0:00

genuinely believe we've got a once in a

0:02

generation opportunity to build a world -class internet

0:04

company out of India. We decided to

0:06

push ourselves through some really tricky moments. We

0:08

went from close to zero in scale

0:10

to like 200 million in like six months.

0:13

And I would not recommend that to anybody. We really

0:15

do it for the love of building. That should be

0:17

the end goal. The end goal should not be, I

0:19

build so I can make XYZ. I'm building so that

0:21

I have the opportunity to wake up in the morning

0:23

and build again. Welcome

0:29

back to another episode of How to Build

0:31

the Future. Today, I'm thrilled

0:33

to be joined by Audit Palachya,

0:35

the co -founder and CEO of Zepto,

0:37

which delivers groceries to your door

0:39

in just 10 minutes. There

0:41

are up to 1 .5

0:43

million orders a day across

0:45

50 cities throughout India, with

0:47

over $3 billion in GMV

0:49

and growing 300 % a year.

0:51

And they're a unicorn at

0:53

a $5 billion valuation. Audit.

0:56

It's great to see you here today. Thanks so

0:58

much for having me, Gary. I appreciate it. So

1:01

when you were getting started, there are some pretty

1:03

big players already in the space. Swiggy,

1:05

Zomato, even Amazon was at it. How did

1:07

you approach that? When we started and we

1:09

were very small, there was no sort of

1:11

grand strategy or ambition. Again, most of it

1:13

was bottoms up from users in the very

1:15

early days. But looking at

1:17

it now, I think the reason why we're able

1:19

to make a dent and create this scale is

1:21

because a lot of the people that were in

1:23

the market in the early days of grocery delivery

1:25

in India, were trying to build its supply chain

1:28

backward instead of user backward or customer backward, right?

1:30

And so they would build out models that were,

1:32

you know, two -hour delivery, four -hour delivery, which

1:34

made sense on PayPal, and like, you

1:36

know, logistically seem more feasible, economically

1:38

seem more viable, instead of building it

1:40

retention and consumer first. The insight

1:42

was, yes, let's do 10 -minute

1:44

delivery, but also let's build out the

1:47

full stack of commerce where we're

1:49

able to run the logistics,

1:51

we're able to have a

1:53

much better sense and selection on quality

1:55

control and sort of that audacious

1:57

thinking put us in

1:59

a position where not only will we able to build

2:01

a better product for users but over time and

2:03

now sort of what's playing out is that

2:05

if you build a better product for users you

2:07

also end up with a better PNL and

2:09

a better economics right so if users love the

2:11

platform your cac is lower the throughput you're

2:13

able to get into your supply chain is much

2:15

higher your costs come down and now we're

2:17

basically seeing like the economics play out pretty beautifully

2:20

better than like the two hour and four

2:22

hour models because of how sticky the

2:24

users are. So I think basically that's the

2:26

dent that we had. We were like

2:28

just thinking first principles, use of customer backward

2:30

versus a lot of the big guys,

2:33

the big behemoths that had like sophisticated PNLs

2:35

and finance teams were thinking supply chain

2:37

backward. I guess if you win the consumer,

2:39

you can win everything. Yeah. And the

2:41

big company sometimes lose side of that. You're

2:43

22 now, but when you were 17, You

2:46

had reached out to Jared Friedman,

2:48

one of our partners here at YC.

2:50

Tell us about that story. So

2:52

we were just actually supposed to come

2:54

here in California and study, but

2:56

the pandemic hit. So the freshman year

2:58

that we were supposed to have

3:00

at Stanford basically got postponed by a

3:02

year. We decided to take a gap year.

3:05

And my co -founder and I were sitting in Mumbai, really

3:07

had nothing to do. No sort

3:09

of fancy internships at Goldman or anything like that

3:12

lined up. So we just decided to hack

3:14

around, work on a project. And it just started

3:16

off as like a WhatsApp group where we were delivering

3:18

groceries for our neighbors because it was a nightmare

3:20

to get groceries in the first wave of the pandemic.

