YouTube2Text

Peter Reinhardt on Finding Product Market Fit at Segment — Transcript

by Y Combinator · 10,257 words · 1,434 segments · language en · Watch on YouTube

Full transcript

  1. 0:00the average person probably doesn't know
  2. 0:02what segment is mm-hmm
  3. 0:03so could you explain for sure so segment
  4. 0:06helps companies give their customers a
  5. 0:09better customer experience and we do
  6. 0:11that by helping them organize all of
  7. 0:12their internal data about all their
  8. 0:14interactions with the customer so for
  9. 0:16example if you go to the bank they
  10. 0:19interact with you at the ATM at the
  11. 0:21teller via a phone call center they have
  12. 0:24a web web app a mobile app they send you
  13. 0:26emails they're interacting with you
  14. 0:27across this huge surface area and they
  15. 0:30need to be able to coordinate that
  16. 0:32interaction right they need to know that
  17. 0:33if you encountered an error on the ATM
  18. 0:35the teller needs to be able to say like
  19. 0:37I'm so sorry you encountered an error
  20. 0:38about to be able to help you and so what
  21. 0:40we do is we help sort of bridge that gap
  22. 0:42of having a single record of all those
  23. 0:43interactions with each customer mm-hmm
  24. 0:45because previously companies would build
  25. 0:46all this in-house or not at all maybe
  26. 0:49yeah there's there's sort of two worlds
  27. 0:50one is they would build it all in-house
  28. 0:52exactly I'm gonna be a rat's nest of
  29. 0:54data pipelines from one place to another
  30. 0:55and so the engineering team would spend
  31. 0:57all their time building these data
  32. 0:58pipelines rather than actually building
  33. 0:59things for the customer that's one world
  34. 1:01the other world is one where it really
  35. 1:03used to be a one-on-one relationship
  36. 1:05with a bank branch manager for example
  37. 1:07and they might keep the information in a
  38. 1:10CRM mmhmm right but if you are similar
  39. 1:14would be an optometrist right you go in
  40. 1:16they have your past orders etc but now
  41. 1:18the world is moving much more to a Warby
  42. 1:20Parker kind of world where you're not
  43. 1:22interacting with a person so a CRM
  44. 1:24doesn't even make sense it's like not
  45. 1:25the right technology for understanding
  46. 1:27what interactions are having with a
  47. 1:28customer and instead it's all these
  48. 1:29different digital channels so that's
  49. 1:31where we come in and your first
  50. 1:32customers were they large scale
  51. 1:33companies like banks or who did you get
  52. 1:36in the beginning at the very beginning
  53. 1:37we actually launched as an open source
  54. 1:39library on Hacker News yeah and it took
  55. 1:42off there blew up basically overnight so
  56. 1:46to be fair it was kind of like the long
  57. 1:49road that's not a year and a half of
  58. 1:52dark times yeah yeah we shouldn't we
  59. 1:54shouldn't totally go over that but okay
  60. 1:55yeah so anyway you you blew up overnight
  61. 1:57again once it once it went live on a
  62. 2:00Chen it blew up overnight and so our
  63. 2:03first customers were the folks hanging
  64. 2:04out on on hacker news it was basically
  65. 2:08small companies for the most part
  66. 2:09founders who were looking for better
  67. 2:10ways to insta
  68. 2:11their web applications and mobile apps
  69. 2:13mm-hmm with this sort of analytics
  70. 2:15tracking and so the initial growth was
  71. 2:17that was just completely unpaid
  72. 2:20customers right so they're just using
  73. 2:21the open source whatever you put up on
  74. 2:23github right so the strange thing about
  75. 2:25the open source library which is sort of
  76. 2:27this data router so you put in one piece
  77. 2:29of data customer did X and then we turn
  78. 2:31around transform it and fan it out tall
  79. 2:32it from places downstream and what's
  80. 2:34funny about the library is if you want
  81. 2:36to turn on a new tool if you want to
  82. 2:37send the data to a new place you need to
  83. 2:39recompile the library and redeploy it to
  84. 2:40your website so it doesn't actually the
  85. 2:43open source library doesn't quite
  86. 2:44actually solve the problem which is
  87. 2:46really a market or a product manager
  88. 2:47wants to use a new tool so what happens
  89. 2:50is you really want to use the hosted
  90. 2:51version yeah so almost no one actually
  91. 2:53uses the open-source version as was
  92. 2:56addressed by design no no this is
  93. 2:58completely accidental really yeah
  94. 2:59there's just no way to make a good
  95. 3:00open-source product around it yeah
  96. 3:02that's crazy so okay so what did you
  97. 3:05apply to icy with we actually applied as
  98. 3:08a classroom lecture tool so the idea was
  99. 3:11to give students this button to push to
  100. 3:12say I'm confused and the professor would
  101. 3:15get this graph over time of how confused
  102. 3:17the students were we thought it was a
  103. 3:19really cool idea we were college
  104. 3:20students at the time and we had a bunch
  105. 3:23of professors who were excited about it
  106. 3:24at MIT and elsewhere I'll never forget
  107. 3:26our in our YC interview we were pitching
  108. 3:29this and PG was getting pretty excited
  109. 3:31and then he turns to professor Miller
  110. 3:35from from MIT who we had talked to and
  111. 3:38had given us the original idea for it
  112. 3:39and says hey would you use this and
  113. 3:42professor Miller says no probably not
  114. 3:47we just rolled with the punches and said
  115. 3:49yeah well you know we talked like twenty
  116. 3:51other professors and and they were all
  117. 3:52excited about it oh man but then you
  118. 3:54went through YC with this all right yeah
  119. 3:56building it out right we went through
  120. 3:57the whole whole YC with this idea built
  121. 4:00it out hundreds of thousands lines of
  122. 4:01code super complicated class or
  123. 4:03electrical product to have like
  124. 4:04presentation view and people could ask
  125. 4:06questions it's very complicated
  126. 4:07we were actually even raised money at
  127. 4:09demo day with this idea about 600k
  128. 4:11finally we deployed it in the classroom
  129. 4:13as the fall semester got started after
  130. 4:15demo day it was just a total disaster
  131. 4:17all students are my laptops and then I
  132. 4:19went straight to Facebook so the way we
  133. 4:22discovered this is we were standing in
  134. 4:23the back of the classroom
  135. 4:24counting laptop screens it'd be looking
  136. 4:26over the shoulders of the students near
  137. 4:27one two three and yeah we discovered at
  138. 4:30the beginning of class about 60% of
  139. 4:31students were on Facebook and by the end
  140. 4:33about 80 percent were on Facebook oh me
  141. 4:35in other words they were supposed to be
  142. 4:37using your desktop app the whole time
  143. 4:39that's right in the professor at the
  144. 4:41beginning class had been like can
  145. 4:42everyone please get out their laptops
  146. 4:43we're going to use this mute and all of
  147. 4:45a sudden all these students are
  148. 4:46distracted by Facebook so we had
  149. 4:47accidentally sort of put in an attention
  150. 4:49attention bomb if you want yeah and you
  151. 4:52hadn't brought on any users during why
  152. 4:55so you were in the suit wait I wrote it
  153. 4:56you were in the summer 2011 bat that's
  154. 4:58right summer 2011 and you weren't
  155. 5:00testing during the batch we tried
  156. 5:03testing but there's not that many
  157. 5:04classrooms that are in session during
  158. 5:05the summer right the school year starts
  159. 5:08in September so we had beta tested in a
  160. 5:10few summer computer science classes at
  161. 5:12both Stanford and Berkeley yeah but
  162. 5:14there was always technical difficulties
  163. 5:16and other things that sort of prevented
  164. 5:17us from getting like a real sense of
  165. 5:18what was happening often times it's
  166. 5:20pretty Yolo I don't I just we didn't
  167. 5:22really get tests rolling until right
  168. 5:24until the fall okay
  169. 5:25and so what was the like the come to
  170. 5:27Jesus moment where you realize you had
  171. 5:29to change the product standing in the
  172. 5:31back of a BU classroom as an
  173. 5:33anthropology class and I remember
  174. 5:35arriving at the 60% and the 80% and we
  175. 5:38went up and apologized to the professor
  176. 5:40and and at that point you're just like
  177. 5:43okay we got to kind of shut this down or
  178. 5:45figure out what to do with the money yes
  179. 5:47right well we had we had just gotten
  180. 5:49wires for these for these checks for
  181. 5:51this money like literally a week before
  182. 5:52right so we called back all the
