Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:06
Welcome to Prime Ventures Partners podcast
0:08
, a podcast for entrepreneurs looking
0:11
to build and grow their startups . Learn
0:13
about uncommon strategies and common
0:15
traps from makers and doers
0:18
of startup ecosystem .
0:25
Hello and welcome to a very special edition
0:28
of Prime Ventures podcast , where I am here
0:30
with my partner
0:32
and colleague , amit Somani . I am delighted
0:34
to be here . Welcome , amit . Thank you , shri Pati
0:36
, looking forward to it . So today
0:38
, our episode is bad
0:41
math in startups and
0:43
we are going to talk about how
0:45
, in various ways , we see math
0:48
being twisted , turned , otherwise
0:51
butchered Maths
0:54
in inadvertent ways
0:56
by founders
0:59
, and our goal in this conversation
1:01
is to highlight some of that and
1:04
perhaps offer our point of view which might
1:06
be helpful to our listeners
1:08
. So , without
1:11
further ado , let's jump
1:13
right into it . So
1:15
, amit , the first thing which I would like to kick
1:18
this off with is a conversation which
1:20
we have in almost every
1:22
pitch and we open with that which
1:24
is the market opportunity
1:26
, or TAM . So
1:29
what are some of the pitfalls
1:31
and mistakes we see in that when
1:34
we are talking to entrepreneurs ?
1:36
Absolutely Shri Pati . So the TAM , as
1:38
it's called , total Addressable Market , is
1:41
definitely one of the several
1:43
pet peaves and we'll cover many of them as
1:45
we go through the rest of the podcast . And
1:48
my biggest challenge with TAM is that most
1:50
people will just do a very simplistic
1:52
top-down model . Let's say
1:54
you were a food tech startup , you'd be like
1:57
there's 1.4 billion people
1:59
in India . Only 10%
2:01
of them ever use this and they eat three
2:03
meals a day . That is , you
2:05
know , 140 million times three . You
2:07
know that's 420 million transactions
2:09
per day And even if I only got
2:11
like a dollar per transaction , this is a trillion
2:14
dollar opportunity . Very
2:16
bogus , right , because that is not
2:18
really the TAM . The ideal
2:21
customer profile may be you
2:23
know an Indian who eats three times a day
2:25
or is mobile and wants to order
2:27
online or whatever it is , but
2:29
you really have to take a bottoms up perspective
2:32
to TAM right , saying what's
2:34
the ideal customer profile , what
2:37
is it that you know their willingness
2:39
to pay for this is and
2:41
what is your ability to address that
2:43
customer right In that total addressable market
2:45
? the addressable part is very interesting and
2:48
relevant . In fact , if
2:50
I go a step deeper , i would even say that the serviceable
2:53
addressable market , which is to say that the addressable
2:56
customer segment , the willingness
2:58
to pay and your ability to reach that customer
3:01
eventually , you know , acquire
3:03
that customer is very important , so that is one that
3:05
really gets me in
3:07
terms of .
3:08
So let me ask you are you really distinguishing
3:10
between a top-down
3:12
and bottom-up , or are you still talking about how to do a better
3:15
top-down analysis
3:17
here ?
3:18
No , i am actually distinguishing between top-down
3:20
and bottom-up . I think top-down
3:22
can be indicative that
3:24
in terms of like a litmus test
3:26
or a laugh test , right , that , okay
3:28
, there's something here . When you're
3:31
building a revenue plan or an annual operating
3:33
plan or a three-year , you know how
3:35
am I going to build this business right
3:37
For this round of financing or the next round
3:39
of financing , you're not going to say I'm going
3:41
to serve India . You're going to say I'm
3:43
going to start in Bangalore or in Delhi and
3:45
CR . I'm going to start with this demographic
3:47
, i'm going to use this channel to get to
3:50
these customers . I believe customers in
3:52
tier one will happily pay a 99
3:54
, you know rupee subscription or a
3:56
transaction convenience fee . So
3:59
I'm really talking about building a TAM model that's
4:01
bottoms up . I think the best
4:03
models are triangulated , right , Just
4:06
like an annual operating plan . So
4:08
it's totally fine , it's dismissing or pooping
4:10
top down at all levels . but
4:12
the best models will be let's do it
4:14
top down , let's do it bottom up , let's do
4:16
it channel based . Let's do it , you know
4:18
, based on the competitive or the existing money
4:21
that people are spending . you know , on
4:23
eating , just , i just randomly hooked onto
4:25
food , but in that category . But
4:28
I would say if you have to pick one model , then I'm definitely
4:30
a big fan of the bottoms up model .
4:33
Got it . So let's take this , just
4:35
the food subscription business here , right
4:37
? So would you say
4:39
that if a company
4:42
is coming and pitching and you're
4:44
listening to their pitch , would
4:46
you also like to combine your
4:48
the company's go to market when they're
4:50
actually thinking about the time or they just
4:53
you know , you'll say , Hey , look , in an optimal
4:55
go to market case , I would actually do it .
4:57
Excellent , excellent nuance . Question Tripati . I would expect
5:00
no less right . I'm looking
5:02
for initially the customer segmentation
5:04
. So for the time , actually just the segmentation
5:06
is enough . Saying tier one , India
5:09
, who makes more than 10 lakh rupees a year
5:11
, That's my target demographic for
5:13
this particular subscription . Offering right
5:15
, That's fine . I don't need
5:17
to know how will you reach them , whether you're going to use , you
5:20
know , Google or Instagram or my gate
5:22
or whatever right That's , you know . That's
5:24
a second order of discussion and
5:26
detail . So , not looking for the full GTM
5:28
, but looking for a sharp customer segmentation
5:31
and clear profiles in each
5:33
of those segments and the willingness to pay , because
5:35
that really gives me a proxy for how much
5:37
you know , the total price
5:39
here could be in terms of the addressable
5:42
revenue opportunity .
5:45
Okay , got it . So
5:47
what would you
5:49
say , makes for an attractive
5:51
time .
5:53
Actually it's basically a summary of what I just
5:56
said right , clear , segment , clear
5:58
, you know , willingness to pay and then
6:00
you know the second order thinking is the ability
6:02
to reach those customers . My
6:05
simplest rule of thumb that we share with the team at Prime
6:07
internally is if you are going to need
6:09
like a PhD degree to compute
6:11
TAM , it's probably not a good TAM , right
6:14
? If you're frivolously saying , look billion people
6:16
or you know 10,000 startups will buy
6:18
my SaaS software only if 10%
6:20
bought , that's also not a good thing . But the simpler
6:23
the model is in terms of usually
6:25
, like I'd say , fit it in one line with 40 point
6:27
Calibri , font right , saying X
6:29
multiplied by Y , multiplied by Z , unambiguously
6:32
is a billion
6:35
dollar market or a half a billion dollar market or a five billion
6:37
dollar market . Like those are the easier ones to
6:39
, and of course doesn't mean that you
6:41
won't dig into it further , but just a
6:43
simple , you
6:45
know definition or a formula is
6:47
what is a good TAM ?
6:49
Which means it's very directly
6:51
connected to your price points .
6:53
Yes , price point , total
6:55
number of customers or businesses you're going to
6:57
reach and some notional ability
7:00
to reach them . Right , because
7:02
you could say , look , i'm going to address everyone . Everyone
7:04
is a target customer . Usually bad idea right
7:07
, not just for GTM reasons , even for TAM reasons
7:09
, because even to address everyone , you
7:11
need to reach everyone , you need to have a brand for everyone
7:13
, you need to have a distribution strategy . So becomes more
7:15
challenging .
