Sheldon Poon on AI, PPC, and Why Bad Data Is Killing Your ROAS

In this episode of Inside Marketing with MarketSurge, Reed Hansen sits down with Sheldon Poon, CEO of Drive Marketing, to break down the real relationship between AI, data quality, and PPC performance.

While everyone is talking about AI in marketing, Sheldon reveals the truth most marketers miss: AI is only as good as the data you feed it.

They dive deep into:

• Why bad data destroys AI performance
• The real difference between predictive AI vs generative AI
• Why Meta and Google’s AI tools sometimes go rogue with your ad spend
• How clean tracking and smart bidding produced a 60x ROAS campaign

If you run ads, manage marketing teams, or rely on AI tools to drive performance, this episode will change how you think about data, automation, and accountability in marketing.

💡 What You’ll Learn

• Why AI fails with bad data – The hidden tracking issues sabotaging PPC performance
• Predictive vs Generative AI – Why most marketers misunderstand how AI actually works in ads
• Meta & Google automation pitfalls – When AI starts spending your budget on the wrong things
• Data-driven bidding strategies – How disciplined optimization can produce massive ROAS gains

🔥 Key Highlights from the Episode

✅ 60x ROAS Campaign Breakdown
How clean data, strong timing, and smart bidding created a record-breaking PPC performance

✅ Why GA4 Data Often Fails Marketers
The most common attribution and tracking mistakes agencies still make

✅ The AI Problem Nobody Talks About
How generative AI inside ad platforms can misrepresent brands and waste budget

✅ The Future of AI Advertising
Why predictive AI is still the most powerful tool in performance marketing today

👤 About Sheldon Poon

Sheldon Poon is the CEO of Drive Marketing, a performance marketing agency specializing in PPC optimization, advanced analytics, and AI-powered ad strategy.

With over 20 years in emerging technologies and AI, Sheldon helps in-house marketing teams transform messy analytics into clean data pipelines that power smarter advertising decisions.

🔗 Connect with Sheldon
 LinkedIn: https://www.linkedin.com/in/sheldon-poon/
Website: https://drivemarketing.ca/en/

🎙️ About the Podcast

Inside Marketing with MarketSurge is hosted by Reed Hansen, Chief Growth Officer at MarketSurge, and explores the boldest ideas, smartest marketing strategies, and real-world insights shaping modern business growth.

Reed is a seasoned digital marketing executive known for helping companies—from startups to Fortune 500 brands—scale using innovative digital strategies and data-driven marketing systems. 

Each episode features unfiltered conversations with marketers, founders, and growth leaders who are actually moving the needle.

📈 Want Marketing That Actually Performs?

At MarketSurge, we help businesses turn marketing chaos into predictable growth systems.

💻 Website
 https://marketsurge.io

🎯 Book a Strategy Call
 https://link.marketsurge.io/widget/bookings/marketsurge/demozvwcxw

🔗 Follow MarketSurge

📘 Facebook
 https://www.facebook.com/profile.php?id=100035121171654

💼 LinkedIn
 https://www.linkedin.com/company/marketsurgeio

🐦 X / Twitter
 https://x.com/marketsurging

📸 Instagram
 https://www.instagram.com/marketsurging/

🎙️ Podcast
 https://marketsurge.transistor.fm/

Creators and Guests

Host

Reed Hansen

Reed Hansen is a seasoned digital marketing executive with a proven track record of driving business growth through innovative strategies. As the Chief Growth Officer at MarketSurge, he focuses on leveraging AI-powered marketing tools to help businesses scale efficiently. Reed's expertise spans from leading startups to Fortune 500 companies, making him a recognized authority in the digital marketing space. His unique ability to combine data-driven insights with creative solutions has been instrumental in achieving remarkable sales growth for his clients. ​

What is Inside Marketing with MarketSurge?

Welcome to Inside Marketing with MarketSurge — your front-row seat to the boldest business insights, marketing breakthroughs, and entrepreneurial real talk.