  183. 5:55investors and we were like well it turns
  184. 5:56out this is a terrible idea so what do
  185. 5:58you want us to do with the money and
  186. 6:00almost all of them said well you know we
  187. 6:02invest it for the team so so go find
  188. 6:03something else okay um okay next day
  189. 6:06next day we're like well what are we
  190. 6:08gonna do and we realized we should have
  191. 6:11been able to figure out some of this
  192. 6:12analysis by not just standing in the
  193. 6:14back of the classroom like we should've
  194. 6:15been able to see some of this digital
  195. 6:16you couldn't see it in the analytics
  196. 6:18metrics at all so we decided hey let's
  197. 6:21build an analytics tool let's build a
  198. 6:22better web analytics mobile web app
  199. 6:25analytics tool to compete with Mixpanel
  200. 6:26and Google Analytics the idea was to
  201. 6:30give really advanced segmentation
  202. 6:31because we also wanted to understand how
  203. 6:33some computer science classrooms at MIT
  204. 6:35we're using it differently than
  205. 6:36anthropology classes would be
  206. 6:38and we couldn't do that analysis and the
  207. 6:40tools we had so that was the idea was to
  208. 6:42build an analytic tool mm-hmm we spent
  209. 6:44about a year building the infrastructure
  210. 6:45necessary to do the analytics and really
  211. 6:49we're not succeeding in getting any
  212. 6:51customers during that timeframe so you
  213. 6:53had I mean did you have a whole
  214. 6:55onboarding process was it like a landing
  215. 6:57page what did it look like oh yeah we
  216. 6:59had a landing page I was going on little
  217. 7:00sales trips I was meeting with people
  218. 7:02tell you so you were trying oh yeah we
  219. 7:04were trying okay but it was it was not
  220. 7:07going well basically what happened is
  221. 7:09people would say well I already have
  222. 7:10this other analytics tool installed so
  223. 7:11like it's not that interested you
  224. 7:14definitely you have the time you were
  225. 7:16not 10x better than Google no no yeah
  226. 7:19yeah were you at parity we were at
  227. 7:22parity in certain dimensions and we were
  228. 7:25exceeded in other dimensions but they
  229. 7:27just weren't them dimensions that
  230. 7:28mattered apparently right okay so you're
  231. 7:31just kind of like blow up overnight with
  232. 7:33accra news and that like this is well so
  233. 7:36positive we're not even this is just the
  234. 7:38analytics tool which we realized in
  235. 7:39December 2012 was failed so we're like
  236. 7:42well over a year into the analytics tool
  237. 7:43we're now have tried two ideas both have
  238. 7:46failed neither is clearly gonna work at
  239. 7:48all yeah we realized that we're screwing
  240. 7:50up so we decide to have office hours
  241. 7:52with YC again I would come back walk
  242. 7:55around a little cul-de-sac by YC with
  243. 7:56with PG he started comes ta stop and
  244. 8:00says so just to be clear you've spent
  245. 8:01half a million dollars and you've
  246. 8:02nothing to show for it gulping moment
  247. 8:06yeah I guess that's true and it was a
  248. 8:10good sort of come-to-jesus moment though
  249. 8:12and you got a pause there and then
  250. 8:15rewind all the way back to the first
  251. 8:17week of YC and it was in that first week
  252. 8:20that we'd been like well we should have
  253. 8:21analytics tools on our classroom extra
  254. 8:23tool say we've googled analytics and we
  255. 8:26found Google Analytics Mixpanel and
  256. 8:27KISSmetrics and we're looking at them
  257. 8:29we're like we don't know which one of
  258. 8:30these things we should use all right
  259. 8:32like they're kind of all similar but
  260. 8:33Google Analytics is a little more
  261. 8:34marketing a KISSmetrics is a little more
  262. 8:35revenue we mix panels a little more
  263. 8:37product II in terms of the sorts of
  264. 8:38insights they can give you and at the
  265. 8:42end of the day that they all collect the
  266. 8:44same data they all collect
  267. 8:45basically who is the customer and what
  268. 8:47are they doing mm-hmm so we decided to
  269. 8:49write this little tiny abstraction that
  270. 8:51could
  271. 8:52data to all three I was like 50 lines of
  272. 8:54code among the hundreds of thousands for
  273. 8:55this classroom lecture tool and then we
  274. 8:57decided we'll just send it at all three
  275. 8:58we'll look at all the tools and we'll
  276. 8:59just pick the one that we need so it'll
  277. 9:01give us a lot of option allottee
  278. 9:02basically for free yep and then we
  279. 9:04forgot about it for like four months
  280. 9:06so months later we clean it up a little
  281. 9:07bit more four months later it cleans up
  282. 9:09a little bit more as I mentioned now
  283. 9:10we're struggling at that point we're
  284. 9:12struggling with this question of why I
  285. 9:13already have mixed panel installed so I
  286. 9:14don't really want to use your analytics
  287. 9:15tool yeah so my co-founder Ilya has this
  288. 9:19idea he's like remember that little
  289. 9:20library wrote that sort of abstracted
  290. 9:21away the differences between these tools
  291. 9:23and what a Sun data to all three what if
  292. 9:24we added ourselves just the fourth
  293. 9:25service that I could send data to you
  294. 9:27and then every time someone has that
  295. 9:29objection we hit him back with an open
  296. 9:30source library that they can use to send
  297. 9:32to us and them that cause seems like a
  298. 9:34clever growth hack it like gets us
  299. 9:35around this problem so we did that
  300. 9:37cleaned it up open sourced it and people
  301. 9:40started replying like all this libraries
  302. 9:41great we'd love to use it couple weeks
  303. 9:43later we'd follow up and be like hey
  304. 9:45well we saw you're using you know the
  305. 9:46open source library but all you have to
  306. 9:49do is copy paste our API key so that you
  307. 9:51can use our analytics tool could you
  308. 9:52just copy paste the API key yeah no like
  309. 9:54yeah so we started just feeling like
  310. 9:58there's some traction on on this little
  311. 10:00routing library which is maybe up to
  312. 10:02like 25 stars on github or something
  313. 10:04like that like not not much but some
  314. 10:06people are issuing pull requests and
  315. 10:08it's the first time we've ever faults
  316. 10:09are like pull like it if you always we
  317. 10:12weren't just pushing a boulder uphill
  318. 10:13this is a little different but subtle we
  319. 10:16had this conversation with Paul Graham
  320. 10:18the next day we sit down this is our
  321. 10:20second time with this we're like okay we
  322. 10:22have 100k left in the bank what's our
  323. 10:23final shot and my co-founder Ian is like
  324. 10:26you know what I think there's a big
  325. 10:28business behind analytics j/s which is
  326. 10:30this routing library like that is
  327. 10:33literally the worst idea I've ever heard
  328. 10:34like it's 500 lines of code to grow a
  329. 10:37little bit's 500 lines of code in it's
  330. 10:38already open source I have no idea how
  331. 10:42you build a business around that yeah
  332. 10:44and so we fought about it all day long I
  333. 10:47was four of us I was the most skeptical
  334. 10:50I think just literally like brutal
  335. 10:51brutal fighting I went home and I was
  336. 10:53trying to figure out how to kill the
  337. 10:53idea was awake half the night finally
  338. 10:56figured it out came in the next day I
  339. 10:57was like alright guys here's we're gonna
  340. 10:58do we're gonna build like a beautiful
  341. 11:00landing page really gonna pitch the
  342. 11:01value of this analytics j/s open-source
  343. 11:03library and it'll have
  344. 11:05a form at the bottom so that we can get
  345. 11:07people to sort of express interest put
  346. 11:09it up on Hacker News and we'll see what
  347. 11:10happens I was thinking like totally kill
  348. 11:12it right I think that's well in hacker
  349. 11:14news right yeah so we'd build a landing
  350. 11:17page put it up on Hacker News and this
  351. 11:18is where we have this you know you're in
  352. 11:20half in the making overnight explosion
  353. 11:21yeah and that that kind of segues into
  354. 11:24your whole startup school talk right
  355. 11:27about legs right basically real product
  356. 11:30market fit yeah and up to that point in
  357. 11:33your life had you launched anything
  358. 11:35where the market I mean this is a Marc
  359. 11:37Andreessen quote because he kind of
  360. 11:39coined the term it pulls it out of you
  361. 11:41right that's right I think you
  362. 11:43experiencing hmm I think that's an apt
  363. 11:45description I had never experienced it
  364. 11:46before and it feels very much like
