In this episode of Below the Fold, we welcome back Hicham Ghazal, one of the region’s top marketing strategists, for Part 2 of our deep dive—this time, into a topic performance marketers rarely talk about publicly: custom bidding. If you're relying solely on smart or manual bidding, you're leaving money on the table. Hicham breaks down what custom bidding really is, why it’s a game-changer for scaling efficiently, and how marketers can start building smarter, more profitable ad engines with it. With decades of experience across agency and brand sides, Hicham shares raw insights on what works, what doesn’t, and how the smartest marketers today are turning bidding into a business strategy, not just a media tactic.
• What custom bidding is—and why it matters now more than ever
• The limitations of smart and manual bidding
• How custom bidding improves efficiency and profitability
• The math behind custom bidding: inputs, outcomes, and what to track
• How to structure your bidding models for real-world impact
• Brand health vs performance marketing—and how to balance both
• Why marketers need to own their numbers and not rely blindly on algorithms
• Leadership, data, and the evolving role of marketing in 2025
✔ Custom bidding isn’t just for data scientists—every performance marketer should understand it
✔ You can scale smarter, faster, and more profitably with the right bidding strategy
✔ Relying on Google or Meta to optimize your campaigns? It’s time to take control
✔ Brand building and bidding are not mutually exclusive—they should support each other
If you’re serious about leveling up your paid media game, this episode is a blueprint.
Loved this episode? Subscribe, leave a review, and tell us your biggest insight!
⏱ Timestamps
00:00 – Intro
01:45 – What is custom bidding (and why it’s misunderstood)
04:20 – Where smart bidding falls short
08:12 – Inputs that matter: defining value in your bidding model
13:50 – Common mistakes in performance bidding strategies
18:24 – Real-life example: building custom bidding from scratch
23:08 – Why marketers need to think like product managers
27:32 – Balancing brand health with aggressive bidding goals
32:10 – How custom bidding increases profitability
36:40 – Final thoughts from Hicham on owning your strategy
[00:00:00] Welcome to the number one advertising and a marketing podcast in mena. Here we celebrate industry giants, but also shine a light on hidden talents that deserve to be heard. Together. We uncover insights and ideas that often go unnoticed, but drive real impact. Whether you are a seasoned pro or just starting out, this space is for you.
Tap that follow button. Your support keeps us up and running. Don't forget to share your feedback in the common section. Or join us on our social channels. Now, sit back, relax, and enjoy the show.
He shall Welcome back to below the fold. It's
really nice to, uh, to record the second, piece with you.
Uh, for those that are [00:01:00] just tuning in and have missed the first part, we recorded a, an episode on media quality with Hisham. Uh, so feel free to go and check that out. And now, uh, we're going to cover with Hisham customized algorithms for those.
that don't know Hisham and haven't listened or watched the first episode. Hisham, could you just give us a look, a very quick intro, please? Yeah, of course. So my name is Hisham. I've been in industry years. I've, mostly worked on the supplier side of the business.
Uh, and I've recently joined double verify.
It'll be around two years now.
So
I did the shift from publishing house, content creation, uh, you Digital ad sales, banner sales, and so on, and move towards
measurement, quality, and, algorithms.
You've seen the tech firsthand, basically.
I
have, I have, and there's so much to see. I mean,
I've been, I've been with DoubleVerify for [00:02:00] two years,
and I'm still
learning the ways.
internally.
So it's, it's, it's a crazy world. And today we're here to talk, or in this episode, we're here to talk about various, about a very special topic, very interesting topic,
Which
is customized
algorithm, customized algorithm. Let's start with
maybe defining what is an algorithm first, right? And With a simple explanation.
Absolutely. I mean,
the best way to explain it is
keep it as simple as think of an algorithm
as a set.
instructions
that takes account an
input
and gives you an output.
Okay. So
one sentence, that's,
that, this is algorithm.
Okay. Now,
if
you want to look into how it can help, think of algorithm as an assistant that helps you answer the who,
What, when,
where, and how
and it
it takes a decision on your
Okay. [00:03:00] For the
the sake of our discussion, algorithms are going to be about
digital ads.
Okay,
So what are these? Who, what, where, when,
how,
Who should I
serve the ad creative should I When should I serve that ad?
placement
should I
I bid on
and how can I make this more The how part
the optimization part.
So this
is where
It
takes into account the bid request, the bid
floor, how much to bid, so how much
that impression
It can
look into the timing of
bidding, it can look into
placements, websites, and So this is what an algorithm does Okay,
clear.
Now,
we then moved into something called, the
Within a few years, I would say that they've done it and they've deployed it, which is the BYOA.
So bring [00:04:00] your own algorithm. And that's something that Google and Meta What is BYOA?
It
basically
allows you
as a,
um,
client,
as a client, as a brand, as an advertiser. yeah. Yeah. To create,
to develop your own algorithm using your data. So you can decide on which ads to bid on
based on your unique business requirements.
So,
this is how it started. This is,
I would say, the, uh, the intro
to customized algorithms.
Is that option was given to brands.
and Like, go ahead,
start building.
