00:00hi everyone welcome to the a six in Z
00:02podcast I am sonal and today's episode
00:04is all about network effects and it's
00:06based on an event we did that was a bit
00:08self based on a set of slides that we
00:09released that you can find online at bit
00:12dot ly slash network effects a 16 Z
00:15welcome to our network effects event
00:17before we get started let me actually
00:18just quickly introduce our guests so you
00:20probably know because you came here
00:22knowing who they are but just to quickly
00:23tell you about who they are
00:24Anu is an electrical engineer by
00:26training she used to work at Qualcomm
00:28where she built the video streaming
00:31solution there and then later she worked
00:33at BCG and she's now a partner in our
00:35deal and investing team and she covers
00:38all things marketplaces with Jeff Jeff's
00:40a general partner here he was previously
00:42the CEO at Open Table which is why the
00:44network effects team is particularly
00:45strong here and president at PayPal and
00:48SVP idea for that and his board seats
00:51include a lot of our companies that have
00:52Network effects like Airbnb and many
00:55more it's actually funny because this
00:56whole deck started off as the result of
00:59an internal debate and you know we're
01:00pretty opinionated as a culture and
01:02there's a lot of strong opinions
01:03floating around but I think it's kind of
01:06funny because this is probably one of
01:07the most internal contentious debates
01:09we've had around what is a network
01:11effect and I think it's because it's one
01:13of those concepts that's really easy to
01:15learn but difficult to master so on that
01:18note let's just start off by defining
01:19what's a network effect yes simply put I
01:22think the simple definition is as more
01:24users join the platform it's more
01:26valuable to existing users so it's
01:29really as simple as that right and it
01:30actually started with hardware so think
01:33of the telephone in the early nineteen
01:34hundred's the more people that had
01:36telephony you could actually make a
01:38phone call so there was a network effect
01:39if more people's it was valuable to you
01:41because if more people had telephone you
01:43could reach more people so it sounds
01:45very simple at the onset but I think the
01:48biggest misconception people have in
01:50fact we internally have debated about is
01:52you know growth versus network the fact
01:55so you know growth is different because
01:57in especially as a as a start-up in your
01:59early days you're growing really fast
02:01that's growth but network effect is
02:04really value how valuable is your
02:06service to the users as more users join
02:08the platform so to be clear though it's
02:11growth and value but what's the
02:13difference between that value
02:14engagement isn't enough to just say
02:15we've got a lot of engagement and
02:16therefore we have a network effect how
02:18do you actually know there is a network
02:20effect I mean we brawl over this concept
02:22because the easy observation is
02:24businesses with network effects
02:26typically are very defensible and author
02:28prone to monopolies yeah and we're
02:30trying to get in early in the company so
02:32that's not completely apparent when
02:34we're investing so you know you're
02:36brought you know we'll have people
02:37standing on the table there's a network
02:39effect here's why and no there's not and
02:41here's why and so I mean it is one of
02:43the key drivers about the investment
02:45decision on a whole bunch of our
02:46consumer side businesses and some of the
02:48business side businesses no I mean we
02:50actually told it I put a slide up
02:51tomorrow please last year network
02:52effects equals moats and you know the if
02:55you if you can achieve it you're
02:57defensible you're less susceptible to
02:59price pressure you're less susceptible
03:00to pressure of consumer acquisition
03:02economics and things like that because
03:04if your service is more valuable than
03:05the other you're gonna create that value
03:07through throughout the entire process
03:09right and it's I mean things valuable
03:11for users which is really key but that
03:13is actually where I think we get tripped
03:14up on what a network of that really is
03:16internally because I think we sometimes
03:17confused oh my god the product has viral
03:19growth there's that a network component
03:21how do you tease those apart yeah and I
03:23think that it's it's hard to do it
03:25especially when you're a hyper growth
03:26startup which is why you know we have
03:29our own internal debates you know
03:30whether the company has network effect
03:32or not I think while growth is speed of
03:35an option think of speed as viral growth
03:37and usually if you have viral growth it
03:39sort of implies that you're not spending
03:41anything in customer acquisition cost
03:43right so Facebook is a great example for
03:46the longest time even to this day they
03:48hardly spend money on acquiring users
03:50right and especially in the initial
03:52years they spend $0 it was growing
03:54widely so how do you actually decipher
03:56whether Facebook as a network effect
03:58because they're ma use and da use but
04:01growing quite fast I usually in the
04:02Memphis monthly actives and the daily
04:05but the one metric and Mark Zuckerberg
04:07has done a great job of saying this in
04:09so many interviews the one metric that
04:11he really focused on in the first two
04:13years was retention and the way he
04:16measured retention was daily actives by
04:18monthly actives and if you see their
04:20chart in the first 18 months it kept
04:23going from 52% to 55% 57% dad showed
04:27that people really stayed in the network
04:29that was a measure of the network effect
04:31and not just the viral growth in their
04:33users and by the way just think about
04:35the level of that number I was talking
04:37to mark in 2006 he reached out and asked
04:41to get together and we were chatting
04:42about the business and um and he got a
04:44text at the time ding does and he reads
04:47it he smiles I go what what made you
04:49smile he goes oh this is it we get our
04:50daily metrics via text I go what metric
04:53he goes um it was this metric he goes
04:55how many of our cumulative historic
04:57users do you think we're active
04:59yesterday and I'm like gum just came out
05:01of eBay and PayPal like two three
05:03percent maybe you think I got bold and
05:05you know six percent and he goes fifty
05:08three and you know fifty three percent
05:10of everyone who had ever signed up to
05:12the service was active the day before
05:14and and that by the way has only grown
05:17ads you know the they're installed base
05:19age it but gets bigger in ages that
05:21numbers grown and you're just kind of
05:23like okay that's a network and I think
05:25the thing that's most interesting about
05:26this whole discussion is it's not by
05:29accident but it feels like it's by
05:30accident cuz you hear these narratives
05:32you know like oh my god Facebook reach
05:33this number and it was taking off at
05:35campuses everywhere and in reality it's
05:38something people actually work at and I
05:40think people here want to hear about how
05:43because I mean what are some of the
