Intro
In the first part of the article — which I wrote nearly a year ago, but as the saying goes, good work takes time — I explained what machine learning is: the foundational concept behind how artificial intelligence works, which is, in fact, surprisingly simple.
I also emphasized that although various A.I. tools demonstrate astonishing, almost magical performance, they are essentially just extremely complex calculators that are incapable of thinking or creativity.
In this second part of the article, I’ll reveal how we can most effectively utilize machine learning in our PPC campaigns, and, using this as an excuse, I’ll also include a hefty dose of PPC history.
AI is still just a super-advanced calculator
PPC Before A.I. – the era of manual targeting
For “old-school” PPC professionals like me — those who learned the field around 2020 — the transition to using A.I. targeting was accompanied by significant anxiety. The reason: the power of habit.
Although machine learning optimization was already available in PPC accounts back then, the technology was still in its infancy, and as such, its use typically led to poor results.
At the time, only careless PPC managers opted for machine targeting and optimization. They had to rely on experimental theology: “If the return is good, God exists!” (It usually wasn’t.)
In contrast, diligent PPC experts would develop and set up audience targeting based on historical data, competitor research, and persona analysis. Then, after launching the campaign, they would manually fine-tune bids and bid adjustments through continuous optimization. Monitoring results closely and adjusting the “knobs” accordingly delivered excellent results.
This is what PPC felt like before AI
It’s also worth noting that online advertising platforms offered far more options back then. Ad account managers had access to much richer and deeper data than the superficial insights available today. Furthermore, advertisers could define much more detailed and specific audiences based on online habits, interests, etc. With clever combinations, we could narrow down the ad reach to extremely precise audiences.
Back then, I used to joke that I could even target people who take their coffee with two sugars — and that was close to reality.
As a result, knowledge sharing among experts in the field mainly revolved around crafting hyper-specific audience combinations and methods for pattern recognition within historical data sets.
For example, when advertising a luxury product, targeting the intersection of relevant interests and frequent international travelers on Facebook Ads worked well. Or in the case of Google Ads, to address the common challenge of knowing which keywords would drive valuable purchases before any had occurred, a great solution was to evaluate how long users arriving via those keywords spent on the website.
These methods worked. It was beautiful. And we rejoiced.
The transition period – how A.I. was received
In this atmosphere — where manual settings were glorified and reliance on machine learning was frowned upon — we initially didn’t take the first reports, studies, and recommendations about improved A.I. seriously. We met them with skepticism. We didn’t want to believe it, but eventually, so many such signals came in that we decided to give it a try.
We started cautiously, testing it first only in our oldest and highest-spending accounts, the ones with the most data. And it worked.
When we tried it in smaller accounts, it didn’t work. So we retreated and reserved the method only for large accounts.
Then came new reports claiming that A.I. now also worked well for smaller accounts — if they had at least some historical data. We checked — it worked well.
Later, it was said that A.I. would work even in brand new accounts because the system now considered not only the data from that specific account but also data from all similar advertiser accounts.
We tested it — it worked well.
Reluctantly, we became friends with AI
But not everyone experienced the same results. At that point, the industry split into two groups: those who clung to manual targeting and optimization due to bad experiences with the new method, and those who managed to harness A.I. effectively.
The “manual” group fell behind badly. A project manager at a leading agency told a story about a PPC specialist who had worked in the field for over 10 years. Despite repeated requests and worsening results, he insisted on manual optimization until the very end. He couldn’t be persuaded, so they eventually had to let him go.
His replacement — a much less experienced PPC manager — produced better results right away, simply by switching to machine learning.
As I wrote at the beginning of my first article:
“…what is a best practice today could become an excellent way to burn money in just three months. Hence, staying up-to-date with industry changes is crucial.”
(Stories like these are part of why I often repeat my belief: in PPC, after six months you’re a junior, after a year and a half you’re a medior, after three years you’re a senior — and any additional years are irrelevant. After all, everything you knew more than three years ago is outdated and only fuels nostalgic “back in my day” stories.)
There’s a very good reason why many people were able to use A.I. effectively from the start and had good experiences with it, while others simply couldn’t manage the same.
And that reason goes far beyond whether someone was unwilling to make a two-click change in their ad account settings.
The key to A.I.-powered PPC
What distinguished those who successfully transitioned to using A.I. in PPC from the very first step was that they consistently paid attention to the creative aspects of PPC. That is, they invested just as much energy into perfecting their ad copy and visuals as they did into technical settings and data-driven decision-making.
Those who followed this approach were able to quickly adapt to the creative demands that come with using A.I. — because this is the area where the rules of the game were most thoroughly rewritten.
To put it simply, in the past, you could “get away with” generic ads as long as your manual targeting was precise. But with A.I., this leads to disaster.
Let’s remember that the core function of A.I. is to detect patterns in large data sets — and it does this orders of magnitude better than humans. Not because it’s more intelligent. Human intelligence is complex; artificial intelligence is optimized for a single task. Even the simplest pocket calculator can extract square roots faster than the world’s best mental calculator — and that has nothing to do with intelligence.
Taking all this into account, the following example makes it clear how A.I. can be used effectively in PPC — and how it can go very wrong.
How to do it wrong
Let’s say the advertised product is a hyper-specific niche software designed to assist with bookkeeping for small convenience stores. Targeting is entrusted entirely to machine learning.
The creative assets, however, are generic: men in suits and women in blazers smiling delightedly at the camera, with ad copy like “Reduce your business expenses” and other such catchphrases.
Such generic creatives no longer work
The system delivers these ads to a broad audience, including many business owners. Since most business owners are interested in cutting costs, many will click. But very few — or none — will convert, because (hopefully) the landing page makes it clear that the product is irrelevant to them.
Yet the algorithm will enthusiastically continue showing the ads to people similar to those who already clicked — in other words, more business owners of any kind. But the statistical likelihood that they run a convenience store is very low.
As a result, the people clicking won’t convert. The system doesn’t get data on who should be targeted for conversions, but it does collect a pile of redundant data — containing misleading patterns.
In short: the campaign learned the wrong things, and we wasted our money.
How to do it right
The product is the same, and targeting is again entirely handled by machine learning. But this time, we craft creative content specifically aligned with the product.
The visuals show people stocking shelves in a small shop, and the ad copy reads “Make bookkeeping easier for your convenience store.”
If you sell software to convenience stores, address them directly!
The system again starts showing the ad to a broad audience. But now, the vast majority of people simply scroll past it — and that’s great!
Why? Because this gives the algorithm a large sample of people who shouldn’t be shown the ad. It can quickly identify common traits among them and begin learning in the right direction.
Soon, the ads stop appearing for those users, and instead are shown to a much narrower group. Most of them will still scroll past — but the few who click are far more likely to be actual convenience store owners. Because the specific creative and messaging speaks only to them.
A portion of this relevant traffic will convert on the landing page — at first a few, then more and more. Eventually, enough conversions accumulate for the algorithm to identify shared traits among the converters, allowing it to further refine targeting and serve ads to people who are statistically much more likely to convert.
In short: the campaign learned the right things, and we achieