Always Be Testing

Sandeep Menon, co-founder and CEO of Auxia and a former Google Pay marketing leader, joins Tye DeGrange, host of Always Be Testing and CEO of RoundBarnLabs, to discuss how AI and agentic marketing are reshaping performance marketing. Sandeep explains why marketers should move from campaign-centric thinking to user-centric thinking, with a “segment of one” replacing broad segments and rigid rule-based journeys. They cover why marketing is harder for AI agents than other functions, what a “10x marketer” really looks like, and why AI adoption is more an organizational challenge than a technology one. The conversation also covers build vs. buy, the three pillars of strong experimentation (a global control group, guardrail metrics, and continuous measurement), and the “re-acquisition treadmill” that ad platforms can create. Sandeep closes with a call for a bias for action over waiting on perfect data, and a candid take on where agentic marketing goes from here.

What is Always Be Testing?

Always Be Testing explores the experiments, insights, and growth stories shaping the future of affiliate and partner marketing in B2B SaaS. Hosted by industry veterans, the show dives deep into real-world lessons from the people driving measurable impact at companies like Google, HubSpot, Ramp, Webflow, G2, and beyond.

Each episode uncovers what happens when today's most innovative marketers challenge assumptions, run smarter experiments, and build programs that scale revenue through meaningful partnerships. If you’re in affiliate, partnerships, or SaaS growth — this is your front-row seat to how the best do it (and what they’ve learned along the way

[MUSIC]

Hello, hello, hello.

Welcome to another episode of

the Always Be Testing podcast.

I'm your host, Tye DeGrange,

and I'm really excited

to talk to Sandeep Menon.

Sandeep, how you doing?

Hi, Tye. Thank you for having me here,

and great to have this conversation.

Absolutely. It's been a

pleasure to get to know you,

and I think everyone's in for a treat.

Absolutely today, based on our chats,

the converging worlds of AI and

performance marketing,

and paid marketing are massive,

and I think we'll get

into all of it today.

Excellent.

Absolutely. Sandeep Menon is the

co-founder and CEO of Auxia.

He's former Google

Payments Marketing Leader.

He's one of the leading voices in AI,

agentic AI, and first-party data.

He's powering a lot of very interesting

things with his product,

Auxia, with what they're doing in paid

media and paid marketing.

I think you're in for a treat today.

It's so good to talk to you.

We had a blast

conversation in prep for this.

Tell us a little bit about what's new in

your world, what's happening.

Thank you, Tye. I want to start off from

where we had dropped off the last time.

In terms of why we

started on this journey,

I do, first of all, want to say that it's

an exciting time to be a marketer,

it's an exciting time to be a builder.

There is so much

change happening around us.

That is one of the

foundational things that inspired me,

along with a group of

engineers and data scientists

that I used to work together at Auxia to

leave Google and to start Auxia.

So the core thing, just to give an intro

to your viewers on what Auxia does,

we are trying to build the context and

intelligence layer for marketing.

We do this out of the belief that if AI

and agents really want to leverage

the true power of what they can do and

turn today's marketers

into 10x marketers, you

need to, as a company,

have a really good

context and intelligence layer.

Marketing, as you know,

and this is something which you've been

doing over the years,

is very siloed.

There are various departments which have

their individual functions,

and it's very multiplayer.

So once you have that

context and intelligence layer,

it is so much more easier for agents to

then go and do

amazing things on top of it.

So that's what we are excited about,

and that's what we're building at Auxia.

That's amazing. I love that.

And it's obviously near and dear to my

heart as we talked about.

I've been in-house,

I've been a publisher,

I've been on the affiliate side,

I've been on the agency side,

I've been on the network side,

and also run a ton of experiments as an

agency and an in-house team.

I run a lot of paid media,

and we also see a lot of

paid media running through

the plumbing and the

ecosystem of partner marketing.

So it's very, very

relevant and very fascinating.

You've spent a ton of time at Google,

which is also super interesting,

leading the business for Google Pay and

leading the marketing there.

What did that experience teach you in

particular about the

personalization piece

that you're looking to achieve with Auxia?