05:45things that people do like how do they
05:46start I mean face looks a unique case
05:48maybe we can start with Facebook and
05:50what other examples but what are some of
05:51the things that people do to get to that
05:52network before there's an effect yeah I
05:54think the the first question you want to
05:56ask us if you have a product is how do
05:59you bootstrap what are some of the early
06:01growth hacks that you want to do to
06:03build the user base and then the second
06:05step is you're sort of testing your
06:06product market fit as you grow the users
06:08and then what you have to quickly hone
06:11on what is the aha moment that gets the
06:14user back to continue to use my product
06:17which actually drives the network effect
06:19and so in Facebook's case contrary to
06:21you know what most people think they
06:23actually went in a very clustered
06:26approach so the the first was they first
06:28launched in Howard and you could argue
06:30that they could easily expand to all the
06:32schools right but I think their goal was
06:34they needed to get at least 80% of
06:36Harvard signed up on Facebook and sort
06:41high bar of 50% Plus engagement so daily
06:44they wanted more than 50% to come back
06:46daily which is a very stretch goal and
06:48in fact if you go back to some of the
06:50early investors one of the criticism of
06:51Facebook was well you didn't grow fast
06:53enough I mean it's a different story
06:55that they did penetrate almost 80% of
06:57the colleges in 18 months and had a high
07:00retention but they did it in a very
07:02strategic way until they hit that high
07:04bar on Harvard they did not roll it out
07:06to Stanford and the discipline that they
07:08got is a result where they were tweaking
07:10the product to get the right fit so
07:13simple things like the relationship
07:14status they realized actually drove
07:17engagement really high because people
07:19felt uncomfortable asking someone else
07:21like what is your relationship status
07:22while you could go to Facebook and
07:24figure that out second thing that they
07:26did was they realized that if someone
07:28was connected to 10 friends in 14 days
07:31they were much likely to return to
07:34Facebook when they realized that they
07:36started we know every new user that they
07:38joined they worked on making sure that
07:40they got to 10 friends within 14 days
07:42they would actually bludgeon the crap
07:44out of you to get you to 10 friends it
07:46would be like you'd show up in like do
07:48you know these 12 people so one of the
07:50best hacks Mark had was um it's well
07:53known he got in trouble for it he hacked
07:54the directories in each of the houses
07:56and so he pre-populated everyone on the
07:59Harvard campus into Facebook and then
08:01let people claim their identities but
08:03you could then friend it was populated
08:06when the first person showed up it might
08:08not have had the same utility but it was
08:09populated versus being the first person
08:11to show up it's like you know I'm in a
08:14stadium by myself and so I mean I think
08:16he did end up in front of the president
08:18of the University for that hack but the
08:20hack populated and they you know brought
08:22a bunch of utility and so you would use
08:24the word growth hack virtually every
08:26marketplace we can think of did some had
08:29some hack or hacks that enabled them to
08:32solve the impossible chicken and egg
08:34problem the chicken and egg problem is
08:35our gated open table it's only useful
08:37the restaurants if consumers are there
08:38and it's only useful the consumers of
08:41restaurants were there where do you
08:42start and so you know there's a whole
08:44bunch of different hacks at different
08:46entrepreneurs have used to try to
08:49address that how do you get the flywheel
08:51spinning yeah well what are some of the
08:53other ways that you see people stumble
08:55because sometimes people discover the
08:57hack by accident or is it deliberate I
08:58mean how do they really I mean do you
09:00actually wake up one day and say okay
09:01I'm gonna figure out the Hat for this
09:02market wins oh yes when we're talking
09:05entrepreneurs you're looking for the
09:06theory what's your theory on how you're
09:08gonna bootstrap yeah how are you gonna
09:09hit the flywheel spinning how you solve
09:11the chicken the egg problem which side
09:12are you gonna start on so opentable
09:14ended up starting completely in the
09:16restaurant side because there was zero
09:18utility to the consumer until you had a
09:20selection of restaurants and so how do
09:23you get a restaurant to adopt it when no
09:24consumers are using it so you you know
09:27open tables case is what Chris Dixon you
09:29know come for the tools stay for the
09:30network they build a suite of tools that
09:32they charge $200 for and laborious Lee
09:35rolled out one restaurant at a time
09:37throughout you know the country that had
09:40enough utility that a restaurant was
09:42saying okay I'll adopt that in the
09:44absence of a network and as they slowly
09:46built that base of restaurants and I
09:48mean slowly I mean they they a good
09:50salesperson would do three or four new
09:52restaurants a month and you know there's
09:54yeah there's 2,000 restaurants in San
09:56Francisco that is slow and then at there
09:59was a point where there was enough
10:01restaurants on the system it was a
10:02subset of the restaurants in the system
10:04but the utility of making an online
10:06order was so great that they'd say I
10:08ignore the fact that you don't have any
10:10percent of restaurants because the 20
10:11percent you do have has a huge amount of
10:13utility as soon as the consumer started
10:16booking any reservations through open
10:18table the system exploded because all of
10:20a sudden the restaurant tour's saying oh
10:21they're they're filling the seats so
10:23that case was come for the tools but
10:25then stay for the network and so the
10:27sales rep that initially in San
10:30Francisco and New York sold three
10:31restaurants a month the same reps in
10:34most cases seven eight years later were
10:36selling twenty I mean there's so they're
10:38ripping off one a day because all of a
10:39sudden I've got all these tools they're
10:41kind of interesting I'm gonna fill your
10:42rest and you know each person I'm gonna
10:44put in the restaurants going to bring 50
10:45bucks and your margins on that 50 bucks
10:48is 30% so you're earning 15 bucks you're
10:51paying us a dollar you know all day long
10:53and so that didn't you know it did the
10:54flywheel goes by the way that was city
10:56by city so the fact that you had
10:58sufficient restaurants in San Francisco
11:00meant nothing in Miami so you started
11:03again in Miami and in Tokyo and in you
11:08it's it's a slow roll up do you actually
11:10advise people then to like think about
11:12their strategy differently you know to
11:13go like I would think people would stall
11:15if you you can only get a local but not
11:17get global like do you how do you sort
11:18of balance I mean I feel there's a
11:20pressure these days to like global fast
11:22how do you make those trade-offs and
11:23decision differently is have different
11:25characters that the local ones are
