Subscribe
Share
Share
Embed
课代表立正的官方Podcast
深度访谈，有用干货，亲身验证的「真本事」
Superlinear Academy创始人，Maven Top AI Instructor
前Statsig布道师（OpenAI收购），腾讯副总监，Meta，Amazon；康奈尔经济学博士
社区：Superlinear.Academy
课程：ai-builders.com
个人：lizheng.ai
so I want to ask you a few questions
about uh kind of a journey of having the
startup yeah yeah because you have a
very unique uh profile right you're
working your your early careers are in
large companies and now you do startup
how just how did it feel like I remember
last time I was interviewing you yeah we
talked about making this uh leap of
faith right and yeah I still feel that
way
uh when was you when were you like most
unsure like or most uh I mean
I think you can never eliminate that uh
I will just be honest right there are
times when I wake up when I feel so
happy excited about the possibilities
the positive possibilities about stats
like
um and then there are days when I wake
up and I'm like there are 45 people
um that work at statsi and they all have
families some of them have kids and so
now all of the livelihoods of everything
is dependent on the success of this
company which puts a lot of pressure and
so some days those things are like you
know ways down right as a Founder you
know you're responsible for building
something that people believe in people
have invested their their time and their
families
um yeah I feel like uh like making
decisions about products yeah are easy
actually like if you work at a large
company if you want to cut some product
you can make the decision right away
yeah it's very difficult to win until
you lost people yeah so in a it's a
never-ending thing so there's a spectrum
of possibility is in a startup on the
one end of the spectrum you can bang you
can go back you can make a bad decision
that could impact the product and we
could go bankrupt right away yeah we
could have a security issue and That
Could Just Kill our business instantly
um the possibility is still there it's a
non-zero probability our goal is to
reduce that to S zero as close to zero
as possible yeah and on the other end of
the spectrum is the you know okay our
startup just goes really successful
someday four or five years from now we
are all very successful uh we're all
very happy and then that probability
even though is non-zero is a tiny
probability but we're all trying very
hard to like increase that probability
so the but what gets me up a day every
day uh waking up and getting excited and
coming into work is the middle the
middle is where I get to do really
valuable really exciting work with
really amazing people
um on you know it's kind of like one of
those
um zero overhead you get to come in
you're productive you come excited you
leave exhausted in the middle you're so
productive that's what drives you every
day like every single day you wake up
and you want to do that right
um and that's what is exciting about a
startup
um I think I told you the last time
if you're an entirely rational person
you wouldn't want to join a startup the
expected outcome of a startup when you
add all the risk in there it's kind of
like you know less than what you would
do in a big company depends I feel like
the rationale is about the financial
income you're talking about the
financial income but rational I think
also means knowing yourself and we just
said right I think you're right yeah
absolutely so I think you're right so
there's an element of like you know
people that choose startups choose it
for non-monetary reasons yeah because if
you're only choosing it for monetary
reasons you wouldn't have to you
shouldn't so extraction right so you
would choose a startup because you like
the people that you want to work in
you're excited by the mission of the
startup you want to like you know have a
much larger impact you want to learn
more those are kind of like the factors
that go into like why would someone join
a startup and that's kind of like you
know that fundamentally that drives you
to like you know go be here in a place
like South Sega thanks so
all good experiences like even the bad
experiences or the fearful experiences
aren't a good part of the oh absolutely
yeah no I mean I will remember this till
till the rest of my life right these are
memories that we're making every single
thing that we do here every decision we
make whether good or bad
um we'll learn but the idea is that you
know I'm I'm I believe I'm making
memories uh that I'll remember forever
nice I I I like I couldn't detect any
difference between now and uh I think
last time we did the interview was uh 15
months ago yeah yeah like do you know
there's any difference
um like yourself
um I think I'm still just as excited as
I was uh about the prospects and the
future of static and
um well I mean obviously the last time
we didn't have customers you know we
have customers so there's a validation
about the product that is really good
and fulfilling uh and so that was that's
always good now obviously we've grown
quite a bit so uh you know we're now
more I guess like uh
larger and more mature and so we're