3:22

And that slowly kept iterating as just like a

3:24

fun project. And we were talking to people

3:27

on on hacker news about it and, you know,

3:29

just sharing updates. And I think

3:31

one day we saw posting get 10

3:33

minutes with a YC partner. And

3:35

we said, wow, this is YC was sort of this,

3:37

you know, for two kids on the other side of the

3:39

world was just sort of this mystical land, right? Where

3:42

we were like, you know, this hobby of building cool stuff,

3:44

people actually do it for a living somewhere in the

3:46

world, right? And so we

3:48

said, wow, we'd love to talk to a

3:50

YC partner and Jared was kind enough to

3:52

like agree after we, I think we, we

3:54

like flooded the comment section saying that, you

3:56

know, give us time, give us time. And

3:58

Jared sort of gave us the benefit. And

4:00

yeah, we had a, like, I think it

4:02

was an eight minute call. We

4:04

gave him a quick sense. He's like, Hey, there's a

4:06

company in the US called Instacart that does something similar

4:08

to what you guys are doing. Maybe this could be

4:10

a business. Have you ever thought of that? We said,

4:13

you know, not really. But then

4:15

he said, no, this could actually be something meaningful. Have you

4:17

ever thought of applying to YC? And that's

4:19

for the first time when we thought, wow, is

4:21

this like, I thought this was just like a

4:23

project, but actually it could be something meaningful. What

4:25

was the first demo like? How did that sort

4:27

of come together? Yeah, I mean, the first like

4:29

real interaction with the customer was basically a WhatsApp

4:31

group. There was like an old lady down the

4:33

road from where he was staying in Mumbai that

4:35

was really struggling to get groceries and she stayed

4:37

alone and so we were just delivering groceries for

4:39

her and you know she would basically say hey

4:41

uh you're going to this grocery store down the

4:43

road why don't why don't you also go to

4:46

the butcher store that was like the first year

4:48

interaction with the customer and you should be able

4:50

to do multiple stores and then she started adding

4:52

her friends to the whatsapp group and so more

4:54

and more people kept ordering um but that was

4:56

the early days and then I think the first

4:58

real product that we built was probably I would

5:00

say two months after the first delivery or like

5:02

a month a half two months after the first

5:04

delivery And at the time it was called Kirana

5:06

Card, right? And it was a

5:08

pickup and drop service similar to what Instacard is in the

5:10

US. That, you know, prototype

5:12

probably would have taken like 72

5:14

hours to build. I'll attribute most

5:17

of the heavy lifting to my

5:19

co -founder KB, right? But

5:21

we built it. We asked the users

5:23

to move to that app and that's how

5:25

it started. There was no like launch

5:27

moment per se. It was just like getting

5:29

feedback from customer doorstep and just cycling

5:31

around. And that just like... iterated into something

5:34

more and more meaningful day by day

5:36

But there was no like one like boom,

5:38

you know, you're a startup and that

5:40

never actually sort of happened until we got

5:42

to YC I guess the initial model

5:44

for it was Karana cart and then what's

5:46

the Karana for Yeah, for some of

5:48

the folks watching sure no Kirana basically is

5:50

the Hindi word for mom -and -pop, right?

5:53

So it's like these small mom -and -pop stores

5:55

that exist in India obviously in the

5:57

US you've got more Walmart's and Costco's and

5:59

Krogos But in India, it's

6:01

primarily these small -moment pop shops that that

6:03

dominant grocery. So that's what a Kirana. So

6:05

Kirana card was like corner shop delivery.

6:07

That's what it means. So the other thing

6:09

that was very important was sort of

6:11

extreme speed. So it wasn't, you know, an

6:13

hour or half an hour. It was

6:15

10 minutes. How did that come about? So

6:17

I think when you speak to most

6:20

people in the US, they think of 10

6:22

minutes as like a convenience value proposition.

6:24

But actually in India, it's very different, right?

6:26

So if you look at India, the

6:28

overwhelming majority of consumption actually happens within

6:30

four kilometers from your house which is very

6:32

different from the US right like the

6:34

frequency of purchase in India is four times

6:36

more for grocery than it is in

6:38

in the US and so people actually just

6:40

buy in these small tickets multiple times

6:42

a week which is why most retail in

6:44

India is hyper proximity right So unlike,

6:46

let's say the big box Walmart Costco models

6:48

that work in like suburban sprawls, people

6:50

are used to doors, the milk delivery guy

6:52

coming to the doorstep or the fruits

6:54

and vegetables guy showing up outside their house

6:56

every morning, or the local mom and

6:58

pop being 200 meters down the road. And

7:00

so that's the dominant format of consumption

7:02

and retail. Your 10 minutes was never really

7:04

like a, you know, let's do 10

7:06

minute delivery for the sake of 10 minute

7:08

delivery. It was more, you know,

7:10

Customers would basically keep telling us on their doorstep,

7:12

and we did a lot of the deliveries for the

7:14

first few months ourselves. They would basically keep telling

7:16

us that, hey, this is great. It's COVID,

7:18

but I would much rather just go to

7:20

my fruits and vegetables guy in the morning. He

7:22

comes pretty much there. Why do I wait

7:24

for two, three hours for you guys? It's

7:27

not a convenience thing. It's like I'm just used

7:29

to buying so much more frequently, so much more easier.

7:32

And so that's when we gradually moved

7:34

shorter and shorter delivery times to have

7:36

the ability to create that doorstep like

7:38

experience now obviously at scale the the

7:40

insight is that it's not just important

7:42

because people are used to it but

7:44

it's just it's a necessity because you've

7:47

got a lot smaller household sizes in

7:49

India you've got a lower penetration of

7:51

four wheelers you've got much more perishables

7:53

that people buy people have got lower

7:55

disposable cash so they don't actually have

7:57

the luxury of buying in bulk and

7:59

so when you fit all of those

8:01

consumer insights in and mainly just again

8:03

talking to people on the ground you

8:05

start realizing that hey this 10 -minute thing

8:07

is actually critical to do the way

8:09

the the style of purchasing people like

8:11

versus like a nice little convenience value

8:13

prop. So so that's how it came

8:15

out but I think it was more

8:17

just just talking to people and like

8:19

constantly saying okay we need to reduce

8:21

it every time by 30 minutes by

8:23

30 minutes by 30 minutes until eventually

8:25

we were like let's just be at

8:28

their doorstep. One of the things at

8:30

YC we often talk about is do

8:32

things that don't scale. Was this one

8:34

of those things where you had to

8:36

go and do those initial deliveries at

8:38

10 minutes and realized there's actually a

8:40

palpable difference in that experience from the

8:42

customer? 100 % so I think in

8:44

the middle of the batch we started

8:46

getting a lot more pushback from users. Basically

8:49

we're just not seeing the retention, right? And the early days

8:51

when you're delivering like in an hour. And

8:53

so people would use the app the

8:55

next week. three, four percent of

8:57

people use it again. And so that's when we started

9:00

talking to people and lots of things came out

9:02

like selection, pricing, quality, but this was one of the

9:04

big ones. And so we said, okay, we

9:06

have to be at their doorstep.