11:26really hard and really slow and but if
11:28you build them they can be really
11:29valuable and really tough to overcome so
11:32you know it is virtually impossible to
11:35to displace opentable at this point is
11:39the strongest never I've worked with
11:40some great network effects businesses
11:41you know eBay PayPal Airbnb you know a
11:44Pinterest it's virtually impossible to
11:46displace open table from a restaurant
11:49right now but they had to go market by
11:51market to do that I mean they're the
11:52eBay was instantly national almost
11:55global because if you were a take
11:57there's ten antique collectors on the
11:59service didn't matter ones in Michigan
12:01ones in Miami ones in Berlin if they're
12:03collecting the same antique there's
12:05utility there and so that one was able
12:06to spin up much more seamlessly the the
12:09notion of the local marketplace which
12:11said I know some of you in here are
12:13doing yeah it has that extra extra
12:17complexity I also think the question
12:19that you have to ask depends on what
12:21type of business you're operating I
12:23think initially the hunch is always
12:26let's go global get as many users but I
12:28think the question you have to ask is
12:30can you truly measure your product
12:32market effect if you are not close to
12:34those consumers or you don't control the
12:37zones you're in so for example whatsapp
12:39right it went global really fast but it
12:41didn't start that way
12:42in fact the yarn used to hang out a lot
12:46at the Russian community in West San
12:48Jose and his the first feature of
12:51whatsup was literally just to show
12:52status saying you know he just said why
12:55can't there be just a status message on
12:56the phone that shows that I'm at the gym
12:58so that someone won't call me in so he
13:00rolled out that at the west and he took
13:02me and his focus group was literally the
13:04fifty people that he used to hang out
13:06with over dinner on Fridays in in that
13:09community and they would give him
13:11feedback on how they're using the
13:12product quickly he realized that they
13:14were using it to send messages to their
13:17family in Russia and he was like well
13:20that was not why how I designed
13:22this app for but he sort of built
13:25features around that to make it more
13:26seamless and that's really how whatsapp
13:28started so contour anyway again
13:29counterintuitive because you see their
13:31growth and you're like wow they went
13:32global really fast which is true unlike
13:34you know like say Airbnb which took time
13:37to build critical mass but even then
13:39they actually in the initial days were
13:41so focused in their small control groups
13:44to make sure that they got the right
13:45product market fit both of those
13:47examples you guys shared with Facebook
13:49open table those are familiar behaviors
13:51for people what about a case like Airbnb
13:54in the early days where there is no such
13:56thing as the so-called sharing economy
13:58like it's actually a completely foreign
14:00concept to say to someone yeah you can
14:03stay in my in my room but that's just
14:04it's crazy at the time how do you think
14:07about like what did they do there and
14:08how do you think about other businesses
14:10that have that kind of a situation or
14:11there anything really new and they
14:12trying to build a network effect yeah I
14:14think with Airbnb and it's a classic
14:18case of marketplace but strong network
14:20effects but they did not have any viral
14:22growth right to that because forget
14:24about renting someone's home you would
14:26not let I mean something more personal
14:27than your phone is your home they see
14:29your bet they see your bedroom they see
14:30your restroom you're not gonna let a
14:32stranger in so imagine how a company
14:34like that could take off so their early
14:37days was really really hard but their
14:40biggest value proposition it was you
14:43know that they could offer homes at much
14:46cheaper rates especially when they were
14:48big conferences or events in the city
14:50and all the other hotels are sold out I
14:52mean that's literally how the concept
14:54was born right they rented their own
14:55place literally rented their own air bed
14:58and they said well why can't we spin
15:00this into a company but you know they
15:02did that for their own home for few
15:04times but it did not take off instantly
15:06so they really had to use traditional
15:08marketing and good design to sort of
15:12build that growth hack so simple things
15:13they started doing where was they
15:15started targeting all events in various
15:18places and they would make specific
15:19advertising in those areas saying oh is
15:22the hotel sold out why don't you try air
15:25in fact even you know in the early days
15:28the the story that goes around is that
15:30they you know they were they found it
15:31very difficult to raise so they actually
15:34remarketed their cereal
15:35boxes Obama and McCain cereal boxes and
15:38they actually used that money to sort of
15:40bootstrap the company so it goes to
15:42shelter during the elections during the
15:43election cycle so it goes to show you
15:45know the different steps that they
15:47needed to take to bootstrap it almost
15:48took thirty six months for them to build
15:50critical mass and see network effect so
15:52it was not easy but once you build that
15:55critical mass you have huge defense
15:56ability so what one of the founders
15:58stories so entertaining and so and
16:00almost always so patently false is the
16:02overnight success i dinner with brian
16:05last night he's been at Airbnb nine
16:06years I mean you know it's so you know
16:09it's you know in the last the last three
16:11or four are gone really well you know
16:12that first I'm Ben Silbermann at
16:15nothing happened there for two three
16:16years nothing you just he just kept
16:19working for product market fit you know
16:21with his dozens of users and then a
16:23hundred years or whatever and then he
16:25got it and you know figured out the hack
16:27to get critical mass of content that
16:29then attracted the users and it goes
16:31straight up but this this overnight
16:33success is is rare in our in our
16:35experience so I think we should drill
16:37down then because if I'm in this hunter
16:39shoes how do you know to hang in there
16:41like what are the signs the metrics like
16:42okay I'm gonna just stick around for
16:44another ten years or I have a sign here
16:47so how do we tell let's walk through
16:49some case studies just to have the data
16:51because we use a lot of words I mean
16:52let's look at the numbers so this is the
16:55first example we're gonna share is
16:57actually Facebook and the reason we used
16:59Facebook was to tease out the difference
17:01between viral growth and network effect
17:03right so this is the chart usually
17:05startups sure at the start like hey
17:07we're growing really fast you can see
17:10that we reached you know eight hundred
17:12and eight hundred million plus I may use
17:13in a very short timeframe so the chart
17:16looks great but this is what we and and
17:18Facebook did this with pretty much zero
17:19customer acquisition cost why they
17:21didn't spanner died but this is what we