7:16
And so what would be a good number to
7:19
make it attractive from an investor
7:21
perspective ?
7:22
Look , every investor will look at it
7:24
differently , but one thing that's common to all
7:27
venture capital is people are looking for large
7:29
, addressable markets . Right , because venture is not
7:31
a business that is a cash
7:33
flow business or a you know , like
7:35
a day
7:37
to day profitability kind of business . To begin with . Right
7:40
, you're building for a large outcome . So
7:42
rule of thumb , i would say , is anything less than a five
7:44
hundred million dollar TAM is probably a non-starter
7:47
to begin with , especially
7:49
if you're at an early stage of your company , because one thing that
7:51
we won't get into today is some
7:53
of the best entrepreneurs will also expand
7:55
the TAM by going into system markets
7:57
and all that . So for today's podcast may not
7:59
be in scope , but you can cover that later
8:01
. But I'd say at least 500 million to
8:03
a billion dollars of bottoms up addressable
8:05
TAM is a good number
8:08
, and the reason I say that is for most companies
8:10
to get to 15 , 20 percent market
8:12
share is a daunting proposition
8:14
, right ? Like you know , microsoft
8:17
in the late nineties got to 20 percent market share
8:19
and got the FTC after them Right
8:21
. So it's not . Everybody says of course I'll get 20 percent of
8:23
the market because I'm the best , but it doesn't
8:25
really happen . So I'd say that basically makes it
8:27
that your obtainable revenues at least hundred million dollars
8:29
in some scenario , say over
8:31
five , seven years .
8:33
So before we get off this , you know
8:36
the TAM subject . What if I come
8:38
in and say that , hey look , i'm actually doing
8:40
a category , creating company right
8:42
In which , at prime , we really
8:45
look forward to evaluating
8:47
and funding , and we have done that several times in
8:49
the past ? So in those cases
8:52
you could say , look , i'm creating a category , maybe
8:54
the time is only 250
8:56
million , but I will still
8:58
actually reach the hundred million
9:00
of my revenue in a in
9:02
some reasonable period of time . How would you , how do you think about
9:04
that ?
9:05
Yeah , no , wonderful question . Actually
9:07
, I'll take it back to product for a second and I'll
9:09
come back to the time . There
9:11
is almost a Jack Dorsey as this quote
9:14
that nothing is invented new on the internet
9:16
. It's all the same human
9:18
needs , the same business needs . I need productivity
9:21
, i need creativity , i need quality , i need
9:23
you know measurability , i need transparency
9:25
, whatever it is that mixing B2B and B2C kind
9:28
of companies here . So really
9:30
, there will be some proxy for
9:32
what you are thinking you're unleashing , right
9:34
, whether it is an iPod or an iPhone or a Sony
9:36
Walkman from days past . Right , you
9:38
say , okay , there's people like listening to music
9:40
and therefore they're already spending XYZ
9:43
. Now you could say I'm going to expand
9:45
the number of people that get to listen
9:47
to that kind of music or use that kind of device
9:50
or are much more mobile than they
9:52
were before . And but to
9:54
say that you know there is zero proof
9:57
and this is like a brand new thing
9:59
. And , by the way , that does happen from time to time
10:01
. Right , we're sitting on the chat GPT revolution
10:03
and there's so many products , right , but for every
10:06
chat GPT or something else , there's also Segway
10:08
, scooter , where you know there
10:10
was no real kind of demand , right ? So
10:13
I would just say , look for proxy
10:15
signals , for is there any spend
10:17
on that category ? Is
10:20
there , you know , the the
10:22
customer sort of economic background
10:24
? Is it relevant , whether it's on the business side or on
10:26
the consumer side ? right , to be able
10:28
to pay for that ? And you're right , there
10:30
are some absolutely truly category creating
10:32
companies where it will
10:34
be difficult to complete the time . And
10:36
you , literally , even as us , as we sees , we
10:39
have to go and back them because they're solving
10:41
a meaningful problem and someday we'll figure out how we'll
10:43
make money .
10:44
Fair enough And I think I would
10:46
say one thing to avoid
10:48
for entrepreneurs is to not do
10:50
a gold seek on on Excel .
10:53
Yes , We've seen that . I
10:55
have seen you know magical numbers like 1.18
10:58
and then you do some regression and it's like
11:00
like wait , this is just extrapolated from
11:03
this thing . So yeah , i would avoid that too . So
11:06
you know . One other thing I'd want to talk about Tripathi is the CAC
11:08
right ? If he's sort of go down and then
11:11
and on the CAC , maybe I'll start and you and you can chime
11:14
in , right That because related
11:16
to the time question and then you said Hey
11:18
, what about GTM ? So
11:20
on the CAC one , i think the challenge
11:22
that I find is that , obviously
11:24
, initially to bootstrap the
11:27
funnel right , because demand you have to create and
11:29
you have to leverage you'll go spend
11:31
some money on performance marketing . Now
11:34
it could be , you know , google or Instagram
11:36
or whatever it is that you're spending money on
11:38
. However , one
11:41
of my pet peeves about CAC for a company that
11:43
has a little bit of traction , it could be a few hundred KRR
11:46
or half a million ARR or whatever it is is
11:48
people always talk about blended CAC , like
11:51
my CAC is $200
11:53
. Okay , it's
11:55
$200 . For what segment ? For
11:58
what channel ? For what price ? prime
12:00
of the product ? right , and
12:03
typically it's like no , no , no , it's everything
12:05
, like all . But no , you , you've
12:07
been around for a while , so you just started spending
12:09
on marketing three months ago , so earlier you were
12:11
acquiring customers organically
12:13
or through referral or through your you know Facebook
12:16
group or your Insta channel . No
12:18
, no , it's all blended . That's not
12:21
, that's not really your CAC . Right , because
12:24
and we'll talk about averages and and medians later
12:26
because cost of customer
12:28
acquisition is very relevant to
12:30
a channel or a or a segment
12:32
. Right , because depends
12:35
on where you're acquiring , so you might actually find
12:37
that your paid CAC to
12:39
distinguish the term is actually $600
12:42
because it's very expensive to get
12:44
people on these channel . However
12:47
, your organic thing is zero , by
12:50
definition , it's organic , whereas
12:52
your SEO CAC , which is a longer level
12:54
gestation period , like you build on SEO
12:56
or ASO . You're spending money
12:58
on content marketing , on SEO , but
13:01
you'll realize the results six , nine months later
13:03
, right , so it's very , very different
13:05
. So it's very important to know what
13:08
is the . At the very least , i would , you
13:10
know , suggest to entrepreneurs to say what
13:12
is the paid CAC for all the paid channels
13:14
, if you must blend it , and what is
13:17
all the organic channels ? right , and
13:19
when did you start what channel , at least at a broad
13:22
level , at monthly , quarterly , whatever it
13:24
is . And one thing that I would I
13:26
would strongly recommend that that really kind of bothers
13:29
me , this blended versus it , and related
13:31
to that , is the fact that CACs
13:33
don't stay stable or grow
13:35
linearly . So it's not like , well
13:38
, i have my CAC used to be 500 , now
13:40
it's 300 . So , therefore
13:42
, once I grow even more , it'll be 150
13:45
. No , it won't . In fact
13:48
, if anything , unless you're in a massively
13:50
monstrous category which is completely
13:52
untapped , and in a category creating kind of startup
13:55
, you will find lot more competition
13:57
comes in . You will find some harder to reach customers
13:59
, you will find customers that are the late majority right
14:01
To use Jeffery Moore's framework
14:03
, et cetera And therefore your CAC . Actually
14:06
I believe it could increase , even though
14:08
your customer base is also increasing , but
14:10
the ratio at which you have to grow So
14:12
I think thinking about CAC not just as
14:14
a static point , but
14:17
seeing how will the CAC evolve as you go
14:19
along Totally is dependent
14:21
on the scalability of the channel that you're using
14:23
and the readiness of the customer to pay for
14:25
it . So those are some thoughts , any thoughts you have .