  365. 11:48losing control
  366. 11:50right like previously you're like
  367. 11:52building a thing and you roll it out in
  368. 11:53a building a thing you're pushing it out
  369. 11:55and all of a sudden you like put a thing
  370. 11:56out there and people start running away
  371. 11:58with it and using it in ways that you
  372. 11:59didn't necessarily expect and you're
  373. 12:01sort of like what it's just an it's just
  374. 12:03a man like stop stop it like we need to
  375. 12:05fix these other things because otherwise
  376. 12:06it's like this feeling of losing control
  377. 12:08and almost like the market is dictating
  378. 12:11to you now what the rules of the road
  379. 12:13are and what needs to get built interest
  380. 12:15so would you're a different feeling yeah
  381. 12:16would you differentiate that from
  382. 12:18overwhelming demand for one particular
  383. 12:20feature versus like we're just gonna
  384. 12:24take this and use it however we want but
  385. 12:26there's a ton of demand there would you
  386. 12:28separate those two things not really I
  387. 12:33think the people always want more
  388. 12:35features yeah but the the thing that
  389. 12:38flipped was people would previously tell
  390. 12:40us they wanted a feature but not use it
  391. 12:43whereas now people were using it and
  392. 12:44they would want a second feature and the
  393. 12:46it's a super important distinction I
  394. 12:48think a lot of founders get caught in
  395. 12:49this sort of I call the the death spiral
  396. 12:51of user feedback yeah where they keep
  397. 12:53going and showing someone their product
  398. 12:55and asking them for feedback they give
  399. 12:56them you know some feedback about how
  400. 12:58they could make it better but they don't
  401. 12:59use it and then they bring it back with
  402. 13:01those fixes and they ask if this is
  403. 13:02better and it's just like the death
  404. 13:04spiral where it never gets anywhere but
  405. 13:05once someone starts using it they'll
  406. 13:06have more requests and that just means
  407. 13:07they're gonna pay you more over time
  408. 13:08right yeah I like how you put it in the
  409. 13:10lecture where you basically if you have
  410. 13:12to ask yourself it's not product market
  411. 13:14fit
  412. 13:14yeah it's it's you really can't miss it
  413. 13:16you
  414. 13:17and said this was now six years ago five
  415. 13:21five or six almost almost six years ago
  416. 13:23yeah right and half you and you're so
  417. 13:25feeling the same way yeah yeah and since
  418. 13:27then we've had a few more sore a
  419. 13:29secondary product market fit moments
  420. 13:30like what about two years in we
  421. 13:35discovered that all of our most valuable
  422. 13:36customers were sending their data to an
  423. 13:39s3 bucket there's basically they're
  424. 13:41keeping log files of the raw data which
  425. 13:44was a little weird because typically
  426. 13:45what people were using the data for was
  427. 13:47to send to an analytics tool an email
  428. 13:48marketing tool and a CRM and help desk
  429. 13:49like places where a business person is
  430. 13:51deriving value log files is a little
  431. 13:53different that's a little weird it's
  432. 13:55unclear what the use case is so we went
  433. 13:57on this sales trip to New York myself
  434. 13:59and our first sales person RAF we met
  435. 14:03with five customers that were using this
  436. 14:04s3 bucket let me just ask them oh my god
  437. 14:08what are you doing with the s3 bucket
  438. 14:09the first customer was like well you
  439. 14:11know we have a data engineering team
  440. 14:13that's taking the data out of the bucket
  441. 14:14and converting it into CSV files and
  442. 14:17then they're uploading it to our data
  443. 14:18warehouse which is a redshift cluster so
  444. 14:20basically they were using it as the
  445. 14:21initial endpoint input into a ETL
  446. 14:23pipeline like oh that's interesting but
  447. 14:26man to the next meeting second customer
  448. 14:28is like well we have a data engineering
  449. 14:30team who's taking the data out of s3
  450. 14:31converting it miss yes okay drops and
  451. 14:35that's interesting and then the third
  452. 14:38fourth and fifth ones all said exactly
  453. 14:39the same thing and and that was the
  454. 14:41point at which I started becoming a
  455. 14:42conspiracy theorist yeah it seemed it
  456. 14:44seemed like some pre meeting had
  457. 14:45happened now but they were all doing
  458. 14:47exactly same thing so it was really
  459. 14:48obvious we just built a way to load data
  460. 14:52directly from segments into a regio
  461. 14:53fluster huh and that was a huge thing
  462. 14:56like you you was really go out that it
  463. 14:58was product market fit again yeah it was
  464. 15:00very explosive
  465. 15:01so we you know we'd grown revenue from
  466. 15:03zero to two and a half million in the
  467. 15:05first year and then we launched this red
  468. 15:09shift connector and the next year one
  469. 15:11from two and a half to ten but people
  470. 15:13weren't asking for that that's right it
  471. 15:16was one step too far for them to realize
  472. 15:18that we could do it easily yeah their
  473. 15:20mentality I think was that oh segment is
  474. 15:22a way that I integrate marketing tools
  475. 15:23and so a data warehouse is in a
  476. 15:25marketing tool it's a bi tool surely
  477. 15:28segments can integrate that just didn't
  478. 15:29click
  479. 15:29ahead to go
  480. 15:30by asking hmmm interesting and and your
  481. 15:33growth how how how does that happen is
  482. 15:36it has it come through developers I go
  483. 15:40to market model is primarily through
  484. 15:41through engineers yeah we talk a lot
  485. 15:45about sort of the way that we've built
  486. 15:47our infrastructure over Jim
  487. 15:49we obviously process a lot of data so
  488. 15:50there's a lot of interesting
  489. 15:51infrastructure problems I think now
  490. 15:53we're processing you know hundreds of
  491. 15:54thousands of user actions per second so
  492. 15:57there's a lot of data going through
  493. 15:58there we write about that
  494. 16:00that's generally interesting to that to
  495. 16:02the hacker news and an engineering crowd
  496. 16:03and yeah typically an engineer is the
  497. 16:05one that brings this brings us in
  498. 16:07sometimes that really technical product
  499. 16:09manager but it's someone who's like yeah
  500. 16:11this is gonna solve this weird rat's
  501. 16:12nest data pipeline problem that I've got
  502. 16:14well and how much of it is open source
  503. 16:16still a good portion of it is is open
  504. 16:19source but most of the value that we
  505. 16:21deliver is actually by running the
  506. 16:22hosted version right cuz it's like at
  507. 16:24the end of the day it's not just the
  508. 16:26develop like you're you're saving the
  509. 16:27developers time but it's these business
  510. 16:29people that really need it yeah and and
  511. 16:32frankly most of the complexity is hidden
  512. 16:34away and how you actually operate and
  513. 16:35scale a data pipeline that that is
  514. 16:37processing the data yeah so you know our
  515. 16:40JavaScript libraries open source our iOS
  516. 16:41SDK is open source or Android SDKs open
  517. 16:43source II wouldn't sort of collected
  518. 16:45data from anywhere and those collection
  519. 16:47libraries are our open source but the
  520. 16:49sort of core infrastructure pipeline is
  521. 16:51not hmm okay and so right before you
  522. 16:54guys launched on HN where did this like
  523. 16:56small tiny micro launch I whatever it
  524. 16:59might be
  525. 16:59were there other avenues that you were
  526. 17:02considering pursuing like in that debate
  527. 17:05in the day before mmm were you thinking
  528. 17:07about other stuff I think the debate was
  529. 17:11whether to build out the full product
  530. 17:12and then test for product market fit by
  531. 17:15trying to sell it to people versus this
  532. 17:17super super lightweightt MVP landing
  533. 17:20page that we wouldn't put on Hacker News
  534. 17:21to see if there was interest in the
  535. 17:23concept yeah and and what drove us
  536. 17:26towards the super super lightweightt
  537. 17:27test was actually the fact that there
  538. 17:29was a skeptical divide among the
  539. 17:31founders and since the founders couldn't
  540. 17:34agree the only way to answer the
  541. 17:36question was to go to customers a SAP
  542. 17:39and get an answer okay it's tough like
  543. 17:42the the launching early thing is always
  544. 17:44a
  545. 17:44because I think there have been
  546. 17:46instances where people are like I will
  547. 17:48just launch this early but because