Hosted by Reed Hansen, Chief Growth Officer at MarketSurge and a digital marketing veteran who's helped scale everything from scrappy startups to Fortune 500 giants, this podcast dives deep into what’s really moving the needle in today’s marketing world. Find us at Marketsurge.io

Each week, we’ll break down the latest marketing and business news (minus the fluff), explore tech trends you actually need to know, and feature unfiltered conversations with the most interesting minds in entrepreneurship and marketing.

Whether you're a founder, a marketer, or just a curious hustler looking to level up, this is where growth happens—loudly, smartly, and with just the right amount of sass.

Subscribe, tune in, and let’s scale something legendary. 🚀

Speaker: Welcome to Inside
Marketing With Market Surge.

Your front row seat to the
boldest ideas and smartest

strategies in the marketing game.

Your host is Reed Hansen, chief
Growth Officer at Market Surge.

Reed Hansen, MarketSurge: Hello
and welcome back to Inside

Marketing with Market Surge.

I'm Reid Hansen, the Chief Growth
Officer at Market Surge, and today's

guest is Sheldon Poon, CEO of Drive
Marketing and a performance PPC

specialist, helping in-house marketing
teams turn better data into smarter AI

decisions, higher roas and lower cpa.

Bridge is something most marketers
struggle with the gap between

flashy content and hard data.

He's deeply technical, obsessed with clean
tracking, and outspoken about where AI

helps and where it wrecks your ad spend.

to unpack why beta bad data destroys
AI performance, the real difference

between predictive and generative ai,
what most in-house teams get wrong

about PPC and what it takes to hit
60 x ROAS without it being a fluke.

Sheldon, welcome to the show.

Sheldon Poon: Yeah, thanks for having.

Reed Hansen, MarketSurge:
Yeah, no, my pleasure.

So, you know, from information we've
received about you and drive marketing

you talk a lot about how better data
leads to smarter AI and a higher roas.

Can you talk a little bit about that?

how that comes to be.

Sheldon Poon: Yeah, definitely.

So obviously the big headline these
days in marketing, right, like everybody

is talking about ai, whether it comes
to tech, when it comes to content,

when it comes to jobs like AI is kind.

Of like the big banner.

I have been working in AI and emerging
technologies for, well emerging

technologies for my entire career.

So like 20 years AI for
at least the last 10.

So, like, I thought I was getting
into AI late when I was working with

TensorFlow and writing Python scripts.

And I'm, I'm quite proud to
say that like we were actually

very early as it turns out.

So as new stuff has been coming out, we've
been able to, to, to stay on top of it.

Fast forward to today, and the
first question I always get whenever

people are asking us, like, Hey,
like, you know tell us what we're

doing wrong when it comes to PPC.

Tell us how we can improve our roas.

The first question is always like,
what are you guys doing with ai?

What is Drive marketing
doing in terms of ai?

And I.

There's always this, this misunderstanding
that AI is the end all be all, when

really it comes down to how you train
the tools and the, the, the training

of these tools and algorithms comes
down to the quality of your data.

So the big gap that we're always seeing
is that, hey, you have terrible data.

Of course, your, your, your
AI tool is not working.

You've trained it badly.

So it's performing badly.

'cause it, it's only as
smart and as accurate.

First thing,

auditing digital footprint.

In terms of analytics for two things
when it comes to the data, it has

to be, it has sounds super simple.

We start diving into the code and
we look at the entire footprint.

It's really complex to stitch that
together in such a way that the

ads platforms and the ais will
perform the way that you want.

And that, that's really
what we specialize in.

That's the reason why even though we're,
we're a marketing company really we're

a bunch of tech nerds that, that kind
of work help us as marketing on it.

'cause most of my team
is very, very technical.

I myself have a computer
science background.

We have a lot of devs on the team, so.

When we talk about bridging that gap,
right, between content and, and tech

and ai, really what we're doing is we're
trying to take the content and the digital

footprint that our, our clients have, like
all the stories, all the case studies,

all the, the flashy brand stuff that, that
is usually excellent and is usually very,

very much on point in terms of messaging.

And we make sure to.

Convert that into actionable, hard
objective data that's clean, that

can be fed into an ads algorithm
that can then take action and create

a feedback loop to be able to turn
that content into actual revenue.