making process decisions and things like
that so I think like we've just moved on
to I would say stage number two from
stage number one when we were talking
before was station number two well I'm
just making it up yeah I feel like stage
number one is fighting for survival yeah
just uh like all energies into building
the product but station number two you
probably need to think about management
uh
the other thing is which is
fundamentally important is
um when you're that early uh the last
time we spoke
um you build products in vacuum where
you have an idea that you believe is the
right you know direction for the product
um but what we're doing now is very
different from what we did before
because now a lot of the features and
the roadmap is directed by what our
customers are asking for and what our
customers are needing and so while we do
still have like you know vision for like
their areas that we invest kind of like
you know future uh product areas a lot
of the existing products and then the
evolution of that is driven by feedback
and our customers are giving some
amazing feedback so which is great
because sometimes you you can't go too
long in a vacuum you want to like of
course get feedback and then that that
you know feeds it yeah like I
interviewed uh veteran in uh to be
business yeah in Silicon Valley he was
the VP I don't have been there
the end there yeah yeah and uh he built
the network structure there because he's
also a serious entrepreneur and uh
investor and he talked he talked about
how like I had a specific question uh
for him like uh part of the market fit
he he emphasized the uh importance of
part of my FIT which I resonate a lot
but the question was do people usually
find product Market fit right away or do
they have to Pivot many times to finite
uh I guess my question too is what is
our biggest pivot
um
yeah
um how do you feel about that one of the
good things about static was that you
know we used all these products inside
of these tools inside Facebook right you
kind of know the destination yeah so we
kind of that you know we kind of skipped
one of the steps of like you know
actually what should we build because we
knew what to build we knew how people
inside Facebook use these products uh
and how they evolved and how it scaled
so a lot of the learnings
um basically or early stage you know
product development was driven by the
learnings that we had from Facebook so
in a way we were fortunate enough to
have that kind of insight you were the
customers correct they know we're kind
of like building it for other people so
it became a much easier problem for us I
mean I I if other people are building
stuff like that more power for them
because it's good yeah that's you know
you save a whole bunch of like back and
forth and a lot of pivots
um but since then we also like added a
bunch of features that didn't exist
inside Facebook Facebook tools things
along the lines of like reviews and
visual Gifts of feature Gates and visual
diffs of experiments you know and then
auditing of these reviews uh what people
are changing across the board
um and then who gets control who acts
you know those are things that companies
want yeah I feel like those are real
demands that uh like just because
Facebook culture and how the teams are
set up yeah I think the culture at
Facebook is like anyone can go modify
anything I think at that point what they
were kind of like Engineers made the
right calls the Judgment was there yeah
um and then here it's a little bit like
you know the companies that we work with
they care about uh there should be
responsibilities and that should go
through this process and like there
should be one person that actually
reviews all of that and approves and so
they ask for those kind of um tools and
we built them nice yes how is answer
about uh pivoting as the technical
insights was always there those are
rarely uh pivoted or getting pivot but
the previous year is how to defend a
customer yeah how like who is a customer
and how to pitch them yeah yeah and
sometimes
um you'll be surprised how different
each customer is and what they focus on
uh what kind of areas that they focus on
and then sometimes you have to build it
to earn their their business all right
my last question about uh kind of still
the challenges of running on to be a
software is uh
sometimes
I think there is a trade-off do you want
to spend more time and resource optimize
our product or do you want to spend more
resource on surrounding our customers
specific need to customize
yeah initially when we had only like a
few customers we were at we were
basically like addressing every single
customer needs now one one thing to keep
in mind is like when a customer asks for
a feature
um I think you can also quickly identify
this is something that is going to be
valuable for all the other customers and
which point we prioritize that much
higher or is this something that is
valuable only for the specific user or a
customer because of their you know
unique needs in those cases what we