  548. 17:49they're like 10% off of what that
  549. 17:51product ought to be or they're not very
  550. 17:53good at communicating it they never
  551. 17:55really get the feedback that they need
  552. 17:57right like how do you how do you kind of
  553. 17:59balance that out like this is kind of
  554. 18:01like fully formed enough or were
  555. 18:02communicating it clearly enough that we
  556. 18:05can launch it like how do you determine
  557. 18:06that usually that test is so cheap to
  558. 18:11run that it's worth running even if you
  559. 18:14decide that it was inconclusive and you
  560. 18:17should go a step deeper mm-hmm but I
  561. 18:19also think that the way frame fit is
  562. 18:21actually an excuse that a lot of
  563. 18:22founders use for not doing the cheap
  564. 18:25early tests when in fact they they
  565. 18:27should yeah I I get the the same
  566. 18:29questions like almost every single
  567. 18:31podcast and I should be a little bit of
  568. 18:34a devil's advocate here but yeah these
  569. 18:36are kind of like straw man arguments
  570. 18:38right I think the really big product
  571. 18:40market for moments for every company are
  572. 18:42pretty unmistakable like the Dropbox
  573. 18:44founders have called a stepping on a
  574. 18:45landmine I just I really don't think you
  575. 18:48can mistake it it really happens in a
  576. 18:51way that you you lose control it's very
  577. 18:52obvious every much metric goes haywire
  578. 18:54people are talking about it a lot it's
  579. 18:57not it's not mistakable or like oh well
  580. 19:01you know this person said that it looked
  581. 19:02valuable and was really exciting and
  582. 19:03blah blah blah if they're not using it
  583. 19:05like it's not there yeah okay well okay
  584. 19:09so there's a question from Twitter it's
  585. 19:11clear that a bunch of people have
  586. 19:12watched your your lecture so this first
  587. 19:15one from Aaron Evan Farrell he asked you
  588. 19:18mentioned in the startup school lecture
  589. 19:19that you had to pivot to analytics j/s
  590. 19:22to find Prada market is it possible to
  591. 19:26purely iterate on something like that to
  592. 19:28find product market fit or should be
  593. 19:30clear from the outset if a new idea is
  594. 19:32something people want there's two
  595. 19:34versions of this I will say the Airbnb
  596. 19:36version of product market fit is much
  597. 19:38more iterative they struggled for years
  598. 19:40and years and made slight iterations and
  599. 19:43iteration that iterations and finally I
  600. 19:44caught on and obviously they're they're
  601. 19:45their runaway success my feeling is that
  602. 19:50that's extremely rare and that again
  603. 19:54this is a really dangerous place to be
  604. 19:57because you can stay in this iterative
  605. 19:59mode for years yeah and it is unlikely
  606. 20:02that the iterations are going to get you
  607. 20:04to a good place
  608. 20:05so I remember very clearly early on
  609. 20:07being really inspired by the Airbnb
  610. 20:09story and it being a logical reason why
  611. 20:12we should keep plugging away at a bad
  612. 20:13idea and I think we abused the Airbnb
  613. 20:16story to just keep stringing ourselves
  614. 20:18along on a bad idea so I would be very
  615. 20:20very careful of following their B&B
  616. 20:22example I don't know many other
  617. 20:23companies that that hit proud of market
  618. 20:25fit that way right and so how long do
  619. 20:27you give an idea at this point I
  620. 20:29actually don't think that it's quite the
  621. 20:32right frame to think about it in terms
  622. 20:33of how long to give an idea okay I think
  623. 20:35what you want is someone either yourself
  624. 20:37or someone else on the founding team
  625. 20:38who's a skeptic so someone who is going
  626. 20:42to have enough context with whatever the
  627. 20:45specific idea is and whatever the sort
  628. 20:47of regime or market you're in someone
  629. 20:51who's skeptical who will question and
  630. 20:54push for the fastest reasonable test so
  631. 20:57in other words if you have an optimist
  632. 20:58and a skeptic and they both agree on
  633. 20:59what a valid test is then I think you
  634. 21:02actually lend up with a good test but if
  635. 21:03you have three optimists in a room who
  636. 21:04all agree on what a good test is I don't
  637. 21:06believe that that's a good test did over
  638. 21:08skeptics but yeah do you have you
  639. 21:10recruited skeptics or did you just kind
  640. 21:12of like luck into that I think we lucked
  641. 21:13into it the first time yeah I do think
  642. 21:17we have some folks on the team that
  643. 21:19segment some early folks who are who are
  644. 21:22skeptics usually about future product
  645. 21:25market fit moments that we've had and I
  646. 21:26think it's been enormously helpful
  647. 21:27that's really interesting what how do
  648. 21:29you test for that like in an interview
  649. 21:31scenario we didn't test for we just got
  650. 21:33lucky again but yeah okay so not even
  651. 21:34just with co-founders like with early
  652. 21:36employees as well that's right yeah
  653. 21:38how interesting like how hurtful can it
  654. 21:40be if someone is like well I really
  655. 21:41think you haven't thought this through
  656. 21:42there's like these three things that you
  657. 21:44should really test ASAP because I don't
  658. 21:46really believe that you have product
  659. 21:48market fit here right that's what you
  660. 21:50want you want someone who's gonna be
  661. 21:52like pushing it and you're like what
  662. 21:54happen
  663. 21:54yeah how would we and then who's willing
  664. 21:56to collaborate with you on how you
  665. 21:57should reasonably test whether those
  666. 21:58things are the case so the sorts of
  667. 22:00tests that we have run for example with
  668. 22:03this mindset recently and even in the
  669. 22:05past year yeah where should we switch
  670. 22:06from a technical buyer to a marketing
  671. 22:08buyer
  672. 22:10unclear how to test that well so the
  673. 22:13this early skeptic who's amazing her
  674. 22:15name is Diana she I was like well I'm
  675. 22:18just gonna go to a conference with
  676. 22:19marketers and I would try pitching a
  677. 22:20bunch of marketers just flew to Florida
  678. 22:21and pitched a bunch of marketers came
  679. 22:23back she's like nope not a good idea and
  680. 22:26I read my own set of tests so like the
  681. 22:29the hacky ways to test these things I
  682. 22:31think are very valuable in it and it it
  683. 22:32comes from having skeptics who and
  684. 22:34different perspectives of people willing
  685. 22:35to go test those things okay
  686. 22:37and so I imagine these these tests from
  687. 22:39skeptics occur on a maybe daily but
  688. 22:42probably like at least a monthly basis
  689. 22:44right in terms of you guys working on
  690. 22:46your product yeah I'd say maybe more on
  691. 22:48like a pre idea basis so like if we're
  692. 22:50gonna launch a new product then it's
  693. 22:51really helpful to have a skeptical
  694. 22:53perspective of like here's why this
  695. 22:54might not actually be a good idea and do
  696. 22:56you rely more heavily on data or actual
  697. 22:59customer interaction like in the early
  698. 23:01part of the product development process
  699. 23:02it's all qualitative it's all talking
  700. 23:04with customers okay because this is the
  701. 23:06thing that bugs me more is like when
  702. 23:08people are just like putting up landing
  703. 23:09pages left and right and like thinking
  704. 23:11that they can like kind of uh I forget
  705. 23:14what it's I will call it like what do
  706. 23:17they do it doesn't know yet like
  707. 23:18Imagineer your way towards them like
  708. 23:20winding down this path to finding it in
  709. 23:24an advanced inefficient with time yeah
  710. 23:26they're scared yeah and and when you
  711. 23:30actually go and talk to a customer if
  712. 23:31you have that conversation in the right
  713. 23:33way you'll learn a thousand times more
  714. 23:35from that conversation than you will
  715. 23:36from them from putting up a landing page
  716. 23:39yeah and I think ultimately we learned a
  717. 23:42lot more from talking to our customers
  718. 23:44after the hacker news landing page than
  719. 23:45we did from the landing page itself yeah
  720. 23:47totally so what are your what are your
  721. 23:49tactics when you're talking to customers
  722. 23:50yeah I'd say the main thing is most
  723. 23:53founders are not familiar with how a
  724. 23:55sales process is actually run and you
  725. 23:58basically want to run a sales process so
  726. 24:00the the sort of typical founder motion