9:08

So what KV and I did

9:10

is we went to one of

9:12

the stores on the platform and

9:14

we basically commandeered the store ourselves,

9:16

right? And we ran it for

9:18

a short period of time, but we did a

9:20

pilot where we said, if we can, if we are

9:22

able to control the customer experience end to end,

9:24

including the delivery time, what's the reaction

9:27

that we get? And so I was basically

9:29

like the shopkeeper for a couple of

9:31

days, KV was delivering and we could

9:33

basically see that, you know, at the customer's

9:35

doorstep, there was a lot more delight. And that

9:37

was obviously the subjective indicator. And then as

9:39

we sort of got to a couple hundred orders

9:41

with that model, we could see that, hey,

9:43

people are actually repeating at like, obviously the data

9:45

was not that sophisticated, but we could see

9:47

the very early signs of, hey, there's these like

9:50

10 people that ordered last week, a good

9:52

chunk of them are still ordering this week. Maybe

9:54

that's a good thing. You did this unscalable

9:56

thing. You did it yourself. You're not running around

9:58

and hiring other people to do it. You're

10:00

directly experiencing it. But as a result, you

10:02

came to sort of ran it to ground

10:04

and you had this like very intuitive idea

10:06

about what the product needed to become. For

10:09

sure. And I think that it started with

10:11

like the subjective intuition, just like what are

10:13

human beings telling you verbally? You know, over

10:15

time, it became more objective, right? Which is

10:17

like, what's the retention of each cohort that

10:19

we're seeing? Are we seeing

10:21

that repeatability? And I think the that

10:23

rigor of testing whether we had

10:25

PMF came from the grinding we had

10:27

to do at YC. YC was

10:29

probably that most intense period where we

10:31

learned all of these frameworks very

10:33

quickly. And so Jared would

10:35

push us saying, okay, you're getting

10:37

good verbal feedback, but

10:39

are people really repeating? And if

10:41

that's the case, then you're

10:43

building something real versus just like

10:45

a hypothesis. You guys also

10:47

pioneered this dark store concept that now

10:49

a lot of people are using. Was

10:52

that also unnecessary

10:54

consequence of becoming a 10 -minute

10:56

provider. Like you needed control

10:58

over what that experience would be

11:00

like. Absolutely right. Like being

11:02

able to control fulfillment logistics was

11:04

critical for us to be

11:06

able to get that like that

11:08

delivery time in SLA for

11:10

the customer. But more importantly

11:12

than that, you know, the people think of

11:14

us as like a temporary delivery provider here. If

11:17

you go into the ground in India,

11:19

we sort of looked at more as

11:21

like a like a hyperlocal amazon not

11:23

just like good serviceability to the customer

11:25

but also great selection, great quality, great

11:27

pricing across and you know today we're

11:29

doing not like you know 700, 800

11:31

products we're doing now almost 45 ,000

11:33

to 50 ,000 products that you can order

11:35

and zap to everything from like you

11:37

know oranges and apples all the way

11:40

to earphones and hoodies we're basically building

11:42

out an internet supermarket chain right although

11:44

people sort of anchor on the 10

11:46

-minute piece the owning the full stack

11:48

logistics is also pretty consequential for

11:50

all the other axioms of customer experience. So

11:52

it's speed for sure, and then there's also

11:54

quality, there's also selection, and there's also price.

11:57

And so all four of those pillars

11:59

of customer experience improve dramatically, at least

12:01

in our experience, when you're able to

12:03

just control the entire stack. So I

12:05

think as a result of owning it,

12:07

we're able to do 50 ,000 products, we're

12:09

able to control the entire supply chain

12:11

on the back end and get better

12:13

quality of produce for customers. Now

12:15

at scale, we're able to cut

12:17

all the inefficiencies in the back end and give

12:19

them better prices, right? And so all of that

12:22

is at the end the day, helpful to the

12:24

customer. So the customer thinks of us, you know,

12:26

temporary deliveries would cut our foot in the door. But

12:28

now they're basically thinking of us as they're, they'll

12:30

go to like supermarket chain in India, or they'll

12:32

go to one stop shop. One of the blessings

12:34

that we had pretty early on is that like,

12:36

and you know, you remember this as well, right?

12:39

Very early in the company's life cycle, we

12:41

got hit by March 2022, where capital markets

12:43

basically fell off a cliff, and we couldn't

12:45

raise capital and we had two competitors that

12:47

had seven times the cash on balance sheet

12:49

that we had, right? And so

12:52

that was like a big forcing

12:54

function, right? Like, execute or die

12:56

was the philosophy in 2022, 2023,

12:58

we had to just create as

13:00

much efficiency, create as much operating

13:02

level as we could very quickly, maybe

13:04

faster than a lot of other consumer

13:06

and trade or e -commerce companies have

13:08

globally, very much early in our life

13:10

cycle because it was existential, right? And

13:12

in retrospect, that was the best thing

13:14

that could have ever happened to us

13:16

because when we are now well capitalized,

13:18

we're able to not only allocate that

13:20

capital better, but we're able to do

13:22

it more efficiently, get more growth for

13:24

a lot less investment. And so essentially

13:26

what happened is that we got to

13:28

this point in March, April, 2024, where

13:30

we had gone from zero in August of

13:32

2021 to two a half years later, we

13:35

had hit about a billion in GMV. And

13:37

the business was actually pretty close back then.