Can you give us, give us a practical example of a B Y or A Yes. Yeah, absolutely. So
the
whole idea of bring your own algorithm is that it's fully customized based on your own data.
[00:05:00] So
what
type of input? So we talked about,
it's a technology, it's a,
These are instructions that take
an input
and
they give you an output. So what,
input Can you have as a customized algorithm?
You can look into your sales data. You can look into your historical ad performance on site interaction. So not just an impression or a click
or a
landing.
how did each user, um, engage on your website? Right. So you have many brands, for example, where they,
The more engagement, the more on site engagement they see from a client,
or from a potential
customer,
they
would say, OK, this is a hot lead, this is a
cold lead.
And based on that, they can start creating algorithms to bid more for that user.
Right, right. So, um,
it can
be a combination, right, off of metrics. For example, it [00:06:00] doesn't, it, it's not just limited to one single metric,
uh, like,
like the conventional, right, Um, optimizations or, or algorithms. Because,
um, it, then would allow a proper
full funnel approach, right? And this is, this is one of the things that,
uh, you, you wanted to, to talk about, which is why, why, you know, these platforms have introduced customized
I guess it would be again uh, find the middle ground between a landing page visit and a conversion,
right? And for each, for different brands, it means different metrics, different combinations of metrics.
And I guess. Uh, this proves very valuable, right? For brands to to capitalize on. It does. The idea off giving that flexibility to clients is because every client has his own set off [00:07:00] and what works for one
doesn't work for another.
The metric might be very important to client A, but it's not that
to client B.
Think of it this way.
You
have clients across multiple industries. So whether it's automotive, entertainment, travel,
and they're
they're all using the same algorithm. makes no sense, right? So
because they're all using off the shelf, the default algorithms that Google gives you,
that
Meta gives you, they work.
Yeah. No one's saying they don't work, but when you are given the flexibility to input your own data, to increase the weight on one metric. Over the other to instruct that the the algorithm like yes, a click is important
put that weight on it But what's more important is my on site engagement. So you factor this in when you're creating that algorithm That's something off the shelf provide you So this is [00:08:00] when you start
really taking, leveraging the power of a I and whatnot. Right. And even on site engagement can be a multitude of,
of metrics,
right? We're not talking about, a particular conversion or an on site or soft conversion. It can be what you define
is
an on site engagement. And
you may come up with, your free to come up with
200 different types of on site engagements, right?
Uh, and feed that
into uh, uh, custom algorithm, right? That's an input.
if you take a look at every client's website, they're built different.
there's
the interaction points are different. So there's maybe add to cart, explore more, learn more, uh, download the brochure. that There's so many touch points within that journey that is different from one client to another, from one.[00:09:00]
from one industry to another.
So why not capitalize on that data? ~Um, and you, you, ~you also wanted to talk about, um, what to keep in mind when looking at
third party
uh, solutions. So there are out there
third party providers of customized algorithms. Yes, there is. Uh, we're, we're one of them.
Okay.
Um, But the question here is
why go to a third party provider? Okay, Uh, if you, if you take as a business proposition, Google's telling your any platform that allows you to create your own algorithm, they're telling you, go ahead,
build your own. We're more than happy for you to deploy it on our platform.
It's great.
Why aren't the brands doing it? The short answer, it's hard. It's not easy to build your own algorithm.
The
long answer,
there are
multiple reasons
for why brands haven't
[00:10:00] jumped
on that wagon yet.
One,
lack of resources and expertise. Not all
brands have the
expertise or they have the right people to build
these algorithms.
Two, building
an
building an algorithm
is not a
one time very high cost of maintenance.
terms of capital to put in,
but
also to actually
keep updating the algorithm.
You
don't build it
and then you set it and forget it. It doesn't work like this with algorithms and you see it. You, on
social media every now and then they tell you we've updated our algorithm,
We've updated. So it's obvious that, okay,
building it isn't.
you don't get one team, like a task force, build me a, um,
build
me my own algorithm. I'll deploy it. And thank you so much. That's two.
Three,
the time it takes [00:11:00] to build an algorithm.
months, okay. Years. keep maintaining it. fast paced industries, you want to move fast.
FMCG, CPG, tourism,
These
are automotive. These are very fast paced industries. Innovation happens quickly.
Competition is
so you
need to move fast. How do you move fast? If you have a third party experts that are able to build that solution for you,
go for it.
So
to your
your question,
should you look for?
What should you look for when you're built, when you want to work with a third party
First
and foremost, make sure that whatever they are building is compliant with data, data laws.
So whether it's GDPR, General Data Protection Regulation,
or C. C. P. A. California [00:12:00] Consumer Privacy Act,
any type of data privacy
law,
make sure it's compliant.
because if it's not compliant, it's gonna get you in trouble.
Two,
make
sure that
It
integrates seamlessly with your,
uh, tech stack.
We all know you use a DMP.
You use a DSP.
web analytics. There's so much that there
are so many technologies that you use and going back to the definition
algorithm.
It's a set of instructions.
that takes
into account an input.