  727. 24:02with running a sales process is the they
  728. 24:04come in and they say okay I'm going to
  729. 24:05give you a demo and it's like really
  730. 24:07shiny polished pitch and then the
  731. 24:09customer decides at the end of that
  732. 24:10pitch whether they're interested or not
  733. 24:12that's not actually how good sales works
  734. 24:14at all the way good sales works is you
  735. 24:17do qualification up front
  736. 24:18so you have some method
  737. 24:20of understanding the customers problem
  738. 24:22better than they understand it
  739. 24:23themselves and then you do the
  740. 24:24computation in your head as to whether
  741. 24:25your product is a fit for their problem
  742. 24:27there's a lot of methodologies for this
  743. 24:30the methodology that we use that segment
  744. 24:32is one called medic ME ddic it's
  745. 24:36literally just a list of sales
  746. 24:37qualification criteria and this is this
  747. 24:39is what sales reps do if a sales rep
  748. 24:40comes back and we're like we're gonna
  749. 24:41close you know this this this deal the
  750. 24:44sales manager says okay well let's go
  751. 24:46through and metrics what are the metrics
  752. 24:48by which this company is going to judge
  753. 24:49whether or not the product works for
  754. 24:50them and if the sales rep can't answer
  755. 24:52metrics economic by our decision-maker
  756. 24:54decision process the identified pain and
  757. 24:58doesn't have a champion if they can't
  758. 24:59have all six of those things yeah
  759. 25:01there's no deal and so when you're
  760. 25:03searching for product market fit you can
  761. 25:04just go through all of those things by
  762. 25:06asking the customer a ton of questions
  763. 25:07mm-hmm and then you can grade whether or
  764. 25:09not you're actually going to build a
  765. 25:10product that will solve the problem
  766. 25:11right well this is it kind of ties into
  767. 25:14this like skeptic versus like optimist
  768. 25:16idea right you have someone who's like a
  769. 25:18champion of the product and in many ways
  770. 25:21I mean maybe this is you like the
  771. 25:22optimist who just sees the world and
  772. 25:24they see this future and it looks
  773. 25:26awesome it's amazing but you need that
  774. 25:28skeptic who sees the world as it really
  775. 25:29is that's right and a sales
  776. 25:31qualification criteria is is a way of
  777. 25:33almost putting the skeptic out as they
  778. 25:34like structured process that enforces
  779. 25:36some level of skepticism yeah I think
  780. 25:39it's so dangerous when you're the author
  781. 25:41because I fall into this camp for the
  782. 25:43most part like when you get good at
  783. 25:46sales you can kind of sell many people
  784. 25:48on almost anything but if that product
  785. 25:52doesn't exist yet it's very easy to just
  786. 25:55kind of mold it in the way when you're
  787. 25:56reading someone you're like oh I can
  788. 25:59totally kind of want it to be like this
  789. 26:00so I'm gonna kind of go down this path
  790. 26:02but then when you actually show them the
  791. 26:04product and they're like like you said
  792. 26:06they won't even install it then you see
  793. 26:08the world that's it really is
  794. 26:09yep and that's the thing and so we you
  795. 26:11guys are just like going out and how are
  796. 26:13you still having these conversations
  797. 26:14with people personally sometimes yeah
  798. 26:17for sure yeah yeah because this is one
  799. 26:22of the things that like people I think
  800. 26:24in large part because they're influenced
  801. 26:26by your startup school talk they have so
  802. 26:28many questions about it and so Benjamin
  803. 26:29Liam asked like you know how do you even
  804. 26:32how do you find that you have the right
  805. 26:34messaging around you're explaining your
  806. 26:36product oh man this is super hard I'm
  807. 26:38not even the right person to ask about
  808. 26:39this
  809. 26:40I should Diana who I mentioned before in
  810. 26:41our VP marketing Holly are the two
  811. 26:44people who have really refined our
  812. 26:45messaging over the years and we're
  813. 26:48always trying to refine it so I don't
  814. 26:51know how you know that you have the
  815. 26:53right messaging you know that whatever
  816. 26:56messaging you can sort of test whether
  817. 26:57alternate messaging is going to work and
  818. 26:59you can do that yeah qualitatively in
  819. 27:01interviews with customers you can try
  820. 27:03explaining it one way and see if their
  821. 27:04eyes light up you can try explaining
  822. 27:05another way and just sort of see what
  823. 27:07what resonates I think a really talented
  824. 27:09early salesperson will also have this
  825. 27:13sort of pattern in their in their habit
  826. 27:17of how they pitch that they'll always be
  827. 27:19testing different ways of explaining the
  828. 27:20product that was definitely true for for
  829. 27:22the first salesperson that we hired he
  830. 27:25was fabulous it just like constantly
  831. 27:26experimenting with different ways of
  832. 27:27doing it so I don't know if you there's
  833. 27:28you never know if you have the best
  834. 27:30messaging but you are constantly
  835. 27:33searching and testing for four different
  836. 27:35ways of explaining it okay but but again
  837. 27:37like if it if it's about you know really
  838. 27:41finding a good product market fit yep do
  839. 27:43you think that leave these like minor
  840. 27:45changes and how you communicate
  841. 27:46something will will make the difference
  842. 27:47I don't think I don't think minor
  843. 27:49changes then it will make the difference
  844. 27:50now once you have product market fit
  845. 27:52then sure you can optimize the messaging
  846. 27:53okay so then we should talk about idea
  847. 27:56generation because that seems more
  848. 27:57important yeah and these like minor
  849. 27:59deviations yep weird where do you begin
  850. 28:03yeah I think the best ideas that we've
  851. 28:06had come so there's a big difference
  852. 28:08between the first idea and these sort of
  853. 28:12like follow-on ideas and and the reason
  854. 28:15so the first idea meaning like the core
  855. 28:16product that's and then the individual
  856. 28:18features that's right okay and not just
  857. 28:20individual features you might have
  858. 28:21entirely new products that come along
  859. 28:23but those are much easier right the the
  860. 28:27problem with the first product and
  861. 28:30product market fit is that you can move
  862. 28:31the product and you can move the market
  863. 28:33because it's fit between these two
  864. 28:34things and so it's unclear and they move
  865. 28:37in some crazy multi-dimensional space
  866. 28:39and so what's the the issue is that to
  867. 28:41get them to both match up you can always
  868. 28:45move either one and in different
  869. 28:47conversations
  870. 28:47in one conversation you might shift the
  871. 28:49product and you might in a different
  872. 28:50conversation realize you need to shift
  873. 28:51the market so that's super tricky I
  874. 28:54don't think there's a repeatable way to
  875. 28:56do that mmm I think you just have to go
  876. 28:59very very deep into a particular market
  877. 29:01and understand the problems that people
  878. 29:03have in that market so do you have a
  879. 29:05particular a process for idea generation
  880. 29:08or you just you get into something and
  881. 29:11you're like man just go super deep yeah
  882. 29:13for that first idea you just have to go
  883. 29:15super deep you just have to understand
  884. 29:16the market and the ecosystem and the
  885. 29:18customers upside down backwards better
  886. 29:20than they do themselves okay so you were
  887. 29:23booted from segment today do you know
  888. 29:26where you would start you'd have to
  889. 29:28start with something that was
  890. 29:29interesting to you personally and then
  891. 29:30you'd go dig in in a deep direction I
  892. 29:33think that it becomes more repeatable
  893. 29:35when you are finding a second product so
  894. 29:38at that point you've mostly lost the
  895. 29:40market side right because you already
  896. 29:41have a buyer you already have a go to
  897. 29:42market motion you already have like an
  898. 29:44area of interest which for us you know
  899. 29:46is these sort of data pipelines and data
  900. 29:48infrastructure customer data
  901. 29:49infrastructure then it's much easier
  902. 29:51because you know exactly who you need to
  903. 29:54go to and you know roughly the like type
  904. 29:56of questions that you need to ask and
  905. 29:57then you can run a process which is much
  906. 30:01a much deeper x-ray of the customer than