17:23would call a speed of an option which is
17:24growth this does not tell us whether
17:27there is the network effect what tells
17:29us whether there is a network effect is
17:31actually this which is even as this kept
17:34growing their retention so this chart
17:37shows daily actives divided by monthly
17:40actives and you can see that it kept
17:42growing from 45% to 57% and Facebook was
17:46actually one of the first
17:47platforms to actually
17:49do that there's because usually you've
17:51never heard where or you increase users
17:53do you actually increase the usage that
17:54was not the norm and so that's the sign
17:57of network effect and so you and this
17:59metric is different depending on what
18:01type of company you are for a
18:02marketplace like for Airbnb you will
18:05actually look at the number of guests
18:07that booked rooms and how that trended
18:09over time right that would be the sign
18:11of the network effect and we wanted to
18:13show the other example Angry Birds they
18:15have a similar curve they did you know
18:16lots of downloads and if you looked at
18:18the user graph it was high so they had
18:20viral growth because people were playing
18:22the game and they really liked it and
18:24they talked about it so they may have
18:25acquired Aldus users with zero customer
18:28acquisition costs but did they really
18:30have a network effect you know not
18:33really because if I'm playing the game
18:35and you're playing the game that doesn't
18:36necessarily mean I'm getting more value
18:38just because you are playing the game as
18:41well and we just sort of fly in the
18:42gaming world you do see where you know
18:44new games come like the next Angry Birds
18:46was you know candy crush and so on right
18:48so I think the key is really to tease
18:50out is this viral growth or is this
18:52network effect the last example our
18:55chair is with medium it's one of a
18:58recent portfolio companies and so we
19:00wanted to include one of the case
19:02studies which was slightly early because
19:03it's not easy to you know sort of see
19:06signs of network effects so medium
19:08mediums use a growth chart is quite good
19:10you know they're adding a lot of unique
19:12visitors and they're acquiring them with
19:13pretty much zero customer acquisition
19:15costs but how do you actually say
19:17whether there is a sign of network
19:20effect so what we wanted to find out is
19:23well how are people finding medium
19:24articles because if if I discovered
19:27medium through Twitter or Facebook then
19:29if I really liked it that much am i
19:31coming back to medium directly or am I
19:33still finding those articles through
19:35other sites so one of the things they
19:37explicitly measure is how much you know
19:40people spend a lot of time reading
19:41articles on medium but how much of that
19:42came directly from medium versus through
19:46social which is Facebook or Twitter or
19:48even other channels and you can see
19:50especially for the non-vital post which
19:52is what they call the tail because the
19:54top widel post will always be viral
19:56people will refer it everywhere and it
19:58goes but for the non viral post medium
20:00is actually the biggest source for the
20:03Vick which was you know early indicators
20:05that they are able to match it's like a
20:07marketplace they're able to match the
20:08readers to the right writers and so
20:11those posts are also getting attention
20:12and it's happening directly on medium so
20:15the question you have to ask for network
20:17effect is depending on your business or
20:19company that you're building what is
20:21that aha moment that gets the user back
20:24to your site or your app without much
20:27effort I mean the other thing is I mean
20:30these are all macro reads you know what
20:33we're trying to make that call you're
20:34trying to make that call early we're
20:36trying to make that call early on your
20:37companies and so you're trying to tease
20:40out what might suggest that these macro
20:44things will kick in overtime and
20:45typically that's micro things and so in
20:47opentable the entirety of my diligence
20:49on the company was show me San Francisco
20:52you've been in San Francisco the longest
20:53show me how the key metrics are going in
20:56San Francisco and so it turned out sales
20:58productivity exploded over time the
21:00number of reservations per restaurant
21:02exploded over time the percentage of
21:04those reservations that came from the
21:07Open Table website versus the restaurant
21:09website changed dramatically over time
21:11the you know the I mean just everything
21:13of restaurants went down over time
21:16because you know the company's providing
21:17more value so it was pretty obvious
21:19early on an open table in when you took
21:22the micro thing you know that ok San
21:25Francisco is working and that gave
21:26before you even got there the management
21:28team and the board the confidence okay
21:30we're gonna step on the gas and start
21:32opening tons of new markets even though
21:34what they took years to turn profitable
21:37they said ok this case if you build it
21:39they will actually come now different
21:42you know so we go into markets you know
21:44Zuckerberg is pretty obvious on Harvard
21:46it was working it wasn't necessarily
21:48obvious the business was working but if
21:50you went into your you know your petri
21:53dish or your test tube and said okay I
21:55made it work that I really work in the
21:56Russian community and San Jose their
21:59behavior is showing increasing returns
22:01to scale and so the best advice I have
22:04is try to you know and typically both to
22:07launch and then the measure is run the
22:09small if you can run the smaller
22:11experiment it makes you know the
22:13validate that it's working for you if
22:15it's working tell us
22:16and then I did the I mean that really is
22:19how you do it in eBay with the early
22:21thing was collectibles collectibles just
22:23exploded on the platform because the
22:25needs of that community trading in the
22:28analog world were so horribly not met
22:31and turns out the the venture guy who
22:33made the investment at eBay was Bob
22:35Kegel at a benchmark capital Bob
22:37happened to come from Michigan and
22:39there's a craftsman in Michigan who does
22:42really beautiful and I'll remember that
22:45it decoy ducks for for hunting it's the
22:48Oscar Peterson was the craftsmen's name
22:50because it's also the name of the
22:51pianist that's the only reason I
22:52remember it Bob went on this black and
22:54white website that she was also had the
22:58Ebola virus and area 51 on it
23:01everything's on here homepage and search
23:04for Oscar Peterson duck decoys and found
23:07like five and he goes I spend my summer
23:10driving around Michigan looking for
23:13these I went online in their five here
23:15and so I mean he was in that literally
23:18what's put him in the deal but you know
23:20it's those little early you know you're
23:22seeking those early signs that you know
23:25the behavior is happening at a small
23:27scale if it's having a small scale you
23:29probably can replicate it on the larger
23:30scale it's not happening on the small
23:32scale you have to yeah you know you got
23:34to figure out how do you get product