14:27
No , just a comment there And I think the
14:30
folks who actually to spend a lot on performance marketing
14:32
will probably identify with this is that other
14:34
reason CAC increases is
14:36
because the keywords which you are bidding on start
14:39
moving from the long tail to the more torso
14:42
of those keywords
14:44
, meaning that there's a lot more competition . So
14:46
if at FinTech you might initially
14:48
be saying that hey look
14:50
, i'm a savings product for
14:53
people who are students in
14:55
studying engineering
14:57
And I'm just making this up right And
14:59
the set of keywords to attract that is different , but as
15:02
that particular customer cohort gets
15:04
exhausted , you might have to
15:06
start advertising on just the savings product
15:08
And then the click through
15:10
required and the bid required on Google
15:12
to actually get those clicks . Now
15:14
you're competing against HDFC Bank , for instance
15:17
.
15:17
Absolutely . By definition
15:19
, whenever you optimize
15:22
your marketing spend , you'll optimize to
15:24
the most efficient frontier . So
15:26
every time you need to scale it 10x , right
15:29
? If you are at 100,000 B2C customers
15:31
and you go to a million , exactly
15:33
your point comes into play , because the low
15:35
hanging fruit is gone . Now you're , like
15:37
you said , you're competing for a savings product with HDFC
15:39
Bank and ICICI Bank and 20 other FinTechs
15:42
there , right ? So I think
15:44
that is one . The other is , which is I mentioned
15:46
earlier , is you know some of these customers
15:48
may not be for , especially for category defining
15:50
or creating products . They may not already be out there
15:52
searching and all that . So those are very
15:55
expensive to acquire . Because there you're trying to convince
15:57
them that you need this new mouse
16:00
trap , right ?
16:01
The intent is not really there
16:03
.
16:03
Intent is not expressed for sure And
16:05
therefore , for example , in CAAC , attribution
16:08
is important , right . So I may
16:10
have seen an Instagram reel one
16:12
month ago and then saw , clicked an
16:14
ad on TripAdvisor or make my trip , you
16:17
know , two weeks ago , And now I come on some
16:19
third channel and I convert . Now
16:21
, where are you going to attribute that cost of
16:25
what you spend on Insta ? you
16:27
know TripAdvisor , whatever , and
16:29
now the last click . So
16:31
it gets complicated , but I would just say at least don't
16:33
do blended CAAC , do paid versus
16:35
this and then have segmentation as
16:38
you go along and you know , ability to express
16:40
the scalability of the channels .
16:42
So maybe one more thing before we move on from the whole CAAC
16:44
equation , which is that how do we answer
16:46
when entrepreneurs say
16:48
that , hello , i'm doing a lot of advertising
16:50
on the SEM
16:53
side , but that
16:55
is leading to organic , ultimately right
16:58
. So that's the reason why I'm
17:00
blending the two . So how do we
17:02
think about ?
17:02
that Great question . So I think the timeliness
17:05
and literally kind of this click
17:07
, first click versus last click attribution
17:09
will lead into that . The simplest
17:11
measure of organic certainly on the search side and
17:13
now that is being disrupted right Through Janiya
17:15
and other stuff , but let's stick to the old world for a
17:17
moment is that the organic
17:20
keyword searches for your brand should be increasing
17:22
, right . So people are not searching for
17:24
your brand or your unique products , USP
17:27
or whatever . You know that all the other investment
17:29
you made in brand marketing or performance marketing
17:31
or SEM is not kind of yielding
17:33
results . So the simplest way I would say
17:35
is that timeline is very relevant of when you
17:37
did the spend . If you did an IPL
17:40
marketing campaign , like three
17:42
months ago , right , it's gonna
17:44
be hard to attribute that in the month of June
17:46
2023 or July , maybe , right
17:48
, 2023 . So
17:50
you can keep some rolling window right . I
17:53
would say like a quarter . Maybe Most
17:55
people will be . I would say should be more conservative
17:58
and do it in the last 30 days or maybe 60 days
18:00
, but you can use a time window to do it And
18:02
, by the way , over time it'll be a rolling average right , because
18:05
what you're gonna spend now on
18:07
SEM or brand should have an impact on organic
18:09
later , so it'll wash out beyond
18:12
a certain timeframe .
18:13
Fair enough .
18:14
Yeah , great . So
18:16
let's switch gears from like a bit to the top of the
18:18
funnel to really
18:20
just going the other end , you
18:23
know , at the bottom of the funnel , which
18:25
is , let's say , you've got the customer , they're paying
18:27
something . Now they
18:30
are beginning to churn , and I know that's one
18:32
of your pet peeves and you've even written a lovely blog
18:34
on it . So , how you
18:36
know , what do you look for when you look for churn and what annoys
18:39
you when you see the kind of lamins
18:41
or the vanity churn metrics that
18:43
people usually throw at us .
18:46
Yeah , so churn
18:48
really is . Whenever we look at metrics
18:50
, we have to ask ourselves what is the objective of that metric
18:52
? right , and the case of churn ? the
18:55
objective is to for us to
18:57
figure out , as an entrepreneur , whether
18:59
the value we are delivering to the customer
19:01
is sticking , because
19:04
the customer churns when they are expecting a certain
19:06
value and they don't see that value and
19:08
they say , okay , i'm not using the service , right ? So
19:10
that's the objective . So I understand how well the value
19:12
proposition is sticking And so
19:14
I would say it's true for all metrics
19:16
. Like you need to be clear about what the objective is , because
19:19
taking a particular definition and then just cookie
19:21
cutter pasting it into your business usually
19:24
results in just the wrong set of numbers And
19:26
garbage in , garbage out you'll make . If you're
19:29
actually not using those metrics to influence
19:31
or inform your decisions , it's really
19:33
not a very useful metric . Okay , so in the case
19:35
of churn , the first thing I would say
19:37
is just imagine a SaaS business
19:39
where you know you have usually monthly
19:42
subscriptions , you have quarterly subscriptions
19:44
, you have annual subscriptions , right ? So
19:46
let's say you have , you're selling annual subscriptions , So you
19:48
sell an annual subscription in January And
19:50
now next year you
19:53
start calculating and a month later
19:55
, sorry , in January of this year you sold 100
19:57
subscriptions And they're all annual . What
19:59
is a churn in February , march , april
20:02
, may and June ? Zero , absolutely
20:04
. Right Now you could come and say , look , i have
20:07
zero churn . And that
20:09
is like the first thing which I feel is
20:11
is probably giving the wrong
20:13
, wrong inference there
20:16
. And the way to think about it
20:18
is that the customer has no choice . In one sense , they
20:20
have already prepaid for it And they're
20:22
not requesting a refund . And
20:24
there are other ways to actually , you know , account
20:26
for that by looking at hey Lord is engagement and
20:29
so on and so forth of the customer . But
20:31
let's ignore that for a second . One
20:33
clean and simple way to look at it is to
20:35
calculate churn based on up for
20:37
renewal basis , and what
20:39
that means is that at any
20:42
particular month you look at
20:44
all the customers who could
20:46
have or who are up for
20:48
renewal , who needed to renew their subscription
20:50
. So let's say you're in February and
20:53
now a whole bunch of annual plans which you
20:55
sold previous February and
20:57
a whole bunch of quarterly plans which you sold
20:59
previous November and a whole bunch
21:01
of monthly plans which you sold in January . They're
21:04
all all coming up for renewal . Correct Right . So
21:07
you take that as the denominator And
21:09
in the numerator is all of those
21:11
folks who did not renew . So
21:13
there's a very active decision being made by
21:15
the customer to
21:18
renew or not to renew And when
21:20
they are actually paying and voting with their
21:22
wallets , that is a right
21:24
measure of both retention and churn . So that
21:26
, i would say , is probably the most common
21:29
mistake . I see , and you see it consistently
21:31
in SaaS companies specifically , much
21:34
more so than , say , b2c or any other companies
21:36
, because the subscriptions keep varying
21:38
the time periods . On that Completely
21:41
.