  907. 30:03you might be comfortable with at least
  908. 30:05it was much deeper than then I was
  909. 30:07comfortable with when I first got
  910. 30:09started like as an example we recently
  911. 30:12were testing product market fit for our
  912. 30:13product we're gonna announce at our user
  913. 30:15conference in September and that's now
  914. 30:18in beta and it's doing really well but
  915. 30:20the initial way that we were sort of
  916. 30:22testing fit there
  917. 30:23we would go in and say like oh hey do
  918. 30:25you have you know a problem with with
  919. 30:27data cleanliness and the person would be
  920. 30:30like oh yeah yeah totally that's one of
  921. 30:31our big problems yeah I get cool cool
  922. 30:33like okay - that is and we might ask
  923. 30:38like two or three more questions but
  924. 30:39like that was sort of the depth but the
  925. 30:41actual level of question needs to be
  926. 30:42like okay well like what do you mean
  927. 30:44Didache like how do you do you currently
  928. 30:45invest in data cleanliness at all no
  929. 30:47like oh well yeah we actually you know
  930. 30:49we have a team of like six people who do
  931. 30:51data QA con all the time like oh well
  932. 30:54those data QA people like where are they
  933. 30:56based or they're based in LA Oh
  934. 30:57interesting so they have like real
  935. 31:00salaries and their knobs
  936. 31:01overseas they're like real like us
  937. 31:03salaries like yeah yeah okay so what
  938. 31:04like 80k a year 100k yeah yeah yeah
  939. 31:07that's about right so like okay so
  940. 31:08you're spending you know like 750 K a
  941. 31:09year for this data QA team and like tell
  942. 31:12me more about that process like what are
  943. 31:13they queueing exactly oh well they're
  944. 31:15clicking this button in the app and
  945. 31:16they're like well which button yeah and
  946. 31:18then are they like in the OP where they
  947. 31:20do when they find a bug and so we would
  948. 31:22ask like literally 45 minutes of
  949. 31:23questions like this and now we actually
  950. 31:26understand their problem and we
  951. 31:27understand what they're doing we
  952. 31:28understand where their cost centers are
  953. 31:30we understand how this thing and then
  954. 31:30we're like oh well what if we did a
  955. 31:31product that did X yeah which is exactly
  956. 31:34what they just explained to us for the
  957. 31:35previous 45 minutes and they're like
  958. 31:36that would be amazing they're like okay
  959. 31:39now wait now this is that was both sales
  960. 31:41qualification and discovery yeah which
  961. 31:44is a standard sales process but now it's
  962. 31:45being used for product development and
  963. 31:47that's such a good learning because
  964. 31:48people aren't gonna tell you no way I
  965. 31:51think a lot of people just get scared to
  966. 31:52like ask these questions totally the
  967. 31:54customers will tell you yeah especially
  968. 31:56if you if you take the champion part of
  969. 31:59medic their last one the C and you just
  970. 32:01start by asking like what's your vision
  971. 32:03for X thing that you do John will tell
  972. 32:05you you're like oh we our mission is
  973. 32:07similar because that's why you got the
  974. 32:08meeting in the first place that person
  975. 32:10is instantly aligned with you they'll
  976. 32:11talk for 45 minutes about their problems
  977. 32:13before you have to tell them anything
  978. 32:14yeah I think that's one of the things
  979. 32:16that most people don't realize it like
  980. 32:18many of the best salespeople don't talk
  981. 32:20that much the best salespeople at
  982. 32:22segment ask why to the point of
  983. 32:24uncomfortableness from everyone else on
  984. 32:26the team including myself yeah
  985. 32:28interesting yeah I wonder what the
  986. 32:30correlation is between sales and
  987. 32:31skepticism it's probably pretty high and
  988. 32:34people who are questioning things and I
  989. 32:36can see the angle yep hmm all right next
  990. 32:40Twitter question so Danny Pro first of
  991. 32:43all he says go Peter and his question is
  992. 32:46about a culture so he says what values
  993. 32:49and standards do you have in place for
  994. 32:51your team at segments and how do you
  995. 32:52actively build that culture into your
  996. 32:54company yes we have four values that
  997. 32:57segment that we're quite dedicated to
  998. 32:59the first is karma which is we want to
  999. 33:03have a positive impact on the world and
  1000. 33:04that manifests itself in a bunch of ways
  1001. 33:06one of those ways is we really care
  1002. 33:10about the customer having sort of
  1003. 33:12getting value out of our entire process
  1004. 33:14so you'll notice that all of our
  1005. 33:16marketing materials for example are
  1006. 33:17often like highly educational we have a
  1007. 33:19really high bar for what an education
  1008. 33:22piece of educational material looks like
  1009. 33:25even in the sales process we want to be
  1010. 33:27helpful if we're not the right fit will
  1011. 33:29tell you and sort of like refer you to
  1012. 33:30the right places
  1013. 33:32separately we really care about doing
  1014. 33:36the right thing by the end-user this is
  1015. 33:39still within karma and that's from like
  1016. 33:41a data privacy perspective so we're very
  1017. 33:42interested in helping companies
  1018. 33:43understand all of their own first-party
  1019. 33:45data so all their interactions with
  1020. 33:46their own customers within their four
  1021. 33:48walls we're super uninterested in
  1022. 33:50helping companies data broker data
  1023. 33:51between different companies schedule a
  1024. 33:53we call it data gossip it's gross we
  1025. 33:55don't want anything to do with it there
  1026. 33:56are plenty of other companies out there
  1027. 33:58that that have stuff like that it's
  1028. 34:00gonna go away and die eventually
  1029. 34:01so that's Karma we care a lot about that
  1030. 34:04the second one is tribe which is a
  1031. 34:07segment we're all there to support each
  1032. 34:09other we're all there to accomplish the
  1033. 34:10same thing and so what we expect is that
  1034. 34:14and what we value is that people really
  1035. 34:16support each other both when they may be
  1036. 34:19struggling with something but also
  1037. 34:21giving them giving them crit so really
  1038. 34:23try to reward folks who are who are
  1039. 34:25willing to go the extra mile to give
  1040. 34:26create when it may be hard could be
  1041. 34:28giving crit across teams or up several
  1042. 34:31levels or whatever that's that's really
  1043. 34:33something that we value hmm the third is
  1044. 34:36Drive much more self apparent we like to
  1045. 34:39get done we value people who are
  1046. 34:41getting done and the fourth is
  1047. 34:42focus which is not just sort of the
  1048. 34:44ability to sit down and get stuff done
  1049. 34:46but more power through something but
  1050. 34:48actually thinking carefully about
  1051. 34:50prioritization we've done a lot of
  1052. 34:52research around how to make the office
  1053. 34:54environment you can actually focus so
  1054. 34:56check out our blog we've we've written
  1055. 34:58about sort of sound decibel levels that
  1056. 35:00we've measured around the office and how
  1057. 35:01we've mitigated that did pretty well
  1058. 35:02that piece did pretty well read yeah and
  1059. 35:04and it was a surprising result for us to
  1060. 35:06discover the different parts of the
  1061. 35:07office had very different sound levels
  1062. 35:09that were not correlated with people
  1063. 35:10talking but were just correlated with
  1064. 35:12the sort of acoustic shape of the of the
  1065. 35:14office and so just moving people around
  1066. 35:16into different places helped a lot
  1067. 35:18depending on how much noise they were
  1068. 35:19willing to tolerate and sort of need it
  1069. 35:21in their role hmm so anyway those are
  1070. 35:24those four values we they are literally
  1071. 35:27the things we value and so we pushed
  1072. 35:28into all the places where you would
  1073. 35:30expect what you value to have an impact
  1074. 35:32so it's who gets highlighted at all
  1075. 35:33hands we have a citrus prize which is
  1076. 35:36someone who's living all the values
  1077. 35:38promotions hiring we have a strict
  1078. 35:40interview in the hiring process we have
  1079. 35:43performance we need a strict interview
  1080. 35:45sorry we have a culture interview where
  1081. 35:47we have these four values and waste
  1082. 35:48specific ways that we're going to test
  1083. 35:49for them when we run performance reviews
  1084. 35:53the performance of you is literally the