So, so that's kind of
what we specialize in.

Reed Hansen, MarketSurge: So you
mentioned about, you know, we're,

we're talking a lot about the
importance of clean data, good data.

where are you seeing the Google Analytics
data GA four data really falling down?

Like, where is, are there common
places where you're like, ah, you know,

like, I knew it, I knew that's where,
you're getting bad data, you know.

Sheldon Poon: Yeah, absolutely.

So one of the things that we see
most often is that ad agencies.

Don't, for whatever reason, even in
2026, they don't tend to understand the

importance of the granularity of the data.

So a lot of the times, I, I always
joke that like a lot of my, my a lot

of our clients come because we're
usually friends that have like a big

red phone on their desk that like,
something's wrong, call Sheldon.

And usually the number one thing that we
see is that there's attribution problems

where both Facebook and Google are trying
to claim that they were the one that.

Brought the revenue or that Facebook
is double firing the revenue or that,

you know, Google was reporting that
we made like, you know, 2 million

for the client this, this month.

But really like when we talked to the
client, there's only 500 K in there.

So those types of disconnects are
usually a thing that, that, that we

see quite often because a lot of the
times these ads, agent ads agencies are.

Are not technical.

So they rely on a third
party web development agency.

And that disconnect makes it so that
the ad agency is saying like, oh, we,

we need this data, install this tag.

The web agency, their technical guys
that don't understand the importance of

the, the marketing and business side.

So like, well, I install the
tag, what do you want me to do?

The tag is firing.

The ad agency then gets into the
situation where, because they don't

have the technical skill to diagnose
why there's a problem, there's a,

a communications breakdown where
the technical guys are going.

Like, as far as we can
tell, we installed the tag.

You told us to install this
tag, we installed it, we don't

know what it's supposed to do.

The ad agency goes, the data's not right.

You guys need to fix something.

And then it just, it's a communications
breakdown and it's usually at

that point that people call.

Me and my team because we, again,
we bridge that gap and we understand

both sides, so we're able to dig
in and diagnose those problems.

And a lot of it just comes down to just
don't just install the default tag,

actually go through the, the process
of, of vetting your data properly.

Reed Hansen, MarketSurge: Yeah.

No, I, that, that makes a lot of sense.

Now I understand that you
hit 60 X roas for a client.

Tell us about that.

How's that even possible and why are
you still doing what you're doing and

not just like on an island somewhere?

So.

Sheldon Poon: So yeah, so this is a client
of ours in the automotive space, right?

So I think in the automotive
space the industry average ROAS

is something around eight x.

Reed Hansen, MarketSurge: Hmm.

Sheldon Poon: So When my guys were kind
of just doing the regular monitoring the,

the the dashboards came up and we started
calculating like 60 something X roas.

And right away my ads guys flagged
this, like, oh my God, something's

wrong, something's broken.

The de like, you know, it just became
like chaos at the office for a minute.

The devs were like, pulled
out of their projects.

Everybody had an emergency meeting.

The ads guys were
reviewing their dashboards.

The devs were told, like, review
everything on the website, find out

what happened, find out what's wrong.

Our accounts guys got on the phone with
the client to apologize and say like,

Hey, like, you know, something's broken.

Really sorry.

I got the short stick.

I had to call the CEO of the company.

And I was like, Hey, how's it going?

Can I ask you what your
current revenue is?

And he is like, why?

I am like, no, just we're
having a bit of a data issue.

just bear with me.

He gives me the number.

I check against our reports.

And I was like, oh my
God, this is 90% accurate.

Reed Hansen, MarketSurge: Wow.

Sheldon Poon: I told my
guys like, drop everything.

Everything's fine.

Like, stop trying to find this red flag.

It turns out that
everything was just firing.

All it was.

I, I, I think it was a
combination of things.

'cause I, I then called the CEO back and I
said, look, listen, like, don't freak out,

but we absolutely killed it this month.

Don't expect this every month.

Reed Hansen, MarketSurge:
Yeah, that's right.

Sheldon Poon: But basically what was
happening was the way that my guys

were bidding because even though
like a lot of it's AI driven, right?

