come now come ask your question Question approaching we have time for really four to five questions. So people have to queue up
33:23
This is lovely. It seems like a dream and I'm sorry to be the past semester realist in this can you talk a little bit about maybe like work-life balance or Like it seems wonderful to have the most three like brightest friends coming together, but something has to give you know that
33:43
You have three which two wins out, can you talk a little bit more about them? Yeah, so I think it's definitely It's an important consideration I'm not gonna you know I would never defend this idea of like oh you just have to sacrifice like oh this guy never sees his family But look he's successful. So I
34:00
Do you think that's but I do think that actually, you know This is something that's that does become easier if you're actually able to You know engineer sort of your own business model and you know have a very direct way of selling something For money you'll have a product a great way about a product is that ideally if you're doing a good job, it's done So that's something that we basically at least aspire to that
34:20
We say we want to build something once and then it's Dan and we sell it and we fix a few small things we of course have to talk to people but the idea is always to ship something that's complete and that basically for us it Makes it much easier to You know have a more natural You know kind of timeline of this We know we're not gonna be chained to this for the rest of our lives and it's like gonna be really, you know terrible
34:43
We have something and we can resell it and it still provides value to people So, of course if it works, that's definitely something we want to optimize for and also, of course, I think I Have to say well, of course we choose to live in a place That's also more affordable than if we lived in the Bay Area, for example, that's a different consideration
35:02
And I also want to say this because no, it's bullshit like oh quit your job and do whatever No, don't if you you know if you need money, but I do think you know You can like at least you know, it's it's definitely something we're hoping to you know Engineer towards and that also helps having a you know, smaller team a smaller scale
35:22
very low overhead costs Which means you can I can say cool I'll take the weekend off and that's that's it and that's how can I can live my life because I'm not dependent on You know all this really one big project that I have to do and otherwise everything falls apart But of course, it's all you know It's a valid consideration and I do say I work I work a lot but also because it's something I enjoy so
35:45
It is very difficult for me to really draw the line there But I am very conscious and I would stop if it was if it got like out of hand Thanks, that was really interesting I'd like to ask if so many
36:01
companies and individuals In this sector are currently Chasing this ideology this dominant paradigm that you so tactfully described as a bit bullshitty Does that leave more room for the people who are following the kind of principles that you advocate or do you think it? Perhaps corrupts the entire marketplace and spoils things for everyone
36:23
I mean, that's that's an interesting question I mean like the bullshit definitely, you know corrupts a lot because I feel like actually it Sells people false promises and it actually keeps people from maybe pursuing things that actually they would be very successful in and drives them towards Other things that do not work as well
36:40
But I do think actually the fact that people are chasing these really large opportunities in a very specific sector and in a very specific Segment and in very specific business types does leave a lot of room for companies that maybe Back in a day would have been filled by a lot of other Companies and that I'm now kind of up for grabs because it's too low and it's too small and it's not worth it
37:03
So I see this as a big opportunity even though of course everything sucks and everything's terrible and like, you know I agree with that but like if there's some opportunity there and some hope then to me is that Would you say that explosion AI is a startup or a small business or is there no difference?
37:23
I mean I would sometimes maybe I wouldn't correct you if you refer to it as a startup because that's just what people You know think of but actually we see ourselves what we aspire to is this idea that you know in German we call Deutscher Mittelstand so like, you know middle-class businesses Germany's kind of famous for that actually, you know
37:42
UK as well at least before Thatcher so, you know This is that's a really that's and that's a concept that like really has proven itself to be, you know, very successful and You know you basically middle so middle stand the idea here is okay. You have a solid middle-class company It's founded by people who you know often made sometimes run by families even over generations and
38:04
You have it solid it makes a profit it often often also provides things and technology and Manufacturers things that are part of a much larger pipeline like one specific machine to put one car part together. That's like super Popular in like Germany and that's maybe a company you haven't heard of but it's a company that's doing very well
38:25
That's how where we see ourselves just more You know in a modern tech machine learning sector Okay, so the model for like the small team with a complementary and diverse skills you've been saying about like I think that that's great for a small company with like
38:43
People with incentive and kind of like senior people with a lot of drive But do you think there's a place for like junior developers in that that can only be like they could do like one thing Pretty productively, I guess I mean the thing is I think the whole junior developer idea is also very much
39:03
Shaped by the more traditional specialist generalist idea. So I actually I'm not even sure in such a and that kind of Let's see You just get somebody who's like fresh out the university and doesn't and simply doesn't have a lot of experience with different stuff So yeah, but I've been often sure but often, you know, they do know stuff
39:21
They can do stuff if they want to do the job There's actually a lot more you can you know, you can use that probably would get lost in a more traditional environment and I think there's actually there's a lot of space for learning new things because you know There are all these different areas that you can go into that. Maybe nobody else has claimed So sure, but of course in the founding team if you want to run the company build the product, of course
39:42
Ideally, it should be you know You pick something that you know that you're good at that you can build and then later you can get more people in Who can learn a bit maybe bring in their other expertise Then learn a bit more about what you're doing So I don't I don't think one thing really excludes the other but I always have to say I don't have like much experience With like hiring lots of what is considered junior developers and seeing where that goes. So you'd have to drive
40:06
Okay. Thanks. Okay now last questions comes from Antonio So great talk I didn't take my laptop out so it was very good So I want to play a bit the devil's advocate over here
40:22
So it's like you showed us a lot of examples of companies have been successful. Yeah, you know, I would say optimizing on loss for the first years and It shows only your company explosion is I as a company that succeeded instead using profit as KPI And all the other stuff. So how do I know this is not survivors bias again?
40:46
Well, sure you don't you don't hear about you know other the people who've tried exactly our model That's a good point and they're not standing here giving keynote today. So that's true But I think it's also if you actually look at the way the economy generally functions Right, and also even if you don't if you're not self-employed you work for a company
41:04
There's a very clear exchange of value there. You do something someone else says cool. You're good. I pay you money for it That's a much more natural flow of things and also to yeah And we've seen this in the past to replicate that that's a very straightforward model that like usually, you know, it doesn't necessarily fail but once it gets more speculative and weird then
41:23
You know all of that changes but the simple thing of you have something I want that I give you money for it You give me what you have I think that's a pretty solid Model that's like proven to be successful if you can do it right and have something that people want to give you money for Thank you very much Enos