The fundamental learning there is

quite foundational to why

we started on this journey.

So I believe that historically,

marketers have been taught to

think in terms of campaigns.

What I mean by that is that you could

talk to any marketing team,

they work towards a campaign,

and every action,

every activity within

marketing is very campaign-centric.

One of the very interesting things that I

learned while at Google is that,

as a marketer, you are able

to achieve so much more if you

shift from what is essentially a

campaign-centric view to

much more of a consumer

or a user-centric view,

wherein everything that,

every experience that you show to a

particular end customer or prospect of

yours is rooted in the deep

data that you have about them,

the understanding that you have them,

so you're able to deliver a much more

hyper-personalized experience.

The problem with campaign

segments is that no individual,

no consumer, no business ever said,

"Hey, I would like to be in a segment."

The ideal segment is a segment of one.

So I think that was a

big aha moment for me,

the power of using AI

and machine learning to

move from that campaign-centric approach

to a user-centric approach.

The knock-on effects it

has on the user experience,

but also on the efficiency of what you do

in terms of your spend,

in terms of the results

you're able to deliver,

and in terms of, very importantly,

the speed at which you can do stuff.

Maybe let's drill into the how a little

bit for the marketers out there.

Like when you're getting that context,

when you're ingesting data,

maybe high-level building

blocks of what you're setting up,

and what does that look like to the

marketer and to the end-user?

Because it sounds

fantastic on its surface.

But I think for the

performance marketers,

maybe has had hands on the wheel a lot,

maybe letting go of some things via AI,

obviously that's

accelerating very quickly.

How do you explain it to a marketer who

maybe is reluctant or

skeptical on some claims?

Let's take the,

and I want to first talk about what I

would call hyper-positization using AI.

And then we can layer on the concept of

how you can use agents to connect the

work you're doing later.

So let's take something which most

marketers spend a lot of time on.

You spend millions of dollars, hundreds

of thousands of dollars,

or millions of dollars depending upon the

size of the company you are,

in acquiring prospects.

So let's say that you have

a prospect on your product.

And it can be any product, any company.

It could be a

retailer, it could be a bank,

it could be a telecom company,

but you've spent a lot of money in

getting a prospect to

come on your property,

or even sign up for your product.

What you do with them in the first, let's

say, 5, 15, 30, 45

days is very important.

That is the classic onboarding journey,

and that is something which marketers

typically spend a lot of time on.

Because every research out there, every

experience out there shows that

if you are able to engage with your

customers really well

in that early period,

if you are able to take them through what

is known as your magic path,

then they are happy,

satisfied, much more sticky, etc.

So what is a traditional approach that

marketers have been using for this?

What they do is that they kind of try to

segment their prospect

base into big chunks, right?

It could be based on some audience data,

it could be based on

the acquisition channel,

it could be based on a variety of

broad-based criteria.

And then even the best

themes, what they used to do is that

based on these segments have a very

deterministic rule-based journey,

which is like, "Hey, Tye

came in on my product,

and he has signed up, he's a user. I'm

going to send him an email,

it's going to be a welcome email."

In respect to what he does,

he's going to get a

second welcome email on day 3,

and then if he comes back on the app,

maybe I'll show him this notification.

All of these are very deterministic

rules-based journeys, right?

And they are based on

some past experiment,

but it is not fully

personalized to what Tye, what you want,

or what feature you are

most likely to relate to.

Contrast that to how a really good

AI-based decision system

would work for the

same onboarding journey.

Tye would come in, we would know all of

the data and details about Tye,

because you've signed

up the product, right?

And then every interaction that you have

across all of the channels,

whether it is on the app, on the website,

even walking into a store,

you should be able to understand and

deliver an experience to you,

which is truly hyper-postilized.

One of the ways I talk to

CMOs about this is that,

imagine that for every prospect or

customer that you have,

you could pair a really good marketer and

an amazing data scientist,

not at a segment level, but

for each and every individual.

What kind of experiences

would you then be able to provide

to that particular prospect or customer?

That is what you should be

able to do with this approach.