We still have to adjust the bids
pretty much every two days just

to make sure that it's stay.

Like what we tend to do is, if
you imagine like a, a graph,

but we create like a channel.

And as long as it falls in between
that we're okay and every time it

dips below in terms of performance, we
adjust it to go back up with budget.

And anytime it starts going
over budget, we adjust back up.

So what ended up happening was there the.

One of the guys on the team Jack, I
think it was, saw an opportunity because

it coincided with the tariff scare

Reed Hansen, MarketSurge: Okay.

Sheldon Poon: and the tariff scare
made it so that people recognize that

these parts are about to get expensive.

So it just coincided with, in there was
a couple of regions where I think it was

just a high season mixed in with the fact
that they had unusually nice weather.

So people were just working on their cars
and we just hit them at the right spot.

So it just, by, just by making sure
that the, the data was, was solid,

mixed in with a little bit intuition.

Jack basically okayed a higher
spend for that period just

to see what will come of it.

And because he okayed the higher spend
based on the data that we had, it was

out bidding a lot of the competition.

So we were able to capture people
at the moment of intent outranked

the competition, and we just
captured more of that traffic.

So it was a very calculated risk
based on very, very good data,

and it ended up paying off and
that's how we, we will get it.

So that was a one-off for that month,
but, consistently, that account

gets around 32 return on ad spend.

And we're quite proud of that.

Like the, the account just fires
really well across the board.

Website's very well optimized.

Their customer service is excellent.

Their inventory is good there.

It is just like one of these
unicorn clients where everything

aligns across the board.

We work very closely with
that team on a regular basis.

They're, they're fun to work with and we
just get really good results out of it.

Reed Hansen, MarketSurge:
Well, that's awesome.

I mean, keep that client forever.

That's a, a great story.

Now you know, you pointed something
out about the bidding and you know,

you're, you're actively bidding on,
on, in some instances Google and Meta

are pushing automation like crazy
and, and in, in all sorts of things.

You know, like even in the image and text
generation, they are, just very active

and trying to enhance the offering.

What's your overall opinion about the
effectiveness of those ais Inside of

Sheldon Poon: Very bad.

Reed Hansen, MarketSurge: Okay.

Oh, interesting.

Is it, has it gotten any
better or is it improving or?

Sheldon Poon: So okay, so when, what,
this is one of the things that I

always have to kind of caveat whenever
I'm introducing people to the idea

of AI a little bit more deeply than
what, you know, the flashy headline.

So, most people when they think
of AI today, they think of LLMs.

So, Chachi, bt cloud Claude
and you know, these kind of AI.

Platforms where you log in and you
start having a chat with Gemini

and it generates something for you.

Reed Hansen, MarketSurge: Mm-hmm.

Sheldon Poon: That type of AI is.

Generative ai and it's captured
a lot of attention because

people it's tangible, right?

Like you can log into chachi pt.com

and start having a conversation
so people have like a better

understanding and a better grasp of it.

Most of what we do is based
on predictive ai, and when you

start getting to predictive ai,
it's not generating anything.

It's using existing data to
predict outcomes and then

action on those outcomes.

Reed Hansen, MarketSurge: Hmm.

Sheldon Poon: And that's.

Traditionally where the ad
platforms have been focused.

So when we're using Google, we use
something called P max which is

essentially predictive AI that.

It takes all of your assets
that you put into a library and

then it decides what is the best
audience for this particular asset.

And it will go across Google search,
Google Display, YouTube, and it'll

just decide what's gonna get the
best result based on your settings.

So that's largely the predictive AI
that we've always been very happy with.

Reed Hansen, MarketSurge: Hmm.

Sheldon Poon: What's happened recently is
that Facebook and Google, to your point,

have started rolling out generative ai.

Reed Hansen, MarketSurge: Hmm.

Sheldon Poon: On top of the
predictive ai, and that's our gripe.

It's done a horrible job with meta.

In particular.

They, they released this new algorithm
called Advantage Plus, which is

their version of Google's p max.

And it's supposed to do the same thing
where just like Google has multiple

platforms, meta has Facebook, they
have Instagram, they have WhatsApp.