ended up you know we quickly understood
like we we didn't know about this we
learned that along the way so you can
tell right away yeah and then so we we
generally prioritize the ones that are
beneficial for a larger set of customers
um than the other ones sometimes we
throw them into the back burner and
sometimes we tell we talk to them how
important is it when do you want this
buy and then we can all work through now
that we have
um a broad set of customers we generally
like put everything into the our backlog
and then we then prioritize that you
know as a kind of like a global
prioritization of like you know who who
wants this and
um how many people want this and when
should we build this does it make sense
and so on now it is important to like
not entire like your product roadmap
should not entirely be controlled only
by customer requests of course there is
uh so I think like we have thirds um so
one-third is driven entirely by
customers uh one-third is driven by like
you know what we believe is the next
level or the next level of
sophistication for the existing tools
that may not be driven entirely by
customers and the third one is
completely new areas of exploration so
we recently launched events Explorer
which is basically scuba inside Facebook
so the real-time query engine so what we
do is we take the data yeah so we take
all of the data inside that the
customers have and then we sample it and
then we take the sample data and we put
it in a physical RAM and then we put a
query engine on top of physical RAM so
that way you can ask questions and
instead of like waiting you know 10
minutes 15 minutes you get sub second
results yeah and the the interesting
benefit of a sub second result is that
you can ask a lot more questions and you
can like quickly identify anomalies
which are masked by averages which is a
good way to like debug stuff scuba is so
powerful and I can't imagine living
without scuba right and so the idea is
that nobody asked for it but we built it
because that's the new area that we can
do as long as we continue expanding the
surface area or or we're going to be
adding value to our customers yeah I
have this painful Expressions because I
I'm jealous of people who get to use
this kind of tools yeah the
infrastructure of the companies
sometimes they're just lacking no I
think it's it's so I think just the fact
that Facebook invested so much in
infrastructure and had the foresight to
like put some of the best Engineers on
the internal tools uh is unique yeah I
don't see that in many companies
and that's basically like you know when
we build these
um there are about 20 tools inside
Facebook that we want to build yeah I
think we're at like three or four right
now and the the possibilities are
endless as long as the team continues to
just crank out features so fast
um I think over the next year or so
we'll have a pretty broad set of tools
so sorry I mentioned last question but I
have one more question no worries data
quality of your customers yeah like I
feel like being able to run good A B
testing yeah uh is heavily dependent on
the quality absolutely 100 and uh for
most companies if they don't have the
right infrastructure they don't have
good quality than how today yeah I think
um garbage and garbage out right so you
want to be careful about like what you
what you put in
um so there are several ways in which we
ingest data from our customers
um the
obviously if you already have a data
warehouse so we can adjust data directly
from whether the snowflake or redshift
or bigquery we can pull that from there
oh
um or the other way is if you have an
integration with any kind of like a cdb
platform so segment or Heap or router
stack or one of those things you already
have a pipe where the data is flowing
through we can also ingest from there
now if you don't have any of those
already we make it easy for you to like
take our SDK and then actually send raw
events directly to us so we make it easy
so the the data actually is
um same and the data is is actually
valid and not garbage so whichever path
our customers are willing to go we're
happy to go with them so typically what
happens is we do um a proof of concept
with an exist like you know if somebody
wants to like try and use stats we sit
down with them our team does a hack on
um depending on where the data is and if
the data doesn't currently exist we work
with them on like integrating our SDK so
we take great care in making sure that
the data that comes in is actually good
data because at the end of the day if we
don't get good data then we're going to
tell you all kinds of bad things and
it's not good for anyone all right yeah
okay thank you for your time cool
awesome well thank you thanks for coming
and visiting statsik yeah it's very nice
to have this rare opportunity to witness
yeah goals and uh I wish static becomes
very big and uh yeah it can Unleash the
Power of the data sense thanks thanks I
appreciate it knock on wood I hope it's
uh it comes true
um you're you're welcome again anytime
all right thank you cool thank you