  1085. 35:55four values are usually like these this
  1086. 35:57is what we value and therefore it's what
  1087. 35:58we test and measure by and I think
  1088. 36:00ultimately it's that cycle of giving
  1089. 36:03feedback and measuring by it that is
  1090. 36:04what drives huh culture to stick and it
  1091. 36:07has this been something that like came
  1092. 36:09natural to you like building culture or
  1093. 36:11did you have to learn it I don't think
  1094. 36:12so I think we I think we learned it mmm
  1095. 36:17we got to about 25 people before we
  1096. 36:20realized that it was something that we
  1097. 36:22should write down and we went off site
  1098. 36:25four founders went off site and we tried
  1099. 36:27to synthesize the values out of what it
  1100. 36:30was that we really liked yeah there was
  1101. 36:33Apple Reddy happening and what it was
  1102. 36:35that we didn't like that we had a scene
  1103. 36:37already happening and not just among the
  1104. 36:39team but amongst ourselves to like what
  1105. 36:40were we not proud of that we had done
  1106. 36:42and what were we proud of that we had
  1107. 36:44and that that ultimately was what got
  1108. 36:46synthesized into those four values hmm
  1109. 36:48and so those were just interactions with
  1110. 36:50other people or like literally product
  1111. 36:52building no interactions with other
  1112. 36:53people and and interactions with
  1113. 36:55partners and customers and things that
  1114. 36:57we were proud of
  1115. 36:58mm-hmm that we wanted to see more of
  1116. 36:59right on so next question
  1117. 37:03Ashwin doke asks how is GDP are impacted
  1118. 37:07segments business model so GDP are for
  1119. 37:10those who don't know is a new EU
  1120. 37:12regulation which basically gives
  1121. 37:14end-users a lot of rights about the data
  1122. 37:17that's collected about them and first
  1123. 37:20off I think it's an awesome regulation
  1124. 37:22both as a consumer but also wearing my
  1125. 37:24segment hat it's interesting in that it
  1126. 37:28impacts the entire globe because if you
  1127. 37:30are storing data about in use it isn't
  1128. 37:32it doesn't matter what Roche diction you
  1129. 37:33run your company in you were still
  1130. 37:35responsible to do that for an EU citizen
  1131. 37:38the biggest impact
  1132. 37:40broadly on the
  1133. 37:42overall ecosystem is it really
  1134. 37:43negatively impacts third-party data and
  1135. 37:45third-party data brokers because they
  1136. 37:46have no real consent path to the user
  1137. 37:49for sharing and buying and selling the
  1138. 37:51data because we help companies purely
  1139. 37:54with their first party data it's not
  1140. 37:57like a existential threat to us in any
  1141. 37:59way and in fact it's something that
  1142. 38:00we're really sort of aligned with for
  1143. 38:03another reason as well which is because
  1144. 38:04we're routing the data out to all the
  1145. 38:06different places where people are using
  1146. 38:07it so we're routing it out to an
  1147. 38:09analytics tool to an email marketing
  1148. 38:10tool to a data warehouse to a CRM to a
  1149. 38:12help desk to ad conversion pixels if
  1150. 38:14that user shows up and says hey I want
  1151. 38:18you to delete me from your system well
  1152. 38:21it's actually like 20 systems for most
  1153. 38:23companies and we're already plugged into
  1154. 38:25those 20 systems so it's actually now a
  1155. 38:27feature of segments that we can go to
  1156. 38:29those 20 systems and delete whatever
  1157. 38:31user is requesting it and clean up that
  1158. 38:33record across all those different
  1159. 38:35systems so for us GDP r1 is like aligned
  1160. 38:38with our values philosophically - is
  1161. 38:41actually an exciting new feature sort of
  1162. 38:44requirement that we can support and a
  1163. 38:46sort of value that we can provide to our
  1164. 38:47customers so we're huge fans nice that
  1165. 38:50was a unsuspecting sir but uh people
  1166. 38:55have been stressed out about it my
  1167. 38:56friend makes Instapaper and they have a
  1168. 38:59big issue with it's a big problem in
  1169. 39:01publishing where they relying on
  1170. 39:02third-party damn yeah especially these
  1171. 39:04like little tiny products that are parts
  1172. 39:06of really big companies even they didn't
  1173. 39:08necessarily know and yeah another
  1174. 39:10everywhere cool all right so next
  1175. 39:14question and roopa cool asks any advice
  1176. 39:17that you have on asking for more money
  1177. 39:19than you're comfortable asking for this
  1178. 39:21is part of your startup school lecture
  1179. 39:23well I guess one of your sales reps was
  1180. 39:26forcing you to ask for more a lot more
  1181. 39:28yeah yeah we had a sales advisor who is
  1182. 39:30well I gotta back a little bit we were
  1183. 39:33initially selling our product for ten
  1184. 39:35dollars a month and you know hundred
  1185. 39:38twenty dollars a year and we brought on
  1186. 39:40the sales advisor and his first advice
  1187. 39:41was well you have to ask for $120,000 a
  1188. 39:43year and I was like that's a thousand
  1189. 39:45acts that's crazy so we were going to
  1190. 39:48the first sales meeting me and him and
  1191. 39:50is with a company called xamarin
  1192. 39:53and I've since told NAT this story which
  1193. 39:56he found amusing but now was the CEO of
  1194. 39:58xamarin and as we're walking up our
  1195. 40:01sales advisor says okay you have to ask
  1196. 40:03for a hundred and twenty K yeah him this
  1197. 40:04meeting and I was like that's the most
  1198. 40:06ridiculous thing I've ever heard I'm not
  1199. 40:07doing it and he's like well if you don't
  1200. 40:10do it then I quit as your sales advising
  1201. 40:11alright I guess I'm asking 420 K so we
  1202. 40:15go in we have to you know demo and
  1203. 40:16everything yeah and he says okay well
  1204. 40:19what's the price and it's 820 K and I
  1205. 40:21turned beet red and he says well how
  1206. 40:23about 12 K a year I said okay well how
  1207. 40:25about 18 and he's like okay fine
  1208. 40:28so from his perspective 85% off from my
  1209. 40:30perspective I got 150 X and it was a
  1210. 40:34successful negotiation I I think I think
  1211. 40:36it's really hard to offend people with
  1212. 40:38price at least if you're sitting in the
  1213. 40:39same room or or on the phone it's
  1214. 40:41probably not a good idea to share
  1215. 40:43pricing information via email if you do
  1216. 40:45that then it's really easy for them to
  1217. 40:46hang up but if you're on a phone call or
  1218. 40:48in person there's a bit of a social
  1219. 40:49contract to a continue engaging
  1220. 40:50particularly in person you can recover
  1221. 40:52so I would encourage you to not be
  1222. 40:55scared of offending someone with a high
  1223. 40:57price yeah but maybe just start in
  1224. 40:59person which is probably the most
  1225. 41:00uncomfortable place to do it but gives
  1226. 41:02you the most opportunity to recover
  1227. 41:03right and the thing is like if it
  1228. 41:05actually banished your business then
  1229. 41:06that's just what it costs yeah you're
  1230. 41:08gonna have well and you have no other
  1231. 41:10way of assessing the value yeah yeah and
  1232. 41:13then in fact what will happen is when
  1233. 41:15they say that's crazy then you say why
  1234. 41:17and then they'll explain to you how they
  1235. 41:19actually value the product and then you
  1236. 41:20say okay and you value it according to
  1237. 41:22their logic and you ask for that price
  1238. 41:24and how long did it take you well are
  1239. 41:26you charging them 120 now for sure yeah
  1240. 41:30we have customers to get way more value
  1241. 41:32than that out of it now yeah exactly and
  1242. 41:33so how many customers did it take you to
  1243. 41:35reach that six-figure amount a dozen at
  1244. 41:40most yeah so it was amazing yeah yeah
  1245. 41:42cool one carlos garza asks how did YC
  1246. 41:47help to get segment where it is right
  1247. 41:50now
  1248. 41:51YC was super helpful the most impactful
  1249. 41:55thing early on is just demo day you're
  1250. 41:58not going to find a bigger concentration
  1251. 41:59of investors who are excited about
  1252. 42:02investing in startups creates a
  1253. 42:03compelling event structures the timeline
  1254. 42:05in crowd
  1255. 42:06way helpful for a first round of
  1256. 42:07financing that can easily get strung out
  1257. 42:09and waste a lot of your time yeah that's
  1258. 42:11the first thing the second thing really
  1259. 42:13is the is the founder network there's
  1260. 42:16not only a lot of reasonably high
  1261. 42:18profile companies now that you can learn
  1262. 42:21from or companies that are sort of