21:42
Thank you , shri Pati . Ask
21:44
people to segment it . Let's say I only
21:46
have two plans , an annual plan and a quarterly
21:48
plan . Yeah , would you suggest
21:50
that they do a quarterly plan
21:53
churn metric differently than the
21:55
annual plan churn metric ? Or would you still say
21:57
, look , you know , renewal is a renewal
21:59
. It doesn't matter when it came up for renewal
22:02
. How would you think about that ? Or , you
22:04
know , you can make it more complicated with the monthly as well .
22:06
Yeah , yeah , i mean it's like similar to the segments
22:08
and CAC which you mentioned earlier , the channels . Actually
22:10
, it is useful to do it by the
22:13
contract , by the duration of the contract , because
22:16
consistently you'll find that and
22:18
there's a fairly common experience
22:20
of startups is that moving customers
22:23
from monthly to quarterly , for instance , immediately
22:25
sees a reduction in churn and also
22:27
follow on churn , it's a higher
22:29
retention , because in one sense
22:31
the cognitive load on the customer
22:34
is reduced from a
22:36
monthly decision to a quarterly decision . So
22:38
I would say that doing that is useful because
22:40
that will help you decide . If
22:43
you want to , for instance , completely scrap your monthly plan
22:45
, correct , it will mean that a lesser number of folks might
22:47
sign up , but by scrapping you might just be
22:49
getting much better retention .
22:52
Great . Another one that
22:54
we often talk about right is this
22:56
notion of , let's say , in particular in SMB
22:58
, saas , right , which is not necessarily always
23:00
the Vogue , where the end
23:03
customer itself goes out of business , right
23:06
. So it's you know , and
23:08
therefore it's not like they renewed , or
23:10
do you double click a little bit on that
23:12
and try to see did they go to another
23:14
competitor or another offering in as much
23:16
as you know ? or perhaps
23:18
you know they went out of business And so
23:20
would you count that in churn or would you count that
23:22
as a no longer using and
23:24
therefore , you know , i get a little credit .
23:27
Yeah , i would actually say it's important to distinguish between
23:29
that and that is specifically true for SMB businesses
23:31
, because that is
23:34
, in one sense , involuntary
23:36
churn , for lack of a better word . Right , yes
23:38
, so I would actually remove that . But
23:40
if you have a large proportion of it , which
23:42
frequently happens in SMB , you might have to
23:44
ask yourself whether you're targeting the right segment in the
23:46
first place . Correct , and
23:48
in SMBs , you know , some things don't
23:51
have a high churn , but most things have a high churn
23:53
precisely for the reason which you're talking about . So
23:56
I think it is . It affects the business
23:58
, it tells you whether you're on the right segment
24:01
or not . if you're , you know if most of your
24:03
customers that actually have very high
24:05
mortality rate , so to speak . But
24:07
it doesn't help because I
24:09
answer the question whether your value is sticking or not
24:12
, because your value might be sticking and the customer is
24:14
going out of business . So it's helps to actually
24:16
just split them into two , got
24:18
it ?
24:19
Got it . So , moving along
24:21
, there's one thing that is very closely tied
24:23
with churn right , Which is the customer
24:25
lifetime value , or the LTV as
24:27
people call it , And I think this will also have a
24:29
bearing on the B2C side , which we
24:31
didn't cover in churn , which is
24:33
, you know , B2C setting . You know I'm not
24:35
buying every day or every week . I mean , maybe some businesses
24:38
are like that where there is an opportunity to buy every day
24:40
. I may be buying three times a year , four times
24:42
a year , etc . So how do you think
24:44
about LTV and your you know kind of
24:46
challenges and myths around LTV in general
24:48
, And then we'll dig into B2C a
24:50
little bit as well .
24:51
But yeah , So probably
24:54
the biggest
24:56
, i would say , confusion in entrepreneurs
24:59
mind is that what actually is LTV
25:01
? right , you might seem like a very simple thing lifetime
25:03
value . But let me give
25:05
an example . Right ? So suppose
25:08
I'm in the business of selling this mug here
25:10
and this costs me high hundred dollars
25:12
, a very expensive gold mug , right
25:14
, and I decide to set up a site
25:17
in which I'm going to sell it for $50 , right
25:19
, yeah , so now I might
25:21
actually get a lot of demand for it , right
25:23
? So now , if you buy it , what is the
25:25
lifetime value of this customer ? Right
25:28
, and the answer would be well
25:30
, i'm getting $50 from this customer . So
25:33
the frequent and
25:35
common mistake here is to use
25:37
that $50 as the lifetime value . Let's , for the moment
25:39
, say the customer only transacts once . The
25:42
customer is actually a negative 50 lifetime
25:44
value , because for every cup
25:46
which you're selling , you're losing money , and
25:49
this is not uncommon . In early days of e-commerce , we had
25:51
negative gross margin in a lot of cases . Right
25:53
, yeah , promotions also in a
25:55
number of businesses . So
25:57
the lifetime value is
26:00
the value piece of it , right , we'll come
26:02
to the lifetime in a second . But the value
26:04
piece of it is actually contribution
26:06
which you as a business is getting from
26:09
that customer . So it must
26:11
must remove your
26:14
costs to both provide
26:16
the product and service that
26:18
customer . So the way I think about to
26:20
simplify things is to remove all variable
26:22
costs associated with that
26:24
customer . In many cases we'll find
26:26
that by the time we acquire that customer , by
26:29
the time and the the CAC is not part of LTV
26:31
because we take it as a ratio , usually LTV
26:33
to CAC . But you have to remove all
26:35
COGS , you know service costs
26:38
, customer service costs , all variable
26:40
costs associated with that customer , and then the
26:42
remaining value is the value
26:44
. And so frequently when you do that the
26:46
so-called LTV to CAC ratio will
26:49
look dramatically different than otherwise .
26:51
So let me , let me come back to you
26:54
and say , let's say that $100 mug I was
26:56
selling day zero for $110 .
26:58
Yeah .
26:58
And the cost of all the other variable stuff is not
27:01
more than $10 . Just , we'll just make an assumption
27:03
right COGS is $100 and then the rest
27:05
of it , do you ? how
27:07
do you still think about kind of LTV
27:10
in that context ? Because I may just buy
27:12
this mug once in life and never buy it again
27:14
. Yeah , What durations do you
27:16
use , you know ? are there any good
27:18
practical ? everyone will say , oh my goodness
27:21
, i sold Shri Pati a mug . He's going to buy 1000
27:23
other mugs from me over the next five years . Yeah
27:25
, shri Pati may never , ever even come back to the site because you
27:28
just saw something clicked it's somebody's birthday coming
27:30
up or whatever and bought it , right .
27:31
Yeah .
27:32
So how do you think about the timeliness of
27:34
the LTV , assuming that the basic
27:36
condition of value is met ?