23:36market fit and that's where if you're
23:38thinking of going to work at Yahoo you
23:41know for a steady paycheck you know that
23:43if you can't get any of the small
23:46experiments working that's when you
23:48might wanna start thinking about I want
23:49to push back on one thing though which
23:51is if you're in San Francisco it feels
23:55like what happens in Sierras is not
23:58gonna happen everywhere else it doesn't
24:00stay in San Francisco exactly exactly so
24:03if you're building a business in Silicon
24:04Valley where you do we are in a little
24:06bubble in a lot of ways like people are
24:08doing very unique things and I think
24:10that's why I'm glad you brought up the
24:11example of like the Michigan example
24:13Pinterest yeah I'd love to hear your
24:14thoughts on what that means for figuring
24:16out whether there is a network attack so
24:17did to amend it a little bit in local
24:19businesses we typically look that they
24:21can replicate it in a market outside of
24:24the market they live in because that
24:26kind of suggests scalability so if
24:28someone in San Francisco
24:30you know a great example and I don't
24:32think he would upset if I use it um
24:33cherry cherry worked great I mean the
24:37cherry car washers it was on demand car
24:41washing they were up and down this strip
24:44outside of Sand Hill Road every single
24:46day there'd be like three cherry guys
24:47washing VC's cars can I see this working
24:53in Minneapolis in the winter and you
24:55know you're just kind of like oh so I
24:57mean that was an example okay we want to
24:59see it you know replicate outside of an
25:01environment where it's you know either
25:03uniquely economically affluent or you
25:06know highly conveniens focused and
25:08things on those but yes we we typically
25:11look in local businesses they can
25:12replicate it outside of the initial
25:14market because that then talks of
25:16scalability right I mean the city a city
25:18is the original network so there is a
25:20certain density there well speaking of
25:22the trend side of things we talked about
25:24the sharing economy so let's talk about
25:26examples like ride-sharing and
25:28crowdfunding and those are new
25:29technologies I mean they haven't been
25:31around before or maybe they've happened
25:32and now smartphones are here to sort of
25:34give us new behaviors around them do
25:36those trends have never defied yeah so I
25:39think I would combine you know I would
25:42call ride-sharing even food delivery all
25:44of those I would put them like the
25:46on-demand category in general I think
25:49it's a new phenomenon after mobile we
25:50just made it really easy for consumers
25:52so the way to think of it is so if you
25:55look at right sharing services right
25:57it's you know 10 times better than
25:59taking a taxicab the product is way
26:01better and consumers use it but it's a
26:03city by city rollout to a large extent
26:06opentable however the difference I would
26:09say in a ride-sharing service it's a
26:11point-to-point service in opentable you
26:13offered the tools to the restaurants and
26:15then you know that's sort of you know
26:17the dude restaurants used it for the
26:18tool and then you brought in the network
26:20which is the diners and you know you
26:22match sort of the two sides in
26:23ride-sharing it's really point-to-point
26:25so I would say the general ride-sharing
26:27we think it's more supply-side economies
26:29of scale and what that means is the more
26:32drivers you have on the platform you can
26:35make sure that you get a good quality
26:37driver within five minutes but that's
26:39what its tense right so it's like it's
26:41almost like the Amazon first party if
26:42you wanted to think of it we
26:44just like it's still a good business we
26:46just call it as different but just you
26:48know supplies that economies of scale
26:49which is scale economy where's there's a
26:51network effect however if you look at
26:54uberpool I think those could have
26:56Network effects because think of this
26:58you're a writer you want more writers
27:00using those because then you can share
27:03the right to San Francisco from Palo
27:05Alto and therefore have a cheaper write
27:06right and you want more writers
27:08everyplace maybe you travel to New York
27:11and you want people to be using left
27:12flying and over pulled air as well so I
27:14think there is a subtle nuance it is
27:16what our view is I mean talk about hacks
27:19I mean ubers early hack was they're
27:22gonna subsidize enough to cars to drive
27:24around a city to provide a minimum level
27:27service that the consumer found
27:28attractive and then as consumers find it
27:30attractive and start booking more sites
27:32they have to subsidize less and
27:33ultimately the subsidize disappears in
27:35the city instacart hack was you can
27:38order from any grocery and they'll go
27:39get it to you they'll mark the crap out
27:41of it and but early on they'll be
27:43showing those grocers said hey you can
27:45get some pretty good demand for people
27:46who can who want to have groceries
27:48delivered they went back to the same
27:50grocers and are cutting deals and now I
27:52believe well over half of their entire
27:54deliveries are not marked up there
27:56through deals with grocers like Whole
27:58Foods but the active jumpstart it was
28:01okay I don't have time I don't know the
28:03time or credibility to do those deals
28:05with the grocers until they experience
28:07the flow of order so the hack was okay
28:09I'm gonna mark it up and then I can
28:11afford to do it for a while and then as
28:13quickly as I can I'm gonna try to
28:14leverage the volumes into a into deals
28:16and then let me provide a better
28:18consumer service so every business we
28:20see has an attempted hack yeah if you
28:24come in and say oh I don't really know
28:25how this you know I'm gonna fit through
28:27the chicken and egg you know not the
28:29best answer you have to try to figure
28:30out how can I get this going which side
28:32starts what is this you know and you
28:34know the deck lays out a whole bunch of
28:36different approaches that different
28:38companies have used to try to solve that
28:40one fundamental problem is in a two
28:42sided network who shows up first
28:44besides this network effects deck you
28:46guys participated in a post we wrote
28:47around 16 metrics and 16 more startup
28:50metrics I think it's worth bringing that
28:51up because you're talking about
28:52discounts and subsidies as part of a
28:55model to create that hack and then the
28:56other thing is in that
28:57same post we talked about the difference
28:59between paid versus organic marketing
29:01and the value of those so if I'm doing a
29:04network effects business and you see as
29:06an investor like okay well you've done a
29:07lot of paid marketing versus organic
29:09marketing to get this network is that a
29:11bad thing is that a good thing I mean
29:12how do you tease it apart I think that
29:14it's really I mean it's there is no
29:16definite of answer it really depends so
29:18I'll take the example of Airbnb you know
29:21it was very difficult to build the
29:23critical mass initially and it was a
29:24strange concept for them but on the