And it's a very big contrast to the

broad-based, rules-based implementation.

That's very fascinating.

Yeah, and so are you able to deliver...

Where is that showing up to the user?

Because I love the flow of

information and the changes

on the marketer level, but on the user

side, how do you think of it that way?

How personal does it get for us?

And so this is...

Without being too scary.

Exactly.

So the point about

personalization is interesting.

We often say that companies...

Look, there are a couple of

changes that are happening.

Number one, I would say,

is that most companies today

sit on a pressure trove

of first-party information.

Right?

This is information that you know about

your consumer, your user,

because they have voluntarily

given that information to you

when they signed up and that every

interaction they're

having with the product.

So the first thing that you need to do is

to make sure that you are using

that first-party data in an effective

fashion to deliver ideal experiences.

Right?

So it's a classic thing, right?

Like when you call a business that you

have had a deep relationship with,

your expectation is that you don't need

to repeat everything

that you have told them about you.

Like, of course, you can do some security

checks, but after that,

they should know you because you have

been doing business with them

or interacting with them.

It's like going to your favorite

restaurant and they

know the cocktail you like,

the table you like to

sit in, your appetizers,

and then maybe also

suggest some new things.

Right?

So you have the data.

So the first thing is using the data

really, really well.

When it comes to experiences,

and that is where a lot of the advances

in AI are going to help you

because marketers have invested a lot of

time in creating beautiful content.

And I do believe that that's an area

where marketers will always play

a very important role

because ultimately, as AI scales,

the one thing that businesses need to and

marketers need to

kind of hold really dear

is your brand because that's what

distinguishes you from another company.

But once you have your brand

and once you have that data,

you're able and you already

have all of these creatives,

then it's all about making sure that what

is the right communication

at the right time to that right user.

And for that, you need

systems that are able to take

hundreds of millions of

decisions in real time

and serve to you that right experience.

And to be honest, if you think about some

of the most well-known

digital consumer brands,

whether it's a Netflix or a YouTube,

they're able to

deliver on that experience.

They have an understanding.

When I watch your podcast, they know,

OK, he's interested in experimentation,

he's interested in testing,

he's interested in marketing.

So they would tend to kind of surface

content like that much more often,

which is personalized to me as opposed to

a topic that I do not

or I do not want to be told about.

Yeah.

And when you talk about agentic

marketing, like how is that done?

How are you able to pull that off?

And what does that look

like within the system?

Yes.

So now that you have that context, right,

like we are able to unlock a few things.

So let me take a step back and talk about

some of the challenges

which make marketing as a function

particularly difficult,

I think, for agents to act out.

So if you look around, if you look around

most white-collar functions,

you see that agents and agentic

approaches have fundamentally

changed the way you do that.

So every engineer working at Auxia now

has an agent for coding, right?

And that has been able to massively

increase their output, right?

And so whether it's cloud code or cursor,

it has fundamentally changed how good

engineers work today.

So the same could be said for legal,

the same could be said for some aspects

of financial services.

Marketing has proven to

be slightly more difficult,

and we believe it's due

to three reasons, right?

The first one is the

topic that I mentioned.

Marketing is very tribal.

And what I mean by that is that what is

marketing at Google, even the same term,

you can have a growth marketer at Google

is probably very different

from a growth market at a Comcast, right?

There might be some parallels, but each

company has its own set of practices.

So you need to first

understand that context.

And you can't just

take, it's not like legal,

where the NNDAs and NDAs and NDAs will

allow you to extend.

Yes, there are some aspects of it that

you can personalize for

that particular company,

but it's not that

transferable across companies.

So building context for a company is very

important, but it's all significant.

The second thing is that

marketing is very multiplayer.

What I mean by that is that it's a

function where anybody

who's involved in marketing

and at this to it, everybody

has a point of view, right?

So your legal team has a point of view,

your CEO has a point of view,

and which also means that every action

that you want to kind of do

needs a lot of coordination.

So any agentic system should enable that.

And the third point is, paradoxically,

what has happened over the past decade

is there has been a

proliferation of point tools.