13:39

to EBITDA and operating cash flow break even. And

13:41

once we were in that point, we were

13:43

able to show, hey, 70 % of our markets

13:45

are profitable. The most mature markets are touching 4

13:47

% to 5 % EBITDA. And

13:50

they're turning profitable faster and faster. And we're still

13:52

able to grow. But why don't we accelerate? If

13:55

the economics are working, then

13:57

we should actually be accelerating into the

13:59

business. Instead of trying to cash our

14:01

chips in at just a billion of

14:03

GMV, we've got an opportunity to build

14:05

the largest you know, grocery player in

14:07

India, which is a 700 billion dollar

14:09

market, right? And so if I'm in

14:11

that position, we can easily build 20,

14:13

25, 30 billion of top line. We

14:15

should be accelerating today. We should be

14:17

doubling down. And that's what's basically happening

14:19

as we speak. What are some of

14:21

the bumps that you hit in the

14:23

road? And like, you know, even as

14:25

a 22 year old founder, how did

14:27

you think about, I mean, Holding teams

14:29

to account like these are sort of

14:31

the more difficult things especially in 50

14:33

different cities like were there cities where

14:36

the numbers didn't work out and you

14:38

had to find a new leader or

14:40

You know did you have to react

14:42

to different like specific things in different

14:44

cities that were surprising? I

14:46

would say some of the biggest roadblocks

14:48

have been people in talent related like

14:50

we like the the most fundamental input

14:52

for our companies is execution excellence and

14:54

like the most important input to execution

14:56

excellence is is high quality talent. And

14:58

so the biggest mistakes I've made as

15:00

a consequence of that are the wrong

15:02

hires in some cases or the wrong

15:04

people management. So in retrospect, the setbacks

15:06

have been, oh, you know, we could

15:08

have done, you know, let's say finance

15:10

better or marketing better or, you know,

15:12

category management better or ops better, if

15:14

we had had the right person. And

15:16

if you those mistakes are very costly,

15:19

especially in that 2022 2023 period, where

15:21

it was existential for us. And so

15:23

there were situations where we had, you

15:25

know, not so much cash, you know,

15:27

we almost effectively kind of died in

15:29

that Silicon Valley bank thing, if you

15:31

remember that, right? Absolutely. We had a

15:33

lot of our cash in that. In

15:35

that period, some of those mistakes could

15:37

have been very costly. And so, so

15:39

yeah, like I think at a high

15:41

level, just getting like the right folks

15:43

on board, we made some mistakes initially

15:45

and now I think we've learned from

15:47

them and that helped us. Yeah. Let's

15:49

go a little deeper into some of

15:51

the existential moments for Receptel. What

15:53

are some other moments that gave

15:56

you sort of the deepest lessons

15:58

into building this company and coming up

16:00

as a CEO. In the very

16:02

early days, I think we had plenty

16:04

of existential crises that the YC

16:06

partners helped us through. Before

16:09

we got into this whole first -party thing,

16:11

we were realizing when we did an

16:13

intellectually honest assessment of the business in the

16:15

early days that, hey, there's no

16:17

retention here. And in this sort of pick -up -and -drop -off

16:19

service, we're not able to do it well. And so

16:21

we thought we were dead. And there

16:23

was a point. in the earliest days

16:25

where we just said, I mean, what

16:27

are we going to do? We've tried

16:29

everything. We've killed ourselves. You know,

16:31

we're working. This was before the 10 minute

16:33

moment. Yeah. This is before the first party stack.

16:36

And so this is before we started commandeering

16:38

stores, right? And we said, yeah, this model is

16:40

not working out. You know, we thought it

16:42

was going to be perfect. We'll just do this

16:44

model. It's going to grow. Everything's to be

16:46

great. Got a reality check. After like a few

16:48

months of like hitting the wall, we said,

16:50

is there even a model that works? And like

16:52

pretty much everyone's telling us. that, hey, grocery

16:54

delivery in the world has never worked really that

16:56

well, other than a few exceptions. India,

16:58

there were lots of companies that came before us. There

17:00

was a graveyard. And so all the

17:02

feedbacks we were getting from investors and from

17:04

people that seemed to know much more than we

17:06

did is that you guys are essentially in

17:08

a dead space and we could see it in

17:10

the data. So in the very early days,

17:13

there was that like super despair period where we

17:15

would have, I mean, very close to giving

17:17

up, right? That would have been probably the highest

17:19

likelihood chance of death that this company had.

17:21

And then obviously as we became like a larger

17:23

and larger company, there were all

17:25

sorts of crises, right? I mean, I

17:27

think we had a, if you've been in

17:29

positions where, you know, we're about to

17:31

raise capital and then a crisis happens, whether

17:33

it's, you know, March 2022, when the

17:36

Ukraine war happened that period, or let's say

17:38

we're about to, it's funny, actually, the

17:40

most of our fundraisers have coincided with some

17:42

sort of crazy crisis that's happening, right?

17:44

And it has actually made it harder, right?

17:46

Like, I think right before we raised

17:48

that round in 2023, which was like, like

17:50

a fundraise from hell almost right because

17:52

it was like the only unicorn round in

17:54

India for that year immediately prior to

17:56

that the Silicon Valley Bank thing happened and

17:59

so uh today's that round took like

18:01

eight months seven months within that I mean

18:03

like there were lots of places that

18:05

we could have died in that were real

18:07

like there was there was a period

18:09

where we had to build a lot of

18:11

financial controls right like we had to

18:13

go through like EY audits that you know

18:15

if we didn't make it through them

18:17

as tightly as we would have wanted we

18:19

would have, you know, not gotten capital,

18:22

right? And when you were competitors with over

18:24

a billion dollars of cash in bank,

18:26

it's existential, right? And we didn't want to

18:28

sell. And so we just had to,

18:30

like, push ourselves through some, like, really tricky

18:32

moments there. But there are probably plenty

18:34

more still to happen, right? I'm sure. I

18:37

think the really cool part of your

18:39

story is that, you

18:41

know, you didn't look at sort of

18:43

how other people were doing it. And I

18:46

sort of accepted that as, you know,

18:48

sort of ground truth. There's an

18:50

X maxim now that is, you know, you

18:52

can just do things. So

18:54

I see a lot of, you can just

18:56

do things energy and what you're talking

18:58

about here. I mean, so a little bit

19:00

of naive and young that helped actually,

19:02

like, you know, it's like, they say it,

19:04

right? Like, if you, if you knew

19:06

how hard it was when you started, you

19:09

wouldn't have done it, right? And so.