29:26demand side they were able to hack with
29:28traditional marketing and you know they
29:31were they had really great hacks around
29:33how they bootstrap demand so they were
29:35really not spending much on demand you
29:37know it's it's contrary to everything we
29:39know because travel people travel only
29:40twice a year how do I do it but they
29:41were really good at doing that in the
29:43initial days but supply was really hard
29:45you know how do you get a host
29:47comfortable to list their home how do i
29:50target the hose so they did end up
29:53spending dollars on it so it's not that
29:55we are you know it's not it sometimes
29:56you may have to figure out you know
29:58there are lots of papers on this on
30:00marketplaces but just what is your money
30:02side and what's your subsidy side I
30:03think for Airbnb that supply was really
30:06hard so they were they spent dollars
30:07right from the early days to acquire the
30:10host but in those cases you have to ask
30:12yourself what is the return on
30:13investment if I am paying to acquire the
30:16user right so what is my order value per
30:19transaction and what's my lifetime value
30:21if you do have good ROI economics like
30:24the CAC versus LTV then I just take cat
30:26versus LTV so let me just my position
30:28cost versus the lifetime value that
30:30you're getting from the customer then
30:32you can spend because there are certain
30:34sectors where you have to travel is one
30:36where you have to because the velocity
30:37of use cases really no but you just have
30:40to be stunned you know the thing that
30:42you don't want happening is where you're
30:44paying lots of marketing dollars to
30:47acquire users and then when you sort of
30:49dial down your marketing dollars the
30:51growth also slows down because that
30:53means you've been spending a lot of
30:54money to acquire and drive the growth
30:56but the company itself or the product
30:58itself is not sustaining that so that's
31:00the thing we really try to tease out is
31:02if you're doing paid marketing is you
31:05know why are they doing paid marketing
31:06what's the ROI a return on investment
31:09from that marketing and is
31:10you know is it truly showing a retention
31:12one observation a lot of the best
31:14networks that our platform here aren't
31:16spending anything on acquisition I mean
31:18it's just it's kind of an interesting
31:19thing in marketplace Facebook know
31:21whatsapp know medium no you know just
31:24open table know just they're not
31:26spending on acquisition the model is
31:28delivering it if they are spending an
31:30acquisition eBay I was the biggest
31:32advertiser in the world on digital
31:34advertiser in the world for a number of
31:35years they were spending but then what
31:37happened is the more they spent the more
31:40valuable the service was to the user so
31:42it was a component of the pl but those
31:45users demonstrated network behavior and
31:47the the Contra example is okay an
31:50e-commerce business you know you can
31:51pour money on it you can grow like crazy
31:53I've been an investor in a few of them
31:55that actually are no longer around that
31:56grew like crazy but you know the only
31:59reason you're growing is the marketing
32:00spend and when you know a new said when
32:03you pull back on it they don't come
32:05anymore there is no network effect there
32:07it was just a growth effect and so they
32:08you know he is how do you tease out that
32:11so it's not that you don't spend it's
32:13surprising how a lot of the most
32:15valuable network businesses didn't spend
32:18virtually anything they they hacked they
32:21figured out the hack well I'll do one
32:23last question then we'll open it up to
32:24everyone for questions it's a question
32:26that probably you know you're
32:27entrepreneurs hate getting from their
32:29investors which is or maybe they want to
32:32know which is when to monetize and how
32:33do you figure that out I should say that
32:35as investors you know we're not going to
32:37push anybody to monetize because I know
32:39it is sort of a debate everybody has
32:41it's more you should think about when to
32:44monetize and what types of monetization
32:46is do you have in the plan but we've
32:49seen companies at all stages right so
32:51Facebook didn't monetize for the longest
32:52time they were building a social network
32:54focused on users and now you know they
32:56did they monetize quite later but
32:58they're doing it beautifully now versus
33:00marketplaces on the other hand you do
33:02tend to see them monetizing a little
33:03earlier why is that the case because
33:05there is a transaction that is flowing
33:07through the marketplace already so let's
33:10take offer up our portfolio for example
33:12right there's a lot of transactions
33:13going on people the sellers are listing
33:15their items buyers want to buy the items
33:18and so there is a transaction happening
33:20now if you actually turn on payments in
33:24it actually improves the overall process
33:26right you don't want to you know meet
33:29the seller and then message them and
33:30just say let's meet outside versus you
33:32actually want to be able to send the
33:34credit payment directly so that the
33:36transaction is done so it improves the
33:39experience and so you can actually turn
33:40on monetization sooner and you tend to
33:42see that in marketplaces so I think it
33:44really depends on the stage of the
33:46company where you are and what sort of
33:49product markets that you have and when
33:51you can turn - I mean I think the
33:52business models are really different I
33:54actually think there is a meme with a
33:56false dichotomy that it's either growth
33:58or monetization and you know it's it's
34:01just a little bit like okay you've got a
34:03hundred people in the company and you
34:05can't have you know any on and you know
34:08you have to pick one or the other and no
34:09one else is allowed to cross over you're
34:11just a little you know it's there we've
34:13had a couple companies that shifted in
34:15the current capital environments and the
34:16uncertainty there shifted from growth at
34:18all costs to you know kind of smart
34:20growth mode or do you want to call it
34:22the amazing thing the profitability
34:24improved dramatically the growth and a
34:27lot of the cases didn't slow down you're
34:30just sitting there kind of like that's
34:35spending much money and it's still
34:37growing the same so for me I worry
34:39sometimes it's a false dichotomy I think
34:41in a marketplace business if there's
34:42money flowing through it you should be
34:43able to demonstrate you can capture some
34:45of it early on or I mean if not I'm
34:47trying to figure out if you're actually
34:48adding value to both sides of the
34:50marketplace and I wouldn't they be
34:52willing to pay so a marketplace without
34:54a monetization model is a little bit
34:56like really you know something like
34:58Facebook you have a little more of a
35:00call on or Pinterest you know being one
35:02you know they were multi billion dollar
35:04value pre-revenue business um you know
35:06the question is yeah there is a
35:08legitimate one question but yeah I mean
35:11typically we think there's more reasons