Like what I mean by that is

that for each marketing function,

you have a specific tool, and very often

these tools don't talk to each other,

which further complicates

the process of putting out

a certain marketing activity out there.

So what a good agentic system is able to

do is to transcend all

of these limitations.

If you think about the day-to-day

activities of every marketer,

whether it's a performance marketer,

whether it's a life-size marketer,

or even a brand marketer, you would

contend that a lot of that time

is being spent on block-and-tackle stuff,

not on the strategy stuff.

It's about like, "Hey, getting this

creative through this approval process,

being getting feedback,

incorporating that," right?

So creating all of these manual segments,

setting up rules in a system of choice,

and then executing in a digital.

These are all work that can

be done very well by agents,

which have that context, right?

So in a world, the world

that we are moving towards,

in the world of the 10x marketer is one,

where marketers shift their activity

to more the strategic

aspects of marketing,

coming on a good marketing strategy,

making sure that it

adheres to your brand,

your guidelines, right?

And then having a set of agents do what

they're really good at,

which is doing repetitive

tasks once they have their context,

much more faster than a

particular individual can do,

and also enable that

coordination between teams.

That's interesting.

The team player aspect, I'm starting to

see more and more of this.

I think the silo removal has

been a trend we've been seeing,

my goodness, ever since the growth

marketer craze of 2015,

and the reforge-like growth

practitioners that were very,

I was a part of, and a lot of that

community did some amazing work

and experimentation and

getting out of the silos.

How do you, I guess I'm

hearing there's benefit to that,

which is really exciting.

Do you see things where marketers or

teammates or companies,

like sort of, how hard is it for them to

get out of their own way,

if that makes sense, to enable Auxia to

really thrive and do what it does?

Because I'm just curious.

You raise a very,

very, very important point,

and there is an analogy that we often use

to kind of explain this.

So when electricity first

happened on factory floats,

the interesting thing

that people notice is that

productivity gains from

electric motors and electricity

did not appear on factory floats for the

first 10 to 12 years,

or even the first 15 years.

It was very confounding, and the reason

for that was very simple.

What people did was they would just

replace a steam engine with electricity.

So they wouldn't change the factory floor

or how the factory operated.

It was only when they started rebuilding

factories around electric power

that you saw the big

productivity changes.

I do feel that we are in the similar

stage with a lot of AI and agent AI.

The point that I'm

trying to make here is that

it is not often a technology challenge.

Like the challenge is

often a people challenge,

it's an organizational challenge.

This is the topic that I spent a lot of

my time talking to CMOs about,

is how do you kind of

almost transition the structure

that you have

currently to an agent economy?

And that would require

some wholesale changes,

and I do think the smarter CMOs are

already making those changes right now

in their teams and

how those teams operate.

As somebody said, the

worry is not about whether...

This is what I tell marketers who are

early in their career.

Your worry should not be

whether AI will replace you.

Your worry should be whether there would

be another individual

who's adapted using AI

tools better than you are,

who will come into place

what you're doing right now.

Right, so I think

that's an important point.

I love that.

As a marketer, I'm

obviously thinking in channels.

The irony is our world

that I operate in is...

I almost view it as a

channel within channel,

multi-channel lever and a

channel within channels.

Hopefully, it's not too micro,

but how does this show up for Google

differently than Meta,

differently than LinkedIn,

differently than influencer?

How are you seeing it impact channels?

Because that's my legacy

brain going to those things

and just thinking

about where things show up.

I'd love to hear your perspective.

Yeah, it's a great point because the

concept that we talked about

in terms of context applies really well

to channels as well.

Typically, what I'd seen

working in large companies

and even our partners is

that if you look at your area,

there are experts.

There's a LinkedIn expert,

there's a Google Ads expert,

and they have a lot of context.

But what very often happens is that in

the unfortunate event

when that growth marketer who is so good

and understands the context

about that particular area

leaves your company,

then you have a big hole.

I think one of the things

that having a good context layer

will enable you to do is

to have the understanding

of how each of these channels work

and how we can react to it.