27:37
Yeah , I think that's a difficult
27:39
question for entrepreneurs to answer because
27:41
they have usually not been in business for too long Exactly
27:44
To actually know that . So let's actually take
27:46
the case of both a B2C and
27:49
a B2B . In the case of B2C , you'll
27:51
have to make educated guess on
27:54
why on earth will that customer come again
27:56
for this mug ? If it's a consumable
27:58
, you can say , hey , look , when I come , what
28:01
is my repeat rate ? Will they actually come
28:03
often enough ? If they're actually buying toothpaste
28:05
, they'll probably come often enough . But if they buy
28:07
something which lasts them a year , then
28:10
actually you'll have to reacquire that
28:12
customer . This frequently
28:14
happens , for instance , in fashion Absolutely
28:16
Right . So you are even in travel
28:19
, right ? If you're actually doing international travel or something
28:21
like that , you'll have to end up reacquiring
28:23
that customer . So the repeat rate here becomes
28:25
really , really important And the
28:28
entrepreneurs have to focus on proving
28:30
and validating that the repeat rate they are presuming
28:33
and the lifetime they are presuming actually
28:35
is correct . And so for
28:37
B2B it's a slightly easier answer
28:40
by actually changing the question itself , and
28:43
usually my suggestion to entrepreneurs
28:45
is to change the question from a
28:47
lifetime value to payback For
28:50
B2B . For B2B , right , yeah , how
28:52
soon is my cost of acquisition
28:54
paid back by the customer's contribution
28:57
, And if that is anything
28:59
less than 12 months is a very good number . And
29:02
in 18 to 24 months you actually start
29:04
wondering because now you need to retain
29:07
that customer for so long . If you're getting
29:09
a payback in one month or in the first transaction
29:11
, you're actually in very , very good wicked , absolutely
29:14
, absolutely .
29:15
And , if I may add , on B2C , i
29:17
think again , like what we talked about earlier
29:19
, seeing the proxy for that demand
29:21
right . So , whether it is , you know , travel
29:24
versus a food startup versus
29:26
a mobility startup , like booking
29:28
an Ola or an Uber or some kind of taxi
29:30
service , you say , look , you
29:32
know , on average , a person going to office 250
29:35
days a year , 300 days a year , books
29:38
, you know mobility or shared mobility twice
29:40
a day . So what's the share of that wallet ? am
29:42
I getting Right ? So I think
29:44
there are ways to model it . But at
29:46
least my peeve is when people say , look , i've
29:48
got them for literally life , like five
29:50
years or something , and you don't really know because the customer
29:53
behavior and the innovation and the competitive
29:55
landscape changes so quickly that
29:58
you have to be a lot more reasonable about the actual
30:00
lifetime . Right , and B2C , certainly
30:02
, customers are a lot more flippant and a lot
30:04
more demanding .
30:05
Absolutely , and I would say that for entrepreneurs
30:08
, they need to be conservative in these assumptions . Yes
30:10
, because you're running a business and if you're not conservative
30:13
, you'll quickly run out of money on
30:15
this one . So making optimistic assumptions on
30:17
that is really
30:19
very risky .
30:20
Absolutely , shri Pati . Let's let's
30:22
pick it up from . You know the average
30:24
order value and
30:27
how do you kind of look at that , and
30:29
also this notion of aggregate revenue
30:31
or aggregate . You know metrics
30:33
as opposed to cohort based metrics , whether
30:35
it's for B2B companies or for B2C
30:37
. Any of your thoughts on that
30:40
in terms of what's the right way to go about it and what is
30:42
a bit annoying to look at ?
30:44
Yeah . So the thing is that when
30:47
we look at revenues , let's for the moment say
30:49
you're a gaming company and you
30:51
have a 300 or 500
30:53
K ARR Right , and
30:55
you're saying , look , i got 500 K
30:57
ARR . Next month I am 550 K ARR
31:00
, and so forth . But the challenge is looking
31:02
at these numbers is that it doesn't provide
31:04
any intelligence on how
31:07
this number is going to trend in the future
31:09
, because what we are looking for is investors
31:11
, and what really you are looking for as an entrepreneur
31:13
is what is my future
31:16
revenue and future profitability going
31:18
to be like ? And if you don't tease
31:20
apart what are the constituents of
31:22
this MRR , you will not have any
31:24
success with it . So what I mean by
31:26
this is that , let's say you have $50,000 of
31:29
revenue this month . You say my MRR is 50
31:31
K . How is that distributed
31:33
between customers you have acquired
31:35
this month , between customers who are in their
31:37
second month , those who are in their
31:39
third month , those who could be a year ago
31:43
customers , and how they are doing So
31:45
? that , for example , is
31:47
a cohort based analysis . You're looking at cohorts
31:50
, which are acquired by time period , and
31:52
you're looking at how they are performing for
31:55
their first , second , third , fourth and fifth months . And
31:58
this is really important , not only because
32:00
now you're looking at what
32:02
is the contribution from a new user , new
32:04
users but also how well
32:06
are my older users
32:08
continuing to contribute to
32:10
my game , to my revenue this month And
32:13
hence to the LTV , which is something which we
32:15
talked about in the last episode . What
32:17
is my lifetime value ? Without a cohort
32:19
analysis , just understanding lifetime
32:21
value itself is going to be very difficult . Now
32:24
, cohorts themselves can come in various
32:27
other shapes . Usually
32:29
, it is a time-based cohort . That is the most common and
32:31
usually most useful . So
32:33
a person or a cohort acquired
32:35
in January versus February , versus March
32:37
, and so forth . But you could also look
32:39
at how am I doing
32:42
with respect to things which happened
32:44
. For example , i
32:46
did a particular promotion in
32:48
a particular month or from a particular channel
32:50
. What happened ? How is that cohort
32:52
behaving ? for instance , you
32:55
might also look at how a segment like how
32:57
are my Android users versus iOS
33:00
users behaving , and so forth , and
33:02
then you can split that cohort again a time-wise
33:05
cohort , but you're not looking at just
33:07
the iOS users or just the Android users
33:09
and so forth , and this might give you a lot
33:11
of insight into what happened . What you'll
33:13
usually find is that not all
33:15
cohorts behave similarly . You'll
33:18
find that suddenly one or two cohorts
33:20
have a much higher retention
33:23
, much higher lifetime value , for instance , and
33:25
that gives insights into
33:27
what is it that the company did at that
33:30
time . You might have done
33:32
a particular channel which actually did
33:34
a lot of acquisition from there , or
33:37
you might have changed some of
33:39
the gameplay mechanics which
33:41
made sure that you had a very high retention
33:43
and a very high monetization
33:45
from that segment . So the long
33:47
and short of it is that looking at cohorts not only
33:49
just how is my revenue divided this
33:51
month , how is the first month of my
33:54
various cohorts doing , how is my retention doing over
33:56
that and how are my longer serving cohorts
33:58
doing across a time period
34:00
is really the only way you can get
34:03
meaningful insights into the business .
34:05
Absolutely . And whatever you said , I think almost
34:08
all of it applies to B2B companies as well
34:10
. I think gaming was just an example . So
34:13
I think I would and it's not
34:15
just from an investor point of view that you should
34:17
do this , I think even to run your own business
34:19
to figure out what are the most attractive
34:21
cohorts or the most attractive
34:24
segments or the most attractive channels from
34:26
where you're acquiring these cohorts . That
34:28
will be very indicative of where you should do your
34:30
capital allocation or your
34:33
intellectual bandwidth allocation . To say
34:35
, oh my god , users acquired an iPhone
34:37
seem to kill it , or vice versa . Or
34:39
users acquired from a particular channel , right , They
34:41
seem to do much better , and so forth .