35:14to do it early than not to do it early
35:15on average so no false dichotomies let's
35:19open up to questions so when you think
35:21about the open table business model and
35:23originally you were selling software
35:25into restaurants I'm guessing you had
35:26competition that was also trying to sell
35:28similar software at what point did so
35:32and that one must spend like a feature
35:33war and classic competition at what
35:35point did the amount of reservations
35:37that you were driving
35:38become enough of a hook for why
35:42restaurants would always choose you guys
35:43for the software that it didn't matter
35:45that a competitor had a certain feature
35:47or you know a couple small things it's a
35:49fantastic question and it's got its got
35:51a unexpected ending there were three
35:53companies trying to do exactly the same
35:55thing at exactly the same time in the
35:57late 90s early 2000s one raised a round
36:01right of before financial Armageddon
36:03came in and a nuclear winter set in for
36:07literally it was a half decade in terms
36:08of funding new companies in the first
36:10half of the 20,000 opentable got raised
36:13enough money and spent it so slowly they
36:16were able to survive the two direct
36:17competitors didn't opentable ran
36:19unopposed for five years which is the
36:21only reason they were able to get to
36:22their network effect eBay when I was
36:25there during that time enjoyed the same
36:26thing I mean there was no company
36:28formation for five years so you know we
36:31were nice in a lot of markets and a lot
36:33of product lines and we just had we were
36:36unopposed for five years so when people
36:39get scared of the current market it's
36:40like yes be scared but also there's
36:42advantages in it if you know capital
36:44dries up to your competitive set and
36:46you've got you know you've got access to
36:47capital or whatever it is it actually
36:49can be it can be an attribute in there
36:52but I mean if OpenTable had actually
36:54GrubHub just went through this they they
36:57didn't have nuclear winter and they
36:59ended up having to roll up three or four
37:00companies to you to win the market they
37:03seamless campus foods you know they camp
37:05campus foods sat out campus seamless one
37:08a couple key markets like New York you
37:10know and so Grubbs a roll up a home away
37:13was a roll up you know there there
37:15wasn't a source of differentiation that
37:17was meaningful other than local
37:18Geographic and so they pieced it
37:20together they they went out of business
37:21yeah yeah and then it to open table ran
37:24unopposed because they they could make
37:26payroll and and that was incredibly
37:28informant formative that company because
37:30the network was so slow to build there's
37:33a slowest network I've ever seen to
37:34build and it's amazing it worked it
37:36wouldn't have worked without the capital
37:37environment anything like its current
37:39form and we have some local investments
37:42they have to go faster than opentable
37:43went I mean in circars going faster
37:45offer ops going faster just get there
37:47more virtual they didn't have they don't
37:48they didn't have to physically hurt each
37:52and the average owner of a restaurant on
37:54opentable owns one unit so it literally
37:57door-to-door trying to do it so it it's
38:01a business that wouldn't work today
38:02hi I have a question for you that I
38:05think is maybe about branding but maybe
38:07not when you are at Open Table were you
38:10ever tempted to offer a private label
38:13version of open table for a segment of
38:16the population that really wanted me
38:18special yeah that's kind of what we're
38:20facing right now where we have an
38:21opportunity to private label to a
38:24certain population that thinks that they
38:26should pay 10x for the same accounting
38:28because they're just really special one
38:33of the things that made restaurants an
38:34attractive marketplace well when I first
38:36met with my predecessor Thomas Leighton
38:38the first thing I does this is an
38:39ingenious model where else can I apply
38:41it and start my own company that didn't
38:43work the what made opentable interesting
38:46was two things one was the incredibly
38:48high degree of fragmentation which was a
38:51bug when you're trying to aggregate it
38:52but once you've aggregated it's a
38:53feature and the second was lead gen was
38:56a really important part of the
38:57restaurant experience most people don't
38:59want to eat at the same restaurant every
39:00night and so open table you know would
39:03put new butts in your seats but you've
39:05never been there that you're trying to
39:07get so we when we we started with
39:09mom-and-pop started winning the chains
39:11the good news is the chains would only
39:13have a hundred restaurants you know
39:15McCormick and Schmidt's Maggiano's there
39:18are a couple other m's and there I mean
39:19Morton's they only had like a hundred
39:21restaurants but they would say like
39:23which were Morton's and you know and
39:25they demanded we we held firm on two
39:28things one is we branded everything and
39:30so if they didn't want it they didn't
39:32and it was such a unique service that
39:34they couldn't afford to build it on
39:35their own at a hundred units the other
39:37was we forced to redirect from the
39:40Morton's website to the open table
39:44Morton's page on open table to create
39:47the brand awareness and that forced off
39:49that force redirect the fact that we
39:51converted restaurant people searching on
39:53Google for Morton Chicago into open
39:56table users is that is the open table
39:58growth hack I mean I that that was the
40:00open table girl attack by far because
40:02the supply base was so fragmented we
40:06able to pull that off and you know
40:07enabled to do that there's yeah I'd
40:10contrast that with Fandango been dangos
40:12basically has probably six or seven
40:14customers there's a large movie theater
40:16chains it's AMC its whoever and those
40:19guys could build their own thing and
40:21every once a while they pull out a
40:22Fandango emcee two years ago I think
40:24pulled out a Fandango and they could
40:26because they had the scale I'd go if I
40:29go to AMC Redwood City if if they're not
40:31on Fandango I'll go to emcee and it's
40:33like oh I can buy it through them
40:34because they have that level of critical
40:36mass so part of it is the market you're
40:37at do you have supply do suppliers have
40:40power or does is the marketplace more
40:43yeah then that's that sounds pretty
40:45fragmented to me so and then branded
40:50marketplaces we would value much more
40:53highly than white label because white
40:55label becomes you're selling utility
40:57you're selling software and the
40:59marketplace standpoint you're selling
41:00customers high do you believe in the
41:03idea of network effects with brands so
41:06for instance Nike the more people that
41:09think Nikes cool the cooler it gets the
41:11more people think Nikes cool and yeah I
41:14think we've had again you will hear the
41:16free actually and it's funny in the deck
41:26we actually that's why one of the charts
41:28that made it there was I don't know if
41:29you've seen it but there are three types
41:31of laws we call it which is Sarnoff slum