Specifically, with

respect to some of the channels

that you mentioned, like LinkedIn,

like I've seen some really good

innovative companies,

the way they use it is like,

again, it is about the speed

of response to the name of this podcast.

You always need to be testing.

Very often, it is

humanly not possible to react

to every change.

Whereas if you set up

a good agent routine,

they can monitor 24

by 7 what's happening,

let's say, on your LinkedIn feed

and have the ability to

either provide you the insights

on how best to

respond to it or over time,

learn how best to respond to that.

Those are things which

unless you had a team of hundreds

or hundreds of marketers,

you are not able to do it.

But today, a small and

a medium-sized business

does have that ability.

Again, it goes back to

the foundational aspect

about how whenever there's a

technology shift of marketing,

the way marketing has done changes.

When I started my career at Google,

the power was the internet.

And internet provided every business

with a unique

opportunity to get global reach,

which was till then,

unless you are maybe a fortune

hundred company, you

didn't think about global reach.

So because it required a

lot of money and effort

if you had a business

to get a global reach.

Come internet and--

Do you feel like

you're enabling companies

that maybe don't have as many resources

to sort of punch above

their weight to some extent?

So from an Auxia perspective,

we typically focus on large enterprises.

So it's not that they

did not have resources,

but one common theme that

you will hear from a CMO

that you speak to is that

they have all been asked

to do more with less.

And I don't mean necessarily less people,

but with less budget,

less resources overall

in terms of what you can achieve.

And I think that is a productivity boost

that we are able to unlock.

That's awesome.

We talked about you

crossing over 200 billion

autonomous decisions a day,

which is absolutely insane.

What do you think, what's like examples

of how that decisioning

works or one of those decisions

that you can kind of

share with the audience

that they kind of can

wrap their head around that?

Yes.

So we did talk about

the onboarding journey.

So I'm going to use that as that example

to kind of bring to life a decision.

Let's say, Tye, you've signed

up to a new product, right?

So and on the second day,

you go to that app, right?

Like what experience

that you have there, right?

Like into what marketing

experience that they have there,

what feature does the

product promote you to try?

That is a real time decision

we take and show on that app,

surface of the app, right?

So the email that you

get, what content is there

on that email and what

again, like what, for example,

if it's a product with a trial period,

what are the three or four

things that you need to try

within that period,

determining what those are

and like constructing that

email and sending that to you

at the right time is a decision.

That's super interesting.

That's super interesting.

That's super helpful.

Yeah, the scale is

exciting and what you're doing.

And you talked a little

bit about the 10X marketer

previously, obviously

that's a compelling one for me

and the audience being marketers,

but how do you sort of

define that for folks?

So this is who we are

building for, right?

Like we are building for the 10X market.

And what I mean by that,

again, using the earlier analogy,

like if you look at software engineering,

what Cloud Code has

enabled engineers to become

is a 10X engineer.

They're able to not just write,

it's not just about volume of output.

It is not just about the, hey, I'm

writing 10 times more code

to use a parallel in marketing.

Like creating 10 times more content

is not necessarily going to kind of,

it might be impressive,

but it's not going to

solve all of your problems

because the quality

of output also matters.

So the way we describe a 10X marketer

is that it should enable

any marketer in your team

to be 10 times more speed.

So whatever action

that you're able to do,

can you do it 10 times more faster,

10 times more quality of output,

and then 10 times more precision,

but you need all three of this.

You need the quality, you need the speed,

and you need the precision.

And if you're able to deliver on that,

then you are moving

towards being a 10X marketer.

And the only way you can do that is,

because we can't stretch time,

is when you have people helping you.

And the people helping you

would be your suite of agents.

And the suite of agents

would understand your needs

as an individual,

but also understand the company's need.

Again, that's a very important design

for building good agents, right?

Like good agents need

to have an understanding

of your context, but

also the company's context.

Because it's not enough to

make just you productive.

We need to make every marketer productive

with relation to the rest of the company.

I love it.

Yeah, it reminds me of Elena Verna

talking about the super IC

or individual contributor.

So I think more is being

expected of the individual

with all of these tools and genteck here

and coming and expanding

and obviously

intelligence getting so expansive.