19:11

That's the Paul Graham Slep Blindness essay in

19:13

a nutshell. So you've

19:15

personified it. I really like how you

19:17

thought about retention. Did it

19:19

feel like you were just asking questions

19:21

about that? And out of asking questions, why

19:24

is retention not where it could be? That

19:26

sort of teased out, well, we could try

19:28

this, we could try that. Were there a

19:30

few other things you tried other than the

19:32

10 minute thing? Yeah, we did. We tried

19:34

a SAS model. for mom and pop shops,

19:36

we try to build like a B2B logistics

19:38

play, right? Like, is there like a better

19:40

model? Maybe like, what if we just do

19:42

the, do the only the delivery part or

19:44

what if you only do the software part,

19:46

right? And we just realized that like, you

19:48

know, grocery in India is, you

19:50

know, can be much better than it is

19:52

in general, as like a, as a

19:54

market and like, we realize that everything is

19:56

so broken, right? Like, whether you go

19:59

into the back end, you go to the

20:01

storefront, you go to the delivery in

20:03

the last mile. If you want to build

20:05

something that really solves the problem end

20:07

to end for the customer, which is across

20:09

like speed, quality, assortment and price, then

20:11

you just need to suck it up and

20:13

own the whole thing. The push basically

20:15

came from a lot of just like radical

20:17

candle, right? And I think that you

20:19

don't really find a lot of that, you

20:21

know, candidly in India so far, which

20:23

is why obviously YC was as valuable as

20:25

it was because, you know, people were

20:27

just like, You know brutally honest

20:29

with us that hey, man, you want

20:31

to make it you want to build it's

20:33

Instacart you want to build or dash

20:35

would have been like a nice consumer company

20:37

This is what the early days of

20:39

those guys would have looked like obviously they

20:41

went through their own struggles, but you're

20:43

not not you're not at a point where

20:45

You can credibility claim product market fit

20:47

and then having that like really brutally honest

20:49

feedback loop really helped us.

20:51

And then we sort of pushed ourselves to say,

20:53

okay, if we're being super honest ourselves, are we

20:55

building an exceptional product? Right? Sure, maybe we're getting

20:58

a little bit of a pop in the early

21:00

days because it's COVID, people are

21:02

locked down, they have to use the

21:04

app. But can I really say that build

21:06

something that somebody actually wants long term? And

21:09

until we had conviction that we were there,

21:11

we didn't go all in. It was just like

21:13

a great feedback loop and like a lot

21:15

of like a lot of tolerance for iteration and

21:17

pain basically. Talk to me

21:19

about that moment where you did feel like

21:21

you got a product market fit and that's

21:23

when you know maybe you were in only

21:25

one location and then you said okay time

21:28

for you know did you do one additional

21:30

location or did you say all right like

21:32

five locations like what was that like? It

21:34

was actually the first location first dark story

21:36

launches in a neighborhood in Mumbai called Bandra

21:38

and pretty quickly that neighborhood became bigger for

21:40

us in the rest the entire city that

21:42

we were operating in and it just sort

21:44

of ballooned And I remember going there because

21:47

it was an experiment, right? And it was

21:49

like, effectively, it was like a cardboard, cardboard

21:51

cut out almost of like what a dark store

21:53

looks like today, because now a dark store is

21:55

like, you've got proper racking and, you

21:57

know, chillers and freezers and everything all set

21:59

up properly. But over there, it was

22:01

just like a bunch of like stuff on the

22:04

ground, basically, and we were just living from there. But

22:06

the experience was so well managed because we were

22:08

able to, you know, curate a good selection, right, and

22:10

work with a couple of people to be able to

22:12

do that. we were able to get the deliveries consistently. And

22:15

so just started picking up very rapidly. And

22:17

long story short, I mean, I did a

22:19

couple of deliveries and I was like, wow,

22:21

people really have a spark in their eye

22:23

when they're getting like a delivery here versus

22:25

like what I'm used to, which is, oh,

22:27

this is bad, this is bad, deliveries delayed,

22:29

items are not right. So yeah, I think

22:31

that was like the aha moment. And

22:33

that is also when we were, we got

22:35

lucky to just keep getting like the right

22:37

mentor at the right point of the journey.

22:39

And so that's when you know, our CDZ

22:41

investors, which is Nexus in India, basically came

22:43

in and said, hey, you know, this is

22:45

real product market fit, probably, right? And we

22:47

said, I mean, are you sure? Like, you're

22:49

very cynical, right? That's when I think the

22:51

right time at the right place, we met

22:53

the people that gave us the conviction that,

22:55

you know, you can actually try this again

22:58

in the second store, try this again in

23:00

the third store. And that's when it started

23:02

just really picking up. We went from basically

23:04

like close to zero in scale. to like

23:06

200 million in run rate scale in like

23:08

six months. And like, I

23:10

would not recommend that to anybody. It

23:12

caught fire. Like people really wanted it.

23:14

You know, like pulling the product out of

23:16

you. Like, when are you coming to

23:18

my neighborhood? Yeah. And even like just, I

23:21

mean, obviously it was 2021. So it's

23:23

crazy, right? But even from a capital perspective,

23:25

we went from like effectively the 125K

23:27

from YC to like a $9 million round

23:29

to $50 million round to $100 million

23:31

round. in like four months. And

23:34

that was obviously crazy back then.

23:36

In hindsight, it was probably the right thing because

23:38

that was like sort of the inflection point

23:40

to just go all in on. And then in

23:42

2022, we stabilized and said, okay, let's start

23:44

building a business out of this. I guess given

23:47

that, like, do you consider Zepto a success? No,

23:49

absolutely not. Like, I mean, not even

23:51

close to it. I think we've got

23:53

a couple of decades before we can

23:55

realistically say that we've won. I genuinely

23:57

believe we've got a once in a

24:00

generation opportunity to build like a a

24:02

world -class internet company out of India.