And the idea is that Advantage
Plus is supposed to do something

similar, where at the profile level,
you feeded all of your assets.

And it will decide.

Okay.

Does this one goes in Facebook?

This one's doing well in reels.

This one's doing well in, in Instagram,
and it just kind of like decides

the, the struggle that we've had
with Meta is that the, for whatever

reason, Meta's started shoving the
generative stuff down our throats.

So it'll take our assets and then it
will decide to create new assets from it.

And then just push that out.

So there's a big problem here,
'cause one, it's using ad spend

that we didn't really approve of.

Reed Hansen, MarketSurge: Hmm.

Sheldon Poon: And two,
we've had situations where.

Some of our clients are in fashion, right?

So they're very conscious of their
brand and they freaked out because

they're like, who approved us?

And we had to tell them nobody.

So they were not happy when they saw their
products being misrepresented when they

saw their brand being misrepresented.

We ended up having to go we had an
emergency call a couple of weeks

ago where I literally had myself
and three of our, our executives.

In Facebook doing screen share, digging
through meta to try to figure out where

the generative AI was pulling these assets
from, why it was generating garbage,

why it was sending people to a blog
page on our e-commerce client's website.

We had to go literally four menu layers
deep to find a random checkbox that

said it's gonna automatically scan the
site and decide what the catalog is.

And it decided on a
blog article from 2021.

we had to get rid of that.

We found a bunch of assets
that generated that.

We just complete garbage.

It was pointing things to, so we,
we had to comb through the entire

system to manually uncheck anything
that had any piece of generative ai.

And it took us, it like I, we
had four execs on that call.

It took us like two hours and
then we had to repeat this

for all of our other clients.

And to make things worse, what would
happen is like so one of our member, one

of our team members on a screen share
would uncheck a box, hit save and meta,

and then we would go back on one of
our computers and check, and we would

see that the check rechecked itself.

Then when she refreshed, it had
reached, so there was a bunch of

problems with the meta platform where
it was just forcing the AI options

and we could not get rid of it.

So in the end, I think that week we
lost probably a good in total like 20

or 30 man hours just going back and
undoing the AI generative stuff that

it was trying to shove down our throat.

And this is.

At the, again, like the AI going
rogue, like that, it's costing us

money because it's using ad spend for
stuff that doesn't work, for stuff that

misrepresents the brand and gets us as
an agency in trouble with our client.

So I, I'm not a huge fan of
the generative AI stuff at all.

Reed Hansen, MarketSurge: Okay.

I mean, are we gonna have a
different conversation in a year?

Like, are you, know, what
do you think, like any

Sheldon Poon: I hope so.

We had other anecdotal stories
like this, like it's not any

better on the Google side.

Another fashion brand also called
and complained because they saw this

random video on YouTube and what
they had done was Google took a bunch

of the videos and the assets that
we had already created manually.

It spun up its own YouTube
channel that nobody had access

to, and it started running its
own generated videos through that.

Reed Hansen, MarketSurge: Oh.

Sheldon Poon: Yeah, it was really,
really frustrating because we had no

control and no eyes on this account.

When we dug into the data frustratingly
those ads that it generated

were outperforming the approved
assets the client had given us.

Reed Hansen, MarketSurge: Interesting.

Sheldon Poon: so the frustrating thing is
that while yes, we have no control over

it and it doesn't represent the brand
well, that to me is a problem that the

way the Gemini is going, I think they'll
eventually be able to, to fix that.

Reed Hansen, MarketSurge: Okay.

Sheldon Poon: And the real strength and
the takeaway from it is that despite

the fact that it was misrepresenting
the brand, which I think they will

fix, eventually, they were able
to generate an asset based on the

data that the AI was collecting in
a way that created an asset that

outperformed our human created assets.

So in a year from now, who knows, like
it might be the case that they find a

sweet spot where it's able to do the
work properly in a way that, and gives

us some sort of eyes and approval process
that mixes in this very, very smart

kind of content creation based on hard
data, based on the objectives that we

fed, the algorithm and the campaign.

So there's a lot of
interesting stuff happening.

Whatever's happening now is definitely
not gonna stay the way it is.