When you think about all the explosion,

obviously Claude and other tooling

that have really exploded

over the last year in particular.

When you think about alternatives,

like buy versus build

and how you're positioned in this market,

what are some of the questions you get

and how do you kind of

overcome those ideas of,

oh, well, we have X or we have Y

versus really testing and

learning and building with Auxia.

How do you think about

the trade-offs for people

that are thinking they

can do this elsewhere?

Yeah, what you can do with a Claude,

any of the foundation

models is quite amazing.

You can achieve a lot.

So typically, we try to think about it

as a two by two matrix.

Let's say on one axis,

you have the

sophistication of your business.

You could be like a

single owner small business

all the way to a Fortune 500 company.

So that's the

sophistication of business on one angle.

And then on the other,

you have what I call simple workflows

and complex workflows.

So my belief is that anything

which falls in any of the three quadrants

of a simple workflow,

irrespective of the size of the business,

or a complex workflow

for a moment pop shop

can be done and you

can achieve a lot of it

by using just a foundation model.

But when you go to

that top right quadrant,

which is complex workflows

in Fortune 500 companies,

that's where things get interesting.

That's where you do want to have a system

that is enterprise ready,

has the ability to learn

and accumulate your context

and is multiplayer.

Because those are the

things that you need

for complex workflows

and for large enterprises.

And that's where companies are playing.

And honestly,

I did my taxes last year with Claude,

but I'm not going to

do Auxia's taxes with Claude

because I think we are a company

and it is empowering in a lot of sense,

but that is where I see

the industry move towards.

And if anything,

more marketing organizations are,

like workflows are fairly complex.

Yeah, I was talking to gentlemen

helping infrastructure and their back end

positioning for tooling

and AI last week recently.

And they're at a growing very large SaaS,

enterprise SaaS company,

maybe upper mid-market,

they're pretty big.

And he was saying how he

has to kind of break down

and understand and organize

how the vibe coding is happening

and spreading throughout the organization

to attribute who's building what.

And when someone builds something,

like how do you ensure that

it's not just totally redundant

and already been built?

And so it reminds me of the complexity

that you're referring to.

Yeah, and then there are

also the additional aspects,

like you got it to be

secure, compliant with this thing.

And it cannot be that

to the point you said,

each small division trying out something

which doesn't tie with everything else.

It's a self.

Yeah, and congrats on the exposure

on the coverage on AppExchanger.

I thought that was really interesting.

You're sharing some of

the insights of that.

Obviously the need for context,

not just more data.

What's the hurdle that

you see most frequently

and just kind of stitching that together,

working across teams,

working across channels.

And with that stack,

is there something that

sort of triggers you to go,

okay, yeah, they clearly

need Auxia at this point?

To the point that I made earlier,

we add a lot of value if

you are a large organization

with multiple teams and processes.

And typically even

from a data perspective,

your data is not in a

very simple, clean format,

easily accessible.

Like most large

enterprises can wish for that.

Even at Google, I used

to say that people outside

think we have pristine data.

That is not the case.

So typically where you have an enterprise

with disparate data

spread over multiple systems,

you have multiple high value actions

that a prospect of a customer can take.

And you have multiple

touch points for those,

ideally digital touch

points for those prospects

your customers to interact.

That's where a solution like

ours can add a lot of value.

That's awesome.

Yeah, it makes sense.

Obviously the show

being always be testing

as you referenced,

your platform is all about

continuous experimentation as well.

Is there an experiment that

you see that has been notable

or learning or

something that has truly kind of

changed things for you or

changed things for your clients?

The one thing that working

in Decade Plus and Growth

has taught me is that if there is one

unique characteristic

that determines whether a company has a

world-class growth team

is the number of experiments they can do.

Right? Because, and this

is, I'm not saying this

because I'm on your podcast, but

ultimately that is the true

determinant of how good

a growth team you have.

Because the more you

test, the more you learn.

Because as any person who

has done a lot of these growth

experiments know that

like there is a lot of value

in your experience failing as well.

Right? Like in good growth

teams, you have a large amount

of experiments failing because that is,

that means you're

doing enough experiments.