  1263. 42:22farther ahead that you can learn from
  1264. 42:23now but there's companies at all stages
  1265. 42:26so there's almost always a group of
  1266. 42:28people in your area or in your market
  1267. 42:33that you can learn from and share from
  1268. 42:34so all the there are tons of little
  1269. 42:36groups that spring up you know like a
  1270. 42:37group of enterprise founders that are
  1271. 42:39all between like 70 and 100 people in
  1272. 42:42San Francisco and you can have dinner
  1273. 42:44once every two months yeah and that
  1274. 42:46becomes an incredible support group and
  1275. 42:49sort of way of learning about about
  1276. 42:52what's going on have you stayed in touch
  1277. 42:54with people from your batch a few yeah
  1278. 42:57exact same some good go right on yeah
  1279. 42:59I've heard of like these informal
  1280. 43:01founder meetups happening quite a lot
  1281. 43:03and again yeah it seems to be great and
  1282. 43:05uh it's a trusted network there's no
  1283. 43:08replacement for that yeah totally I
  1284. 43:10definitely didn't get that from college
  1285. 43:13all right Juan has another question in
  1286. 43:16the early stage what's the thin line
  1287. 43:18between ignoring a customer's suggested
  1288. 43:20feature or moving a customer's requested
  1289. 43:22feature to the core of your application
  1290. 43:25or product I think what I think what he
  1291. 43:28is trying to ask is basically like at
  1292. 43:30what point do you say like hey this
  1293. 43:33customer is requiring or asking for this
  1294. 43:35feature and we have to kind of hold the
  1295. 43:37line because we don't want to become a
  1296. 43:39custom dev shop
  1297. 43:40so should we integrate this or tell them
  1298. 43:44to you know find someone else the best
  1299. 43:46defense against that is having a clear
  1300. 43:49product vision for where your product is
  1301. 43:50going to go long term and if you have a
  1302. 43:52clear product vision for where it's
  1303. 43:54going long term it's a very simple
  1304. 43:55question of yeah is this thing in that
  1305. 43:58picture long-term or not and if it isn't
  1306. 44:00that picture long-term then you can
  1307. 44:03prioritize it to be sooner or later
  1308. 44:05depending on whether a customer is going
  1309. 44:07to pay for it or not yeah and if it's
  1310. 44:09not then it's not and you probably
  1311. 44:12shouldn't build it right yeah I think
  1312. 44:13that's like that
  1313. 44:14the infamous customers you don't want in
  1314. 44:16Scenario yeah where you just have to let
  1315. 44:18them go yes I
  1316. 44:19I guess the important thing is like
  1317. 44:20imagine the entire timeline of
  1318. 44:21everything you're ever gonna build feel
  1319. 44:23free to move things around we do this
  1320. 44:25all the time we move things around based
  1321. 44:26on like what customers actually want has
  1322. 44:27it's a reasonable signal of what's
  1323. 44:29actually more important yeah
  1324. 44:31but I wouldn't I wouldn't add major
  1325. 44:32things or remove major things from it
  1326. 44:34just based on one customer mm-hmm so
  1327. 44:36since you've done YC and it's been
  1328. 44:39several years now would have been the
  1329. 44:41biggest learnings since oh my gosh so
  1330. 44:43many a huge bucket or a huge area of
  1331. 44:47learning for me as weird as finance like
  1332. 44:50I mean I came from an aerospace
  1333. 44:51engineering background and then we were
  1334. 44:54doing software engineering for the first
  1335. 44:55couple of years and so you just start
  1336. 44:57completely unprepared for preferred the
  1337. 44:59like business side of of things so I've
  1338. 45:02learned a tremendous amount about about
  1339. 45:04finance as we've raised money and and
  1340. 45:06and you know learned to manage your
  1341. 45:08business with a piano and all those
  1342. 45:10things not that you should rush into it
  1343. 45:12but it's it's a huge area that can be
  1344. 45:14leveraged I think and you would have if
  1345. 45:17you were to do it again hired someone
  1346. 45:19earlier on who knew what they were doing
  1347. 45:20in on the finance side I actually think
  1348. 45:22we did a reasonably good job of that so
  1349. 45:23we hired a part-time CFO around town
  1350. 45:26that raised our series a so we were
  1351. 45:27about at about a million in revenue and
  1352. 45:29we were raised to fifteen million
  1353. 45:31dollars here he's a we had had a
  1354. 45:32bookkeeper up until that point but we
  1355. 45:34were like I feel like we should have
  1356. 45:36someone like you know point us as to
  1357. 45:38what we should be doing with the money
  1358. 45:39and maybe have like a plan or a model or
  1359. 45:42something so that was definitely the
  1360. 45:44right time to hire a part-time CFO and
  1361. 45:46Jeff berkland was super impactful over
  1362. 45:50the years
  1363. 45:50intro would have been the other big
  1364. 45:52important hires for you that I made like
  1365. 45:53a huge difference I'll have remembered
  1366. 45:56the exact hammer a whole bunch of people
  1367. 45:58but advisors maybe is an interesting
  1368. 46:00category sure part time CFO I think is
  1369. 46:02in that bucket
  1370. 46:03we had an HR advisor who was really
  1371. 46:05impactful we've invested more in HR than
  1372. 46:07most startups yeah of our size and I
  1373. 46:10think that was the right thing a lot of
  1374. 46:12startups like uber for example do not
  1375. 46:15and end up with I think really high
  1376. 46:16prices for this way I think it's a
  1377. 46:18challenge right because if you if you go
  1378. 46:20around and start googling like should I
  1379. 46:21hire CMO should I hire CFO should I hire
  1380. 46:24XYZ I think you can always find someone
  1381. 46:26strongly advocating for any particular
  1382. 46:29role but the challenge is like okay you
  1383. 46:32know you only have so much
  1384. 46:33and so much time and you can only find
  1385. 46:35so many great people so like where do
  1386. 46:37you and where and when do you decide to
  1387. 46:39hire those like optimal people for this
  1388. 46:42stage in your company right and so yeah
  1389. 46:44just kind of curious if there are any
  1390. 46:46big turning point moments for you it was
  1391. 46:50a huge turning point around 10 million
  1392. 46:52in revenue when we hired the first team
  1393. 46:54sort of execs to the team yeah one is
  1394. 46:57our VP of engineering and the other was
  1395. 46:58a VP of people they were the first
  1396. 47:00people who had previously been managers
  1397. 47:02and our VP of engineering had managed a
  1398. 47:05team of 150 at Dropbox so we went from
  1399. 47:07literally zero management experience
  1400. 47:09aside from what had been picked up along
  1401. 47:11the way going from zero to 50 people to
  1402. 47:13having someone who really knew it or two
  1403. 47:15people who really knew what they were
  1404. 47:16doing that was hugely impactful and we
  1405. 47:20should have figured out a way to do that
  1406. 47:21earlier 50 people in 10 million revenue
  1407. 47:23or whatever it was was way too late
  1408. 47:24yeah cool if you weren't working on a
  1409. 47:28segment right now do you have an idea of
  1410. 47:29what you would oh man I get to
  1411. 47:33occasionally invest in YC companies nice
  1412. 47:35and there's a lot of cool things
  1413. 47:38happening there I was blown away by this
  1414. 47:41like breadth of things that are
  1415. 47:43happening in the batch this batch there
  1416. 47:45was a really exciting company building
  1417. 47:46in space rocket engine and another that
  1418. 47:49was doing industrial inspections by
  1419. 47:51drone I just can't imagine a world where
  1420. 47:53we continued have people in harnesses
  1421. 47:55hanging off of wind turbines that's I
  1422. 47:57can't imagine that that continues for a
  1423. 47:58long time so that seems like an obvious
  1424. 48:01market opportunity so flying things I
  1425. 48:05have a background in aerospace
  1426. 48:07engineering cool man so if people want
  1427. 48:12to learn more about segments where
  1428. 48:14should they go if they want to learn
  1429. 48:15more about you yeah segments just go to
  1430. 48:17segment com or you can tweet at me on
  1431. 48:21twitter i'm ryan pique rei NPK okay cool
  1432. 48:25yeah we'll link it all up all right
  1433. 48:26thanks man cool thank you
  1434. 48:35you

About this transcript

This page contains the full transcript of Peter Reinhardt on Finding Product Market Fit at Segment by Y Combinator, generated from the public captions YouTube serves with the video. The transcript has 10,257 words across 1,434 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.