34:43
Absolutely , and one of the things we talked about last time
34:45
for a SaaS business was paybacks . So
34:48
one of the things we look for in
34:50
evaluating companies is if you look
34:52
at various cohorts , how is the
34:54
payback trending over time ? So
34:57
if suppose you actually
34:59
had a payback of 12 months , is that
35:01
actually becoming shorter ? And the way
35:03
to figure that out is look at the revenue
35:06
realization from
35:08
each of your cohorts over a period of time And if
35:10
these cohorts are becoming steeper , meaning
35:12
that if the Y axis is revenue and
35:14
you're actually going to a higher
35:17
revenue earlier in the cycle
35:19
, that's indicative of a good business
35:21
trend .
35:22
Absolutely , and one of the metrics that I know both
35:24
you and I are big fans of the dollar-based
35:26
net retention . That will also
35:28
come very easily from the cohort
35:30
data , because , yes , there's logo
35:33
retention . I had 100 customers last year
35:35
, i have 110 now , two churned
35:37
, and so on and so forth , but I could also have
35:39
two churned . So I only have 98
35:41
customers , but the cumulative value of the
35:43
dollars I'm getting now is 115%
35:46
of what I was getting last year . So I'm actually more
35:48
valuable to my customers , either because of my product
35:50
offering or my pricing power or
35:52
increased number of users or licenses or whatever
35:54
.
35:55
Absolutely , and
35:57
we talked about churned last time , and it's related
35:59
to that , which is that if you look
36:01
at your revenue cohort
36:03
versus user cohort , you
36:06
might actually get other insights . So
36:08
, for instance , if you find that
36:11
your revenue retention is increasing and
36:13
you have 110% , 120%
36:16
revenue retention , but if you look
36:18
at your user retention
36:21
and you see that it's actually going down , what
36:24
that means is that your customer mix is
36:26
changing And it's going towards
36:28
a higher value customer . Now
36:31
it is a great thing if that is what you intended
36:33
to do , but not so great if
36:35
that is not something which you plan to do and you
36:37
wanted to actually have a large number of customers . Now
36:39
what's going on is that your average order value , which we'll
36:41
talk about in a bit , is increasing And
36:44
your product probably
36:46
is now moving up market And
36:49
it appeals more to a certain segment than
36:52
more than the previous what you were doing
36:54
earlier . So cohorts actually give you a very
36:56
good indication of
36:58
how things are moving in your business over time
37:00
, which averages almost universally
37:02
don't .
37:03
Yes , we'll talk about averages and
37:06
medians later . There
37:08
is one other scenario from what you just described
37:10
, the last example . That could be the case that
37:12
your dollar-based net retention
37:14
is increasing but your actual user
37:17
or business retention is decreasing . It
37:19
could be that your aggressive marketing strategy
37:22
is not working , so you're just randomly
37:24
acquiring customers because you're
37:26
trying to grow your way into your
37:28
next kind of business milestone , but
37:31
you're not acquiring effectively And
37:33
therefore sometimes I find that that behavior
37:35
tells you of the customers you don't want to
37:37
acquire because they'll be rapidly churning
37:40
or they'll not be high retaining or they'll certainly
37:42
not be high increasing in terms of their
37:45
contribution to your business . So
37:47
I think that cohorts are absolutely
37:50
the way to go and you should literally
37:52
run your business on it , not just from a
37:54
reporting perspective . Let's use
37:57
this as a segue to talk about the average order
37:59
value . So
38:02
there's a lot of this confusion
38:04
on ARPU first
38:06
, transaction value , et cetera
38:09
, et cetera . Any thoughts on when
38:11
you look at average order value or
38:13
ARPU , for whether it's a SaaS company
38:15
or anything else , how do you think
38:17
about it ? What are some of the challenges in terms of
38:20
how people report these metrics ?
38:22
Yeah , so when you're thinking about , for
38:24
a SaaS company , the ACV
38:27
, the annual contract value
38:29
is , the way you look at it , right
38:32
. And annual contract
38:34
value presumes , first and foremost , is that the
38:36
contract is annual , which
38:38
might or might not be the case , that's right . And
38:41
the second thing is that it becomes a little bit harder because
38:43
in a lot of SaaS companies you have subscription
38:46
and you have transactions , so
38:48
the total value is actually
38:51
a blend of that , really what you're
38:53
going to realize from that customer over
38:55
the year . So it is
38:57
very important to kind of both calculate
39:00
and delineate very clearly what
39:05
your annual contract value is going to be for
39:07
that customer , especially since more and more businesses
39:09
are now moving from subscription into
39:12
a usage-based model
39:14
, because , being driven by the customer side , because
39:16
they are not using , they don't want to pay for
39:18
it . It's probably the first thing
39:20
which comes to mind from a SaaS thing , but that's you
39:22
can talk about from a more B2C
39:25
perspective how you think about
39:27
AOVs .
39:29
Yeah , no , i think I
39:32
had just one more thing to kind of add like
39:34
a rejoinder to the B2B side , which
39:37
is that sometimes for certain class
39:39
of B2B companies right could be on a
39:41
mid-market or enterprise
39:43
there is a one-time set-up fee
39:45
or an onboarding fee or a
39:47
services fee or whatever , which
39:50
is not going to be a recurring revenue
39:52
. So certainly when we are looking at it and again
39:54
everything is aggregated and blended into one whole
39:56
thing , saying we are a $50,000
39:58
ACV , but when you double click on it
40:00
you're like well , 15,000 is onboarding
40:03
and services for set-up , which
40:05
will probably never happen again . Then really
40:07
the annual contracted value is really $35,000
40:10
. I'm not saying that the $15,000 is
40:12
not relevant , but that revenue is
40:14
a little . you know , blue dollars versus green dollars
40:17
. So that's one thing that .
40:19
Yeah , I think it's an excellent point , Amit , And what
40:21
I would say where I found that useful is
40:24
when customers are able to say wait
40:26
, my set-up or initial
40:28
fee covers my CAC .
40:30
Yes .
40:32
Right .
40:32
Including the variable cost of the people providing
40:34
that onboarding or whatever other service , exactly
40:37
Plus the variable cost .
40:38
Correct . So I think that's a very clean way of thinking
40:40
. To me that shows good understanding
40:42
of running a business in a capital
40:44
efficient fashion . because you go and acquire
40:46
a customer for $1,000
40:48
. But , as you correctly say , all
40:51
loaded in costs of getting that customer onboarded
40:53
, the fee you're charging is $1,000 . You
40:55
can say , look , i'm not losing money , So my payback is there
40:57
as soon as the customer comes onboard . But
40:59
, as you correctly said , for calculating LTV
41:02
for that customer now you'll have to
41:04
understand the recurring nature
41:06
of the revenues
41:08
, correct So
41:10
. but , amit , when you're looking at B2C
41:13
businesses , the
41:16
almost always companies
41:18
will be providing promotions . They will
41:20
be doing , they'll be refunds , returns
41:23
, right and going on . So what
41:25
is a clean and fair way
41:27
of accounting that ? because almost
41:30
any business with stars has to . we
41:32
will face these kind of things as the product matures
41:35
and the value proposition kind of firms up .