There are a couple of

things which are foundational.

And when we designed Auxia,

we made sure that we build it

into the product from the ground up.

The first and the most

important thing is to have

proper control group to

understand whether you are

the results of your test.

You would be surprised how in, even in

some very large companies,

how "unsophisticated" it is the way

experiments are run.

So from a platform

perspective, we enable it.

You don't need to

worry about the complexity.

You always have a global control group.

You need to have very rigorous

counterfactual markings

to make sure that control versus tested.

And you can actually understand why a

particular uplift happened.

Right? So I think that

is very foundational.

And we bring it as part of our platform.

The second thing which

people often forget about is

the importance of not just following the

metric you're testing.

Right? And this is one thing which I

learned a lot at Google.

Right? Like you might be

optimizing for metric A,

but it is very important that you

understand the impact of

that particular test on

all of the other metrics.

Because very often, if a good example,

we work with a very

large newspaper publisher,

and we help them and we help

them do a lot of experiments.

And one of the things

that editorial team is very,

very, very particular

about is that, okay,

I'm okay with you running experiments to

understand how people

subscribe more, etc.

But we are ultimately a newspaper and we

stand by our editorial, this thing.

So there should not be any negative

effects on page views, etc.

So those are important guardrails and a

good system should be

able to run experiments

while also monitor how we are doing

within these guardrails.

And the third and final point I would

make is, again, to the

point about always be testing.

It's important to have that global

control group on a continuous basis.

You cannot stop it because very often

what people do is that

they run an experiment,

they're like, "Oh, uplift is there.

Uplift is done. Now let's move on."

The beauty of having a global control

group is that you will see how these

values drift over time.

Even the best growth experiments, the

results that they derive

can plateau or even decrease

after time, then you need to come up with

new things, which is why

it is important to have

that global control group always

measuring uplift techniques.

That's awesome. I love that framework,

those three pillars.

VentureBeat mentioned you guys are kind

of tackling this

re-acquisition treadmill.

What is it and what do you think about it

in terms of what matters to marketers?

This might be slightly, I wouldn't call

it controversial, but at

least one pet peeve that

I have is that I spent all of these years

at Google, over a decade

at Google, and I feel that

in many ways that was the true big

innovation in marketing,

like all the last decade,

programmatic ads. I think it's beautiful.

You can assign a goal and Google and Meta

can help you bring traffic from anywhere

in the world with the

right parameters and budgets.

But what it has paradoxically done is it

has made many marketers "lazy."

What I mean by that is they don't spend

enough time and effort in

making sure that once you

have a prospective customer, you engage

with them and you can

stay with them and you can

continue to kind of derive value from

them by giving good experiences.

I spoke recently to the CMO of a single

brand fashion retailer and

she was talking about how

Instagram is a very effective channel to

bring the first

customer in, but unfortunately,

they need to rely on Instagram to also

get that same customer to

make the second purchase.

That, in my view, is a re-acquisition. If

you have a really good

deep understanding of the

customer and if you're doing really good

marketing, you should

be able to cultivate

a relationship and get them to stay with

you and deliver value both ways.

That makes a lot of sense. Yeah, and it

kind of dovetails into the measurement

piece, right? You have the agent

optimization. You've got

things happening really quickly.

How does that think about or how does

that inform the

conversations and results you see with

marketers when it comes to measurement?

How much does it really change

measurement? How much does

it not? Just so that I understand the

question correctly, are you

referring to how the use of

agents and can influence measurement?

Yeah, so that's basically

the foundation of my question,

absolutely. So again, a lot of what we do

here comes from some

shared pain over the years.

So I'm pretty sure you also had this

experience. You run a beautiful

experiment and then you need

to spend a couple of weeks having a data

scientist or marketing

analyst review all of that data to

really understand what happened and what

insights you have from there,

right? That is a good example

of what can change with a really good

agent. You can have a data science agent

or an analyst agent,

which is able to kind of do that analysis

much more quickly than a

human can do. And what's more

important is that that agent can be

working 24 by 7. So it can constantly

kind of share new insights

and give you ideas around what to kind of

change. That is one of the

things that we do. We have in

our product, we have an analyst agent,

and we also have a feature called Sparks

because initially when

we launched the analyst agent, our

expectation was that, hey, a marketer

should go in and ask

it the right questions so that it can

give smart answers. What we

realized is that, no, no, no,

man, that is too much cognitive overload.

