No cookies, no problem: the new key to leverage machine learning in advertising

No cookies, no problem: the new key to leverage machine learning in advertising

October 01, 2021

Third-party cookies have long provided a way for businesses to identify and target leads for advertising. But with privacy concerns on the rise, it’s now up to businesses to leverage first-party data to get results.

Watch our full-length video with B2B advertising experts Ian Maier and Zoran Arsovski, who share why first-party data is the key to the new world of Facebook and Google advertising.


Or read on to learn about the new shift that's enabling B2B advertisers to not only continue to reach audiences without cookies but also reach them more effectively.

What are third-party cookies?

First, a quick refresher.

Cookies are small data files that websites send to your computer or device to store information about you. While that may sound creepy, cookies can improve your browsing experience, like how a website generates first-party cookies so that you don't have to log in every time you visit that site or your shopping cart won't empty every time you leave the page.

Third-party cookies are created by domains outside the one you're directly visiting. Ad tech companies use third-party cookies to track people and their activity across different sites, build up profiles of people and their interests, and then serve up relevant ads.

Those shoes you were looking at earlier now following you around the internet? That's third-party cookies at work. The upside is ads that fit your interests. The downside: increasing uneasiness about online tracking.

And while people have been fighting this system since the '90s, when the first ad blocker was launched, tech giants are finally heeding the consumer:

As consumers opt out of — and platforms do away with— traditional online data tracking, cookie-enabled advertising opportunities also decline.

Why shift to using first-party data

Much of the power behind advertising platforms like Google and Facebook lies in their machine learning algorithms. Campaign settings like audience targeting and budget affect how the algorithms learn — and conversion signals, in particular, teach the platforms what to ultimately aim for.

For a long time, cookies provided valuable signals — of intent to use for advertising like retargeting and of quality to feed the algorithm.

Without cookies, data doesn't get sent to the machine learning algorithms operated by Google, Facebook, and similar platforms. So companies have to start thinking about different ways to target and convert high-quality, high intent prospects.


The answer: take that tracking offline and put it in the hands of companies themselves.

According to Ian Maier, general manager of Clearbit Advertising, the future revolves around server-to-server conversion tracking. With this method, instead of relaying information to Google and Facebook via third-party cookies, the company sends it directly.

"As a business, you are collecting the data about the form submission, about what ad they clicked on, about what browser they're using, what their IP addresses, etc. And then you're sending that data server-to-server to Facebook or Google," Ian says.

"What that means is you're sending that data directly from your computer to their computer, which cannot be blocked by an ad blocker, by browsers, by operating systems."

Using first-party data ensures that you can leverage the platforms' algorithms to reliably target high-quality prospects. It allows you to use and scale conversion data but with a finer degree of control to teach the systems what to aim for, so you have a much better chance of converting quality leads into closed/won revenue.

"If we're looking for more form submissions for our product signup, or for a demo request, or for an ebook download, or in this case for a webinar registration, by sending that data, we are telling Facebook and Google what a good outcome is," Ian says.

"They're then going and looking at this massive pool of data that they have on their users and trying to pattern-match and predict who, based on historical patterns, is likely to convert if they show them an ad."

Feed the algorithm better quality signals

Algorithms need a large amount of data to function — and cookies used to be a critical part of their diet. The difference now is that you can feed in the right data on what to look for.

The downside to cookies is the lack of differentiation in what's getting sent as signals back to the platforms. Cookies don't really help algorithms distinguish the quality of a lead — seeing anyone who visits or fills out a form is an equally valuable lead, including students and spam bots.

"When we do that, we are just reinforcing bad behavior. You're reinforcing bad outcomes," says Ian.

Zoran Arsovski is the managing director at VertoDigital, a digital marketing agency that specializes in B2B performance advertising. His team has been working with companies like Google on new methods for generating high-quality leads without cookies.

Zoran explains that when you collect your own data, you can segment leads using their attributes. "Let's say we're trying to send offline conversions, we would look for something which is on a lead level," he explains. "You can select specific attributes that you would like, and this can be a combination of attributes."

That means you have granular control over the attributes of people you want ad platforms to find more of. For example, if someone starts a trial, uses two specific features, and shares that they work at an organization of 1,000+ employees, those are two attributes (engagement and company size) that you can identify as desirable.

Similarly, if you’re only targeting enterprise business leads, you can send those conversions along so that algorithm learns to disregard leads from small and medium-sized businesses.

Send your own data to Google and Facebook and then, you can cherry-pick exactly the type of leads you want more of.

Learn more: How to send offline conversion data to teach ad algorithms


Shift from quantity to quality

The examples above illustrate how you can reach good leads in a first-party data world. According to Zoran, this is a shift to a more dynamic, responsive, and accurate targeting model. Companies can identify the quality and status of every lead, feed that into the advertising platforms, and benefit from the resulting optimized performance.

"Once you start doing this, it means that you're moving away from ‘just bring me somebody who will fill in a form’ to somebody with some attributes and with some qualifications that I'm determining," Zoran says.

This dynamic method also enables a shift to value-based bidding, where leads can be better differentiated by potential ROI and even associated with a dollar value in order to optimize bidding.

"We are moving to a model where every lead has its own price, or its own value," Zoran says. "Instead of optimizing for maximized conversion bidding models, we can start to think of using return on advertising spend.

"If I'm investing a dollar in an advertising campaign, I'm trying to maximize that dollar. And getting the most out of that dollar means attracting the most valuable leads for me."

This goes back to the ability to set attributes, get creative, and send optimal signals back to platforms like Google and Facebook. "If you start doing this, then you're unlocking a whole new potential and way of thinking," Zoran says.

Don't try to beat the machine learning system

When relying on machine learning systems to optimize your advertising, you have to trust them.

The algorithms aren’t a Las Vegas casino where you count cards and trick the system. The truth of the matter is that machine learning works — but only when you keep things simple and send them the right signals. As Ian points out, it's really hard to get all the elements right manually and by yourself.

"There's a strategic mindset shift that we have to make, away from what I see is actually hurting a lot of advertisers," he notes. "It's this mindset of, 'I am going to try and game the system,' or 'I am going to try and basically be the machine learning algorithm' — 'If I can only find the perfect calibrated audience, or test these 10 different ad sets in one campaign and move around budget to find the ideal audience, the ideal amount of spend'...

"You're actually fragmenting the whole system. You're preventing the system from being able to learn in the first place."

The advancement of digital advertising arises from the progression in machine learning — and the end of cookies is not the end of the digital advertising line but a new beginning. It's a chance to get even better at it as a B2B marketer — leveraging technologies by feeding in higher-quality, more specific training data.

Watch Ian and Zoran's full video to learn more about B2B performance advertising in a cookieless world.



Buying signals: Definition + 15 examples

Marketing & Growthby Justin Tsang on June 06, 2023

Discover the art of identifying buying signals in sales and marketing. Check out 15 examples of buying signals and pro tips for handling them effectively.

Marketing analytics: a guide to measuring performance

Marketing & Growthby Emily Pick on May 25, 2023

Marketing analytics is the study of data to make sense of marketing efforts and improve ROI. Learn more about this crucial process in our guide.

The Standard in B2B Data

Now reinvented with Artificial Intelligence—Clearbit is the first AI Native Data Provider. Enrich your records, reveal buying intent, and connect with your ideal customers.

image-hero