24:04

And you know, sitting in places like

24:06

this, you get really humbled because you

24:08

realize that, wow, like look at

24:10

the kinds of quality of companies that have

24:12

been built in the US and in

24:14

other parts of the world. And I don't

24:16

think we're there yet. And until we

24:18

can credibly come in and say that, hey,

24:20

we're able to hold our own to

24:22

some of the best globally, like an Amazon

24:24

or a DoorDash or a Mercado Libre

24:26

or all these other companies, then I don't

24:28

think we won. I also think that

24:30

the ambition is less about like

24:32

oh this valuation this that like it's

24:34

now come to a point where you

24:36

just really want to build something exceptional

24:38

that can be like a benchmark for

24:41

the rest of the ecosystem. So I

24:43

think it's going to take 20 -30

24:45

years to get there. We are

24:47

very excited to do that for the

24:49

next 20 -30 years minimum and then much

24:51

more. And we just have to execute like

24:53

crazy to make that happen. So that's

24:55

what that's what we're trying to do. Yeah,

24:57

people are always overestimating what they can

24:59

do in one year and way underestimating what

25:01

can happen in 10 or 20. Yeah,

25:03

100%. So you just you once tweeted, nothing

25:05

against work life balance. In fact, I

25:08

recommend it to all our competitors. Yeah.

25:11

You know, stupid idea. I think about that

25:13

often. I mean, I think it is

25:15

possible to have balance. But you know, you

25:17

can also do you can run your

25:19

work like a really good marathon. Yeah, yeah,

25:21

for sure. You know, I think the

25:23

people that like work at a company

25:25

like Zeppelow, I'd say that, you know,

25:28

join here if you're really looking to do the best work

25:30

of your life, right? And like, this is the place that

25:32

is going to get that out of you, right? Because

25:35

it is a very execution focused culture. Everyone

25:38

is really deep into the game. When you're

25:40

just like in these super execution focused cultures,

25:42

then you're really like accountable for real problem

25:44

statements every day. you can actually just become

25:46

like the best version of yourself. And I

25:48

think the, it's not for everybody, right? That

25:50

sort of mindset and thought process, but it

25:52

is for like the super ambitious, the super

25:54

capable that really want to, to get the

25:56

best out of themselves. And

25:58

it's not forever, right? You're not going have

26:00

to do this, like work so hard for

26:02

many, many decades, but like you definitely need

26:04

like, I think I believe portions in your

26:06

life where you take yourself to the maximum,

26:08

right? And really achieve a lot. And

26:10

I think Zepto is sort of that place. It

26:13

doesn't feel like work, at least for me. And

26:15

I know that that's not a realistic expectation for,

26:17

you know, today we've got like 3000 plus corporate

26:19

employees, right? And mainly ops folks,

26:21

but like with 3000 people. And it's not realistic

26:23

for me to tell everyone that, hey, you know,

26:25

love what you do, work seven days a week.

26:27

It's not realistic, right? But at least for KV

26:29

and I, I mean, just, it feels like the

26:31

journey of a lifetime, right? And like, there's nothing

26:33

more exciting than this to be spending time on.

26:36

So. That's, I guess,

26:38

the thought process. But what I will say

26:40

though is, if you don't really have

26:42

a mission that excites people, then you can't

26:44

push them to that level to really

26:46

give them their all. And I think that

26:48

SpaceX is a good example of a

26:50

company that is just able to get the

26:53

maximum out of people because it's such

26:55

an inspiring mission. Obviously, we're not setting rockets

26:57

to space. We're delivering groceries. But

26:59

I think that the mission that at

27:01

least I like to try to bring out

27:03

is that, guys, let's build a truly

27:05

great... company in India and let's kick off

27:07

the internet revolution in India that should

27:10

have been much bigger than it is today.

27:12

I think we can create 10x more

27:14

value for the country than we have so

27:16

far. I think if you

27:18

look at even places like China, the

27:20

dynamism came from the internet companies

27:22

and then that spread everywhere. So

27:24

maybe we can play a small

27:27

part in that. That would be great.

27:29

What are some of the unique

27:31

advantages and challenges you've experienced building a

27:33

startup specifically in India? So I

27:35

think the advantage is that the talent

27:37

is incredible, right? And

27:39

I initially didn't realize

27:41

this deeply, but,

27:44

you know, we interviewed like American engineers

27:46

and obviously their exception would be said,

27:48

hey, the guys we have are

27:50

as good, right? And I think

27:52

a lot of people underestimate that. That's

27:54

why, you know, a lot of great internet

27:56

startups set up offices in Bangalore, right?