41:37
Yeah . So typically though I mean there are different ways
41:39
to treat it . I'm not an accountant but certainly , evaluating
41:42
it from a business point of view or from an operating
41:44
point of view , i would say I would
41:46
cleanly separate All
41:48
variable costs , right . Just like you
41:50
got revenue at the top of the Kind
41:53
of top line , you eliminate all
41:55
the variable costs with respect to cac
41:57
, onboarding , setup , whatever it
41:59
might be , payment , gateway charges , etc . And
42:02
then I would have another line item for refunds
42:05
or promotions or whatever . So I included
42:07
in the variable cost because it
42:09
is , it is a variable cost from a PNL
42:12
perspective , right ? Rather
42:14
than saying like , let's say , you just to take
42:16
a degenerate example Which is not that far off , in
42:18
certain fashion , kind of verticals , right Like
42:21
you might have 30 , 40 , 50 percent return rates
42:23
. So now I'm booking crazy revenue on the top
42:25
line saying I'm at 5 million GMV , i'm a
42:27
50 million GMV , but 25 million
42:29
dollars . That comes back right Every
42:32
every month or every whatever
42:34
unit of frequency it is . So
42:37
therefore , your revenues really not 50 million dollars
42:39
, right ? So when we
42:42
I don't think we covered the gross margin and so forth , but
42:44
maybe we can talk about , you know , gross margin
42:46
versus net margin versus contribution margin and
42:48
them Yeah this is a good time to perhaps
42:50
talk about that . Yeah , so . So I would say
42:52
that these days especially since we are recording
42:55
this in June 2023
42:57
, you know , people are not looking at GMV
42:59
and top line and all that stuff . Right , people
43:01
are looking at . In fact , there was a
43:03
little , you know , quirky quote going around
43:05
that the gross margin is the new revenue , right , so
43:08
it's not even like your cost of goods and all that
43:10
which technically , even by accounting standards , is
43:12
revenue . So I would say , in the
43:14
modern definition of a
43:16
capital efficient business , I think you
43:19
have to eliminate the Cost
43:22
of goods and obviously refunds and
43:24
kind of any of those top line kind of variable costs
43:26
right In your actual
43:28
revenue . Now , that may be a little bit harsh , but
43:30
that is the reality , because that was that revenues
43:33
never going to really come to you , right , it has no ability
43:35
to hit your bottom
43:37
line because it's right off the top , right . So I would
43:39
eliminate that and and really
43:41
look at it from that point of view In
43:44
terms of just gross margin and contribution margin
43:46
, right , gross margin is just less cost of goods
43:48
. Very simple , whatever product or service
43:50
you're providing , that's what it
43:52
is . And then I
43:54
get very annoyed with CM1 , cm2
43:57
, cm2.5 , cm3
44:00
. Look , it's very simple . Contribution
44:03
margin is Basically the
44:05
contribution at a unit economics level
44:07
per unit sold , less
44:09
all variable costs . So
44:12
, like I said earlier , payment gateway
44:14
goes from that , amortize refunds go from that
44:16
. You know , whatever
44:19
right a a cat goes from
44:21
that , onboarding goes from that and
44:23
that is your real ability to
44:25
figure out the economic Viability
44:28
of this business . Right , is there juice quote
44:30
unquote in this business or not ? And
44:33
I think all this notion of well , cm1 doesn't
44:35
take that out . This one doesn't . I mean to the extreme
44:38
level that I would even say There are certain
44:40
fixed costs That might
44:42
be amortizable on a variable basis
44:44
. We talked about it in the earlier episode , seo
44:46
. So you might have a massive investment in
44:48
SEO or doing video creation For
44:51
your YouTube channel or your in Surreal channel
44:53
, which is amortized over time because
44:55
you're using that to build brand and that's going to affect
44:57
your cat or hopefully reduce your cat
44:59
, but that needs to be amortized . You can't just
45:02
say , well , i just have ten people doing content
45:04
and that's just like . That's different , that's
45:06
a fixed cost . No , it's not fixed costs , because
45:08
if you're trying to grow the business , you might have 50 people doing
45:11
content or , if that's
45:13
not an effective strategy , might have two people doing content
45:15
and therefore there should be amortization of that .
45:18
So would you rather that we actually stop using
45:20
, you know , cm1 , cm2 , cm3
45:23
and so forth .
45:24
Hundred percent . I would just say look , there
45:26
is your top line revenue , there's your gross margin
45:28
, there's your contribution margin , which is less
45:30
of all variable costs . And if
45:32
you're a higher bar startup , i would
45:34
say even the fixed costs that are naturally amortizable
45:37
, of course , you know like office space
45:40
and you know software engineers and so on , may
45:42
be harder to amortize . Yeah over variable transactions
45:44
, because those are really assets you're building for the long term
45:46
. So just keep it simple right real
45:49
revenue , real variable
45:51
, you know I mean revenue less variable cost
45:54
and then eventually , hopefully someday
45:56
, you have a bit and Pat Right so like profit
45:58
after everything .
45:59
Yeah , and for SaaS companies that would almost
46:01
definitely include the cloud costs . Of course
46:03
, absolutely significant variable cost
46:06
is cloud cost right , So absolutely
46:08
has to include that for sure and
46:11
I would add that initially , you know a lot of companies
46:13
have credits and things like that , so those
46:15
costs are hidden initially . But
46:18
the right way to think about it is that what happens when those
46:20
promotions and credits and everything else expire
46:22
and then , and then What
46:24
is your margin ? and actually you'd be surprised that many
46:27
SaaS companies we see Do
46:29
not have that much of a very
46:32
high gross margin because the cloud
46:34
costs and now increasingly The
46:36
gendered AI , chat , gpt and all these
46:38
costs are going to be non-trivial part
46:40
of delivering the value and they all , they all need
46:42
To be properly accounted for absolutely
46:44
. So let's move to a
46:46
related topic , which is
46:49
averages and medians
46:51
. Yes , so I know that that's something which Which
46:55
you Really have
46:57
a lot of thoughts on , so let's , let's hear it .
46:59
Yeah . So I think again , we've
47:01
covered this in both the episodes right that
47:03
don't look at aggregates , look at cohorts
47:05
, don't look at total revenue , look at revenue
47:08
by segment , don't look at total cac , etc
47:10
. And this should be leading the witness to
47:12
to kind of the jury , in this case , to
47:15
to what the outcome is . Averages
47:18
are extremely mid-lead misleading for
47:20
any metric , right ? We looked at a few in
47:22
the past and give a few more examples and the reason
47:24
they're misleading for any metric . It could be
47:26
engagement time , it could be session
47:28
time , it could be , you know , whatever other metric
47:30
which are very relevant may not be purely
47:32
from a PNL perspective , but from
47:35
a product success perspective . Right , because
47:37
they hide the behavior of the
47:40
power users Versus
47:42
the name , or , you know , new onboarded users
47:44
versus your typical user
47:46
. Right , because averages
47:48
and medians are synonymous when they , when
47:51
the distribution is uniform . Right , or
47:53
like a more like a normal distribution ? Yeah , when
47:55
, but in early stage startups , probably
47:58
even later stage or mid-stage startups Distribution
48:01
is hardly but uniform . There are certain
48:03
set of customers that will absolutely love
48:05
your product and double down on it and spend hours
48:07
. Right , we have couple of gaming companies in
48:09
our portfolio where two , three , five percent
48:11
of users will dominate right . Yeah , our
48:14
users number of games , times in app
48:16
purchases , etc . But if you did an
48:18
average of somebody who spent a dollar
48:20
per week or per month
48:22
on a game and somebody else who spent $500
48:24
real example from real company right . But
48:27
you do the average of those two users 250
48:30
dollars and 50 cents , extremely
48:32
misleading , right . So I would
48:35
say you should , and and medians
48:37
, of course you know will work better with skewed
48:39
distribution , which is the case right . Also
48:42
, i would basically appeal to the founders that it's
48:45
also a good way to run your business . So you need to
48:47
know where the money is coming from
48:49
, right , where the engagement is coming from , where the
48:51
retention is coming from . So , in
48:53
even beyond a median right
48:55
, i would which is the highest frequency
48:57
, sort of recurring data point , i would
48:59
say look at the 90th percentile user
49:02
and Or the business
49:04
that that you're serving and the 10th
49:06
percentile user . That gives
49:08
you a true picture . So it's almost like three
49:10
quote-unquote , three medians , right . Or , if I drew
49:12
a graph right , this
49:14
is That at
49:16
the 90th percentile , which is the highest engaged
49:19
users or businesses or whatever . This
49:21
is my retention , this is my revenue , this
49:23
is my AOV or ACV , etc
49:26
. At the newest right
49:28
and I'm talking newest in the sense of either
49:30
could be onboarding but , like you said , on a time basis
49:33
, or could be by how much they're
49:35
producing for me . So I could have a gaming user
49:37
who's been with me for a year and has only
49:39
ever spent $5 or 50 cents . Still
49:42
a very loyal user , but they're
49:44
in that 10th percentile . Yeah , if you do
49:46
the cohort not by time right . So
49:49
I would definitely say that those are couple of things
49:51
that I would really Encourage people to
49:53
do is look at medians , look at 90th percentile
49:55
, look at 10 percentile And , in fact , run
49:57
your business on it , because that will lead you to the
49:59
ones that you want .