So in a year from now, who knows?

It might be a very different
conversation where I'm like, yeah.

We are fully, fully like on
board with the generative ai.

Reed Hansen, MarketSurge:
That's an interesting situation.

You know, like, you know, I don't
think many agencies or brand are

as quite as diligent with the
data as you are, and I think.

or, you know, and, and working in fashion,
you know, I think that's, that might be

one of the most brand and visual, you
know, like adhere to standards industries.

And, and, and, you know, interesting,
you know, you point out that AI.

Generated stuff.

It didn't match the brand
standards, but it performed well.

And you know, I think there's a lot
of companies that would say like,

well, my brand is making money.

my brand standard is like, I
don't care what it looks like.

But where, where do you
think that balance is?

Like ideally, like if you were a company
that hadn't been advertising and.

You know, and was to, wanted to start
advertising for some additional growth.

What would you recommend as far as
a balance between what's available

on AI and the human diligence that
is required to be really precise?

Sheldon Poon: So, so that's
kinda like the main question.

I think that that's the real
question I think people are asking

when they come to us and say
like, what are we doing with ai?

Right.

That, that's what they
want to get down to.

I, I, I think.

For us, the, the best use of AI has been
for us to look at our own processes very

clearly map out the standard operating
procedure for a given process so that

we can tear it apart and say, well,

This is grunt work.

This is grunt work, and hand
those pieces off to AI so

that we can be more efficient.

And then when we do let the AI kind of
run a little bit on its own, we just need

to make sure that we understand that.

The, it's a black box.

And on the other side of the, the, the
output, we just need to be diligent

about quality assurance and making
sure that we stay on top of it.

So anyone who's, who's looking at starting
ads like running ads today, I would say.

First and foremost, make sure
your foundation is solid.

Make sure that you know your
website is, is running properly.

If you haven't done an audit in
terms of your data, get somebody

like us to go in, audit your setup.

Make sure that GA four is
collecting stuff properly.

Make sure that all of your data
is clean and accurate according

to your actual revenue numbers.

Make sure that that foundation
is clear before you spend your

first dollar on ads because.

Again, if your foundation is not good,
your data is not good, you're just

paying meta and Google and you're
not getting anything out of it.

So first and foremost is just to
understand your own company and make

sure that your foundation is solid.

Okay.

Reed Hansen, MarketSurge: Awesome.

Sheldon I can tell you are one of
the real experts in this field,

and I am glad that you came on.

If people would like to work with
you, you know, the that really

want high performance, they want
high quality, you know, where would

they, find you in your agency?

Sheldon Poon: So the two easiest
ways to find us is if you go on

our website drive marketing.ca,

not.com,

dot ca.

And I always joke, it's like,
you know, the Canadian like ca.

So Drive marketing.ca

is our website.

All of our contact information is there.

We're about to post our case studies
for the year which have some in

info, interesting information
and interesting numbers on it.

So I'd say look out for that.

And then the other really easy way is
just to get in touch with me directly.

She, I think I'm the only Sheldon Poon
on LinkedIn, so I'm pretty easy to find.

Reed Hansen, MarketSurge: Great.

Sheldon Poon: So Sheldon Poon based
in Montreal, just shoot me a dm.

I'm pretty active on LinkedIn and I
check my messages pretty regularly.

Reed Hansen, MarketSurge: Well, fantastic.

It was so great to have you on.

thanks for making the contact.

And we are delighted
to to be connected now.

Sheldon Poon: Yeah, it's my pleasure.

Thank you so much for
having me on the show.

And yeah, I mean, like I, if you're open
to it, I'd love to do this again as like

new stuff comes up and we have like new
random stories I can tell you about.

Reed Hansen, MarketSurge: I'd love that.

Well, let's keep that in mind.

We'll check in, in a month or two.

Sheldon Poon: Yeah, sounds like a.

Reed Hansen, MarketSurge: I.

Speaker 2: Want to stay ahead of what's
actually working in marketing right now.

Head over to Market surge.io

and see how we're helping businesses
grow smarter, faster, and louder.

That's market surge.io

because your next breakthrough
shouldn't be a guess.