27:58

And hire people there. And so the big

28:00

advantage is that you just have such

28:02

an incredible talent base and really although it's

28:04

very competitive, it's still a lot less

28:06

competitive to get such high quality talent than

28:08

it would be sitting in San Francisco,

28:10

I'm sure. I think some of

28:12

the hurdles are as an ecosystem, I think

28:14

we're very much like post 2001 right now,

28:16

because we had like a big reality check

28:18

in 2022 -23. And there's still a lot of

28:20

fear and lack of ambition in general. And

28:22

I'm sure you've seen some of that like

28:24

in the post 2001 kind of period, I'm

28:27

sure, where people were less ambitious

28:29

than they could have been. So we're sort of

28:31

coming out of that and saying, you know guys

28:33

why we you know if we execute well we

28:35

should go we should swing for the big leagues

28:37

and so i think the the challenge is like

28:39

whether it's hiring a senior senior executive whether it's

28:41

you know bringing on board a new board member

28:43

i feel sometimes and sort of scavalier like a

28:45

little bit in the minority camp ever seeing it's

28:47

very easy to be at like three four billion

28:50

of gmv and just take it uh take it

28:52

easy right and just get to break even and

28:54

then build a nice little company and go public

28:56

you do well right but wouldn't it be much

28:58

better to go for Five x

29:00

that six x that and if there's opportunities

29:02

there why not go for it? And

29:04

so I think in the US that mindset

29:06

is sort of taken from granted Everybody

29:08

has that here. Yeah, right. It's and it's

29:10

such a special place because of that,

29:12

right? But in a place like India, it's

29:14

still germinating I mean, I think you're

29:16

one of the leading importers of the philosophy

29:18

itself. Yeah, hopefully we've got the talent

29:20

We just need to get that mindset the

29:22

dynamism basically, right? So what else do

29:24

you have on the plate and sort of

29:26

in your vision for Zepto over the

29:28

next 10 20 years? I think from our

29:30

perspective there's a lot of areas to

29:33

innovate in like Indian consumer internet and we

29:35

want to I mean we don't want

29:37

to get distracted where we just you know

29:39

try multiple verticals like we really want

29:41

to just go deep into the compounding flywheel

29:43

that we have a product market fit

29:45

and say how do we just keep improving

29:47

our product market fit I believe that

29:49

PMF is like it's a it's not static

29:51

it's dynamic right you can keep improving

29:53

it over time if you keep innovating and

29:55

so there are like a few exciting

29:57

innovations we're working on so for example there's

29:59

Zepto cafe which is basically first -party food

30:01

delivery that we've now started to build

30:03

into our dark stores where in like a

30:05

three a half thousand square foot facility

30:07

we have 250 square foot that's allocated for

30:09

coffee, tea, snacks, like if you've been

30:11

to 7 -Eleven in Japan for example even

30:13

in the US you have all that amazing

30:15

in Japan not so good here yeah

30:17

yeah yeah exactly right and so that it

30:19

came from the same inspiration and we

30:21

build that out there and that's doing phenomenally

30:23

well so we've scaled that effectively starting

30:25

last year and we've gone from zero to

30:27

over 100 ,000 orders per day. And that's

30:29

the beauty of all these consumer internet

30:31

businesses, right? When you have like 15, 20

30:34

million people opening your app every week,

30:36

and then you add like a new little

30:38

use case for them, you can just

30:40

funnel hundreds of thousands of people just by

30:42

moving pixels around. That's like this super

30:44

exciting thing about this business. That's why I

30:46

love it so much. Sounds like it's

30:48

the true everything store, but it's not only

30:50

just the everything store that will come

30:52

to you tomorrow or maybe later next week.

30:54

Yeah, it's like literally in 10 minutes.

30:56

Yeah, it's like the you know, hyperlocal everything

30:58

store for hyperlocal India, right? Like that's

31:00

the idea. But there's a lot of other

31:02

cool stuff. I mean, even if you

31:04

look at our advertising business, right? Like similar

31:06

to like the other really successful e -commerce

31:08

companies, we've gone from like a 40

31:10

million ARR in ad revenue last year to

31:12

this year, we've just crossed 200 million

31:14

ARR, right? And that's just, I

31:17

mean, it's obviously phenomenal business. And

31:19

we're just innovating there in an

31:21

exciting fashion. We built out like

31:23

our search stack in -house, the

31:25

entire relevance engine. meeting, attribution, campaign

31:27

management, automated keyword suggestions. I

31:29

think we've built out a really high quality performance

31:32

ad stack out of India. For

31:34

one of the first players to have actually been

31:36

able to do that, obviously Flipkart and few other

31:38

folks have done it, but we built it really

31:40

quickly and it's giving us a lot of value

31:42

in that business also scaling. So ads is another

31:45

big focus, cafes are a big focus, and it's

31:47

all sorts of other stuff. And we're, you know,

31:49

we're starting to get into, you know, electronics as

31:51

a category, general merchandise and a paddle, cosmetics.

31:54

So we're just getting more and more into

31:56

as many use cases as we can

31:58

find. And all of it user backward. And

32:00

so I basically spend time with the

32:02

team looking at, hey, what are people searching

32:04

for that we're not able to give

32:06

them? And then let's start adding those use

32:08

cases. So when people started searching for

32:11

coffee, we said, let's put in Zepto cafe.

32:13

When people started searching for lipstick, we

32:15

said, let's start adding cosmetics. So the flywheel

32:17

is spinning as a result of that,

32:19

where we're getting better retention and we're getting

32:21

a better PNL end state as a

32:23

result of that. There's a lot more to

32:25

push on that, so very early. But

32:28

I think we could, maybe we'll be talking hopefully

32:30

if we execute, well, it's a big F, but

32:32

if we execute well a few years from now,

32:34

we'll be talking about a new form of e

32:36

-commerce that we've sort of invented out of India.

32:38

And I think the Chinese guys were able to

32:40

invent new forms of e -commerce too, right? So why

32:42

can't we? You know, going through your description of

32:44

how Zepto works in India. it

32:47

actually sounds way more awesome

32:49

than what Amazon is doing. Amazon

32:51

tried to do some of

32:53

these faster delivery type things. They

32:55

scaled them back. We sort

32:57

of don't know if we'll ever actually

32:59

even see them. If anything, it's heartening

33:01

and pretty exciting for the Indian ecosystem

33:03

in that there is this very powerful

33:05

energy where you can just do things

33:07

and it's alive and well. It's happening

33:09

in the other side of the world

33:11

too. Yeah, so we're in this age

33:14

of intelligence as Sam Altman calls it

33:16

and I'm a deep believer in that.

33:18

100%. What are you seeing at Zepto?