50:01
Actually , one of the terms which is used
50:03
in SAS metrics is this thing called the smile
50:05
curve . Yes , right , which , if you look
50:08
at , if you draw a curve
50:10
wherein on the suppose There's
50:12
a product which is being used on a daily , weekly
50:14
basis , and on the x-axis
50:16
you Plot the number
50:19
of the users were coming once , twice
50:21
, three times , four times , so many times
50:23
. Like to think of a game , if people who come daily
50:25
would be Clocking at the 30
50:28
times a month They are coming , right
50:30
on the y-axis is , of course , the number of people and and
50:33
really interesting companies
50:35
and SAS companies have this smile
50:37
shape to this curve wherein you have a large
50:40
number of folks who are coming for Just one or two or three
50:42
days in a month , so they are the
50:44
infrequent users in your , in your distribution
50:46
, and then it goes and
50:49
as the number of days goes in , increases , it
50:51
goes down , but The
50:53
there are a small fraction of power
50:55
users who are coming at
50:57
the 28 , 29 , 30 days a month
51:00
, right , and it again trends up . So
51:02
that trending up is a very healthy
51:04
sign , because what it
51:07
indicates is that the
51:09
focus on acquiring that kind of customer
51:11
becomes the focus of the company , because
51:13
Understanding what is the behavior of that customer
51:16
? Which channel are they coming from ? What
51:18
are the things which can be done in the product
51:20
to incentivize that kind of behavior ? Becomes
51:23
like the gives us an insight so
51:26
that it really leads to a high value business .
51:28
Absolutely , and I think the exact same thing applies
51:30
to gaming startups , content startups any
51:32
kind of beauty engagement startups , because
51:34
you'll acquire a lot more at the top of the funnel
51:36
and then you will have some
51:39
PMF right for the ones that
51:41
really liking it . And then really the
51:43
ones on the on the other end of the smile curve Will
51:45
be the ones who are the highly retained , highly engaged
51:48
, highly paying customers right .
51:49
So I think that behavior is quite common
51:52
to To to be to see
51:54
as well exactly and how
51:56
that kind of curve trends over time Gives
51:59
a very good insight into how the business health is
52:01
.
52:01
Yeah , that's . That's another very good point , right ? Oftentimes
52:03
, entrepreneurs will critique this notion
52:05
of , well , all this cohort and all this is fine
52:08
, but really , you know , my product
52:10
was very shabby six months ago , right , or 12
52:12
months ago now It's getting much better . Yeah , but
52:14
you will see the trending of the entire curve going
52:17
up right . So when we are looking at it as well , we're
52:19
not just looking at an absolute data point saying
52:22
, oh , month three must be this and month eight must
52:24
be the or six must be This , right , we're
52:26
saying is the overall curve trending up right ? So is your
52:28
onboarding getting better ? is your acquisition
52:30
strategy getting better ? is your day seven
52:32
, day 30 ? Retention , or for a B2B
52:34
SaaS company , might be quarterly revenue .
52:37
Plan or whatever .
52:38
Yeah so the curve ? a beautiful point
52:40
that the , the directionality
52:42
of the curve , is what we're looking at , not
52:44
often not the absolute number For
52:47
any given month or quarter .
52:49
Absolutely .
52:50
Yeah .
52:51
Well , i think we kind of like covered
52:54
We covered cohorts
52:56
, we covered aggregates , we covered
52:58
medians and averages
53:00
and AOVs . So Which
53:03
is where we would like to wrap up This
53:06
episode ? but one thing I will leave our
53:09
listeners with , and especially entrepreneurs , is
53:11
that's very important to visualize
53:13
the data , because if you don't visualize
53:16
the data and by that I mean actually drawing the
53:18
trend lines , drawing the right kind of kind
53:20
of curves there So that you
53:22
can draw meaningful inferences in a quick
53:24
and efficient fashion It is
53:26
very difficult , because we can get all kinds of data
53:28
and it's just noise if you
53:30
cannot actually make meaningful inferences out
53:32
of it .
53:34
Yeah , and if I may add my closing comments before
53:36
we wrap this up , shri Pati , is that , in
53:38
addition to the visualizing , look , we talked about a lot
53:40
of different data , a lot of Bad math , about
53:42
different metrics , etc . I
53:45
think you also want to quickly zone in
53:47
on the , you know , two to three or maybe three
53:49
to five metrics that really matter , because
53:51
you you don't , as a you know , have the ability
53:54
, as a young team , to focus on 18 different
53:56
things . You should measure everything , you should
53:58
visualize everything . Yeah , you should
54:00
capture everything , because you may come back with an insight
54:02
later And oh , we never measured that . Now
54:04
I don't have ability to figure that out right
54:06
. But then really pick
54:08
the three to five that are defining to your business
54:11
. This kind of customer
54:13
, this kind of cohort , this kind of ACV
54:15
leads to this kind of retention , of this kind of churn
54:18
. Everything else is noise , okay
54:20
, and you're seeking that . You
54:23
know highly repeatable , you know
54:25
Exciting customer
54:27
, cohort or behavior that that you're going to
54:29
try to tap to . So I would say measure everything
54:31
, but focus on a handful of few
54:34
in terms of driving the business . Absolutely
54:36
All right with that . We'd like to call
54:38
it a wrap for this episode of
54:40
the prime venture partners podcast . If
54:42
you like the episode , please write into us . If you'd like
54:44
to hear more about stuff like this , or any other
54:47
metrics that are your peeves , please
54:49
do send it to us on our LinkedIn or
54:51
our Twitter handle . Thank you .
54:55
Dear listeners , thank you for
54:57
listening to this episode of the podcast . Subscribe
55:01
now on your favorite podcast app for
55:03
free and you'll be the first one to know when new episodes
55:06
are available . Just
55:08
search for prime venture partners podcast
55:10
in Apple podcast , spotify
55:12
, cast box or however You
55:14
get your podcasts , then hit , subscribe and if you have enjoyed
55:17
the show , we would be
55:19
really grateful if you leave us a review
55:21
on Apple podcast to read the full transcript . To read
55:23
the full transcript , find the link in the
55:25
show notes .
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More