Why do you need somebody

to go and ask it the right

questions? We know what, like, you know,

man, within marketing,

there are a set of questions

that you need to ask and we can

automatically ask those questions and it

can surface those insights

and that's what we do. So every day when

a marketer logs into Auxia, they see a set

of sparks. These are

specific queries that the analyst agent

executed on the marketer's

behalf. Around the experience

it's been running and surfaces insights

on certain actions it

needs to take. It could be

things which are purely informational,

saying that, hey, performance of this

campaign is trending in

this direction. Or it can be

action-oriented, saying that,

hey, there are some treatments

for creatives which are performing much

worse than expected. There

is an opportunity to do a

refresh there. Love that. Very

interesting. What's something a lot of

enterprises get wrong about AI

personalization? It goes back to the

point that I made about data. When I

speak to a lot of senior

marketers, there is a belief that you

need to spend years in

"cleaning up your data" to get it

to a pristine state before you can use a

good personalization

system. I'm not saying that

completely corrupt or bad data will

provide magic, but what many enterprises

don't realize is that

they already have good enough data to

start the process. Very

often I run into marketing teams

that are on multi-year journeys on data

standardization and coming

up with data architecture.

That's good. It's important. But in

today's world, you can't

spend two years in data

standardization before you start using

AI. That's too much of an

at-least time. I do believe that

a lot of the models today have the

ability to understand

structured and unstructured data,

try to make sense out of it. Generally,

having a bias for action

is something which I would

strongly recommend. Yeah, I love that.

It's one of the many Amazon

principles that stuck with me.

It gets used a lot, but it's very much

needed. Letting

perfection and over-analysis

is just a great con grip. Organizations,

we've seen it a lot. I

love that. The whole agentic

thing is crazy. Agenic marketing, where

do you think it goes

in two to three years?

Do we end up with marketing teams running

agents or agents running

marketing? Where is it going to go?

This is a very dangerous question. The

pace of change is unlike

I've ever seen before. You and

I have been very fortunate to be being

part of some of the

earlier waves of technology, but

the pace of change is just amazing. I

think one of the most

foolhardy things you could do is to

venture, I guess, on the exact state in a

few years' time. What

I can say and what I

genuinely believe is that we are still at

the beginning of this

shift and in what it can do.

If you think about the last wave, was the

wave to cloud. Even now,

I don't remember the exact

figure, but one of the large consulting

companies had done a survey on what

percentage of global

enterprise data is on the cloud. I think

it's still hovering around

the 30% mark after a decade

plus a movement on cloud. I think the

shift is happening much more

faster, but we are very much

in the early stages. I know that's like a

pop-out answer to the question.

It's okay. It was pretty spicy. Sandeep,

you've been awesome. It's

awesome to get to know you,

and you spent a ton of time. You've

covered a lot. I'm really excited about

Auxia and what you're

building and all the growth and speed of

learnings, which is near and

dear to my heart, obviously.

Come on anytime and more exciting

conversations to come.

Thank you. Thank you, Tye. Thank you for

the opportunity. It's

always fascinating to

talk to a fellow traveler in the space

who has seen the previous

generation and is excited about

what's to come and has fundamentally the

same belief about

experimentation being the

bigger out that you need

to continue to focus on.

Yeah, you've doubled down on how

important it is and you're

building something that's

emphasizing that. For folks that want to

reach out and learn more,

what's the best place for

that and what do you suggest?

You can just drop me a note. I am quite

active on LinkedIn as

well. My email is Sandeep at

Auxia.io. I can drop you a note and would

love to kind of

continue the conversation.

Beautiful. Thank you so

much. Thanks, everybody.

Thank you, Tye. Have a nice day. It was a

pleasure talking to you.

Bye. Always.