33:20

So the way that I look at

33:22

it is, you know, we're like a

33:24

very nascent internet native company. And we,

33:26

I was actually talking to someone senior at

33:29

Microsoft yesterday and he gave us, gave me

33:31

a really good framework, which is on the,

33:33

we're basically the application side of, of the

33:35

AI acceleration, right? And if I

33:37

have to figure out what applications I

33:39

need to prioritize because In the end,

33:41

all the applications are going to be

33:43

fully fleshed out, either through agents or

33:45

something else that I don't know. But

33:47

my job right now as CEO is to say,

33:50

OK, what are the applications that I need to

33:52

expedite that I need today that are core to

33:54

my business? And what are the applications that would

33:56

probably come out on their own and are non

33:58

-core to my business? But I'll them, but are

34:00

non -core to my business. So for example, let's say

34:02

non -core would be accounting. Can

34:04

I make my accounting 10 times more efficient,

34:06

10 times more accurate? Eventually, 100%. There'll

34:08

be an agent that does that for me.

34:10

Can I make my legal contract work

34:13

that I do at scale? Can I make

34:15

that much more efficient? Absolutely, I can.

34:17

And we have a lot of that stuff

34:19

that we have to do in India.

34:21

But those are not really core. And someone

34:23

is going to solve that for me

34:25

later. And I'll pick up an agent or

34:27

whatever is much more than I do

34:29

a year from now or nine months from

34:31

now. And I'll solve that problem. But

34:33

today, what I'm focused on on the application

34:36

side is, what are the things that

34:38

are super core that I need today? And

34:40

I will build some of the capability

34:42

in -house to be able to apply that

34:44

as soon as possible to my business as

34:46

an internet native company. And so, for

34:48

example, let's say the ads business that I

34:50

was talking about. So automated keyword suggestion,

34:52

for example, or bidding and attribution, we

34:55

were able to get much

34:57

better return on ad spend for

34:59

our clients. on the

35:01

ad side, like Unilever and Proctin Gamble

35:03

and Coca -Cola, we were able to

35:05

do that because we started training on

35:07

Lama to try to build out a

35:09

better, I mean, to basically build out

35:11

the relevance engine in -house and that gave

35:13

us much better results and we were

35:15

able to search the ad's business as

35:17

a result of that. So that had

35:19

a big impact on bottom line. Or

35:22

for example, the easiest application that everyone

35:24

knows is customer support. And now that's

35:26

obviously playing out, but we said we

35:28

need this customer support stack much quicker.

35:30

So we've built out like a crack

35:32

team that's, again, training on some of

35:34

the foundational models, looking at all our

35:36

data. let's say like rotten bananas for

35:38

example because we do one a half

35:40

million orders per day so there'll be

35:42

rotten banana there'll be delayed order or

35:44

there'll be all those sort of customer

35:46

support tickets I'm basically saying let me

35:48

just train this model myself or whatever

35:50

let me use what exists out there

35:53

and give them my data and build

35:55

out an application for myself and start

35:57

automating every ticket and so today you

35:59

know greater than 50 percent of the

36:01

tickets that are raised on Zepto not

36:03

just like a rules -based chatbot are

36:05

actually sold dynamically by a generative chatbot

36:07

that's been modified for our exact use

36:09

cases, right? So we'll focus on these

36:11

key things like search, ads, customer support,

36:13

forecasting, supply chain forecasting, another core part

36:15

the business we built a lot of,

36:17

not generative AI, but a lot of

36:19

deep machine learning work that we've done

36:21

there. That's exciting. I mean, I think

36:24

one of the observations is that basically

36:26

these are not solved problems yet and

36:28

that's actually exciting. and or even this

36:30

idea that you could do the everything

36:32

store across every category and you can

36:34

basically snap your fingers and add almost

36:36

any new category. Yeah, all of these

36:38

things are actually empowered by the technology

36:40

themselves and those are coming at a

36:42

more and more rapid pace actually. Maybe

36:44

to wrap things up, I guess my

36:46

favorite question often is what advice would

36:48

you give the 17 -year -old version of

36:50

yourself? And it was not not

36:53

not too long ago But you it

36:55

feels like in founder terms, you know

36:57

the founders live many many lifetimes every

36:59

single year. Yeah What would you say

37:01

to that person right now when I

37:03

started I was still a little bit

37:05

extensively motivated But I didn't really internalize

37:07

that or like I'm doing this just

37:09

purely for the love of building and

37:11

then that is sort of the the

37:13

maturity journey that I've been on when

37:15

I think the first year or two

37:18

years or three years of the company

37:20

when everything was breaking and we were

37:22

gonna die multiple times, right? At

37:24

that point, you just basically have

37:26

a gut check where like, what are

37:28

you really doing this for? And

37:30

if you're doing it for, you know, money

37:33

or like... know, or or fame or all this

37:35

other nonsense, then there are far easier ways

37:37

to do that. And I joke with my team,

37:39

like, you know, there's far easier ways to

37:41

make a buck than to build this crazy contraption,

37:43

right? But yeah, so I would basically tell

37:45

them that, you know, 17 year old version or

37:47

whoever else watching this that like, really do

37:49

it for the love of building it, that should

37:52

be the end goals. The end goal should

37:54

not be I build so I can make XYZ

37:56

or I can get ABC out of it. But

37:58

I'm building so that I have the opportunity to wake up in

38:00

the morning and build again. that's basically the

38:03

mindset that I'm in right now every day I just

38:05

wake up and I'm like, wow, I'm

38:07

so excited for the next problem statement. And so it's

38:09

just great. Adit, thanks so

38:11

much for spending time with me today. This is awesome.

38:13

Yeah, likewise. Thanks, Gary. It It was was great talking to

38:15

you.