41:33at Casa la in REITs law so for brands we
41:36actually think it's more like Sarnoff
41:38law you know some people argue yes it
41:40has some Network effect but not directly
41:42incredibly value to the user which is
41:43why the value is not as steep as
41:46Metcalfe's law for example but the other
41:48side of the argument you usually hear as
41:50well that's like you know it's almost
41:52like Angry Birds you can argue they
41:54didn't have a brain they died on and
41:55they go but when you're playing the game
41:57and the game is hard and still having
41:58conversations about it and you know but
42:01you have to start the question is is
42:03that really value and sustainable value
42:05and is that something we would call a
42:08network effect it's a really interesting
42:09question I think it's a I think it's a
42:12different source of value and a
42:13different sustainability you also run
42:16into the diseconomies once everyone's
42:17wearing Nikes that people stop wearing
42:19so you're kind of I worked at Disney
42:21when it was a long time ago it was very
42:24very hot on the consumer product side
42:26and you know we couldn't sell enough and
42:29then five years later you can't give it
42:30away to the same audience and I just
42:32kind of you know Disney became you know
42:34was that I was there in the Lion King
42:36and all those and then there's any
42:38animation hit a hit a pause button Pixar
42:41clean their clock and you literally
42:43couldn't give away Disney's stuff so
42:44it's you know one of the most valuable
42:46brands in the world has done an Nikes
42:47done a little bit of it too they're
42:49they're hot and not back and forth so I
42:50you know the brand and the fashion
42:52overlay I think I mean I think they're
42:55unbelievably valuable commodities I'm
42:57not sure I'd call it a network effect
42:59going back to your statement about
43:01comfort the tool and stay for the
43:03network or the community it seems like
43:06the some of the strongest network
43:07effects actually happened when people
43:09not only use a tool from the top down so
43:13Nike branded shoes or an app from uber
43:17but when customers or users give back
43:19data or create something which then adds
43:22to the network that we benefit in a
43:24stronger and more coherent way when we
43:27have a creative tool or something that
43:29someone even if it's a post on a
43:31Facebook wall and I guess I'm curious to
43:33know if you think that there's room for
43:36more creativity there that these tools
43:38are actually better at creating network
43:41effects or faster or it looks like
43:43Pinterest took a long time but I'd be
43:45interested in in that perspective when
43:47pitchers started working it went very
43:49fast but you know so it took a long time
43:51to get the product market fit to make it
43:53work there's some fascinating ways
43:55entrepreneurs come in with the you know
43:57what is the value that more masses come
43:59we're seeing more and more data network
44:02based effects I mean just you know
44:04companies are saying like I'm doing this
44:06to get the data and with the data I can
44:07provide a vastly superior service and
44:10here's why ABCD you know the hottest job
44:14write down the world's the data
44:16scientists both my kids are in college I
44:18so you sure don't want to get a PhD data
44:22science but you know we're getting more
44:24and more things coming at us with you
44:26know either a corpus of content or a
44:29basically a data we've called a database
44:31network effect and there can be some
44:33powerful ones where you know the
44:35aggregated data and the system Pinterest
44:37should be able to put better content in
44:38front of you than anyone else because
44:39they have so many more signals and so
44:42much more content yeah it could either
44:44in the form of personalization it can
44:46also come in the benefit of b2b I think
44:48a lot more for example every you know a
44:51lot of theses and B players use shipping
44:52right and they can't negotiate great
44:55discounts like how Amazon does because
44:57they consolidate but he there is the
44:59shipping provider that is sort of
45:00servicing all of them a tech company and
45:03it's collecting all the data they can
45:05actually give better information back to
45:07the players in terms of hey if you're in
45:09Dallas or in this micro zone use this or
45:11like you know if they say it's priority
45:13you know two-day priority but it
45:15actually goes overnight 80% of the time
45:17so you can do money-saving servicing
45:19interesting startups that are sort of
45:20tackling those problems using data that
45:22customers are contributing because the
45:24transaction happened on the platform and
45:26using that to sort of give
45:27recommendations better recommendations
45:29on who to use I was really curious if
45:32you had any advice if there are adverse
45:36socio-economic or political or
45:38Geographic effects or even seasonal
45:40effects that could adversely affect the
45:44network effect I realize I'm saying
45:46effect is there any one advice on how to
45:51guard against or at least plan or buffer
45:54for these adverse effects as they're
45:56coming in that could actually disrupt
45:59this seasonality or probably the best
46:05example is 9/11 or everything kind of
46:09took a crash and if if you started a
46:13company that day odds are you would have
46:16very hard time getting customer
46:17acquisition same thing for you know at
46:21the onset of the Occupy Wall Street
46:22movement if you were trying to sell
46:25something to affluent individuals
46:27probably wouldn't have as much success I
46:30was managing a network effects business
46:32during 9/11 which was eBay
46:34you know our volume literally just
46:37dropped 40 50 percent overnight new user
46:40stopped completely it was a big divot in
46:43the business for two or three or four
46:47and everything popped back we never
46:49regained the lost users we had during
46:53that month they never came back but it
46:55recalibrated as managing OpenTable
46:58during at the financial crisis meltdown
47:00we did our org meeting on a mid August
47:04to be an act for the IPO in mid-august
47:062008 we told the bank's on Thursday
47:10before the Monday orde meeting who got
47:12the business Lehman Brothers didn't get
47:14the business they went bankrupt Saturday
47:16on that news Merrill Lynch did get the
47:19business so they traded the Bank of
47:20America on Sunday over the weekend the
47:23business the entire restaurant
47:25consumption market dropped 15% because
47:28of the panic and we award Monday morning
47:31that one lasted for like 15 months but
47:34very few of the startups or long-term
47:35yeah you need capital to pursue it but
47:38just you know the if it's a month you
47:40typically you know it's it's not going
47:42to completely bury it you know there if
47:45the shocks more than a month it's a it's
47:46a hell of a shock that's so it's a I
47:48mean I think the externalities other
47:50than you know I am thinly capitalized
47:53now have a lot of time to run the
47:54experiment other than that typically you
47:56you can understand it outside of it
47:58thank you guys for your great questions
48:01thank you and let's hang out and talk