So your DSP data, that's an input, your web analytics,
that's an Third
party measurement, your quality measurement, This is, this is an input that you can factor in and to your algorithm.
So make sure that whoever's providing you with a customized algorithm can seamlessly integrate with all these data points.[00:13:00]
sure that this integration doesn't
greatly
disrupt your workflow. Okay.
Now,
I'm
saying this, but obviously there will be some disruption, okay? You're changing
a whole algorithm and what you use to bid on, uh, media.
So obviously there will be some disruption at the beginning. But then later on, it's gonna make things worse.
Much easier for you. Yeah, I, I, I think this is, these are, you know what, actually,
I
just, I just, uh, remembered one. didn't think of it until now. Customer service. You can have the best algorithm, but if you face a problem and no one's helping you fix it.
Yeah.
So
make sure that whoever you're dealing with has a proper proper customer support.
Exactly. Yeah. And this is like you said. After sales. Not before sales after, after sales. After sales. After sales. So like you rightfully said, this isn't a widely [00:14:00] adopted approach.
right? It does require time. It does require a certain level of expertise.
This is where third party solutions, you know, can be helpful.
Um, And
obviously, you know, you would need the right support from this, third party solutions in order to help you implement
the custom Um, before we move to the last point on customer
algorithm, could you just maybe elaborate a little bit more on what sort of disruptions you're referring to that you, you would come across when implementing a customer, uh, a
custom
algorithm? so one of the disruptions you'd face is mostly in the testing
Yeah. Okay.
You want to test, and which brings us to the last point. Very good. Yeah. It's uh,
okay.
You found a third party solution that checks all four. [00:15:00] Okay. It's working fine. Everything is working fine for you. Now you want to make sure that the actual algorithm will drive efficiency for your brand.
So
you need to test it.
Now,
what's the best way of testing it?
A B
So you run a business as usual line item, and then you run another line item with the customized algorithm.
That activity
is disruptive to the way you do work.
because you're used to setting a budget in one place, but then, now, you have to split the budget
You have to make sure that every change you do in line,
the
business as usual line
item, um,
has to
be reflected in the
custom
algo line item because you cannot change the targeting.
You cannot change the audience. You want to compare apples to apples? Absolutely. This at the beginning can be time consuming and we all know humans are resistant to change and To be [00:16:00] fair, the people working on these campaigns, they're so overworked. So you come in and tell them, Hey, do extra work.
Yeah. With a little bit of a, uh, maybe also
risk to
their eyes, right? Because they're walking in,
uh,
uncharted or on uncharted territories. Right. But with the right partner in place, Absolutely. You need the right guidance. Okay. It's, it's,
when we do those AB tests, it's not to say,
Oh, look, we're better.
It has nothing to do with that. Okay. If we do well, you're doing well. The agency is The whole ecosystem is now doing
Why not?
Why not?
So once you do those AB tests and you say, you know what, I think I'm going to adopt
the custom algorithm as my business question is,
is a very fair question,
Let's
say everything is working fine now. [00:17:00] be sure that in three months that algorithm that is being updated and that is being fed new input
doing better than off the shelf
algorithms, right?
It's the same connection off. Okay, I did my measurement, you know, and the part one of our discussion.
Yeah, it's working fine. Now, why should I keep doing it?
So with a custom algorithm, why should I keep using custom algorithm? And this is where we tell clients
test it.
Do an A B test. Don't do against benchmarks. Run head to head.
Okay, test it. Test for few campaigns.
Few months. See which one is doing better in terms of frequency, C-P-M-C-P-V. Look at the metrics that interest you as a media buyer
and
decide. It's very data driven. There's no opinion
on it. Okay. It's either doing better or it's not doing You deploy it,
Three months down the line do another AB test. Yeah. [00:18:00] Why not?
other
platforms, they update their algorithms. We know that.
.We update the customized algorithms. Now it's time after three months do the test again and let's see
and if
and and and then
we're more than happy to tell Them do it two three times a year. It's right as long as the results
the results speak for themselves.
Okay. That's the beauty of data.
That's the beauty of algorithms where
you're really just comparing a list of key metrics. Okay. It performed
a
dollar CPM. This one 0. 7.
Speaking of CPM, we've seen improvements in CPMs by up to 47 with customized algorithms.
It's almost cut in
So the
the question is why aren't more people Jumping on that wagon again. It's difficult. But as you said with the right partner,
it can it can be done.
Very good
Uh,
well, I hope that after this, you know, more, maybe more people
will, we'll test [00:19:00] it. We'll test it. And more importantly, we'll continuously measure right, in order To make, sure that they
continuously improve their custom algorithm.
I do want to ask,
um,
the listeners and viewers to share their input in the comment section.
Uh,
share any questions they have for Hisham. on custom algorithms. We'll, we'll ask Hisham to, to reply to those. Hisham, thanks a lot.
My pleasure. My pleasure. Thank you for having me. Yeah,
quite the discussion. I'm really happy with the data. Yeah. Yeah. Same here. Same here. Um, yeah, I mean, um, onto the next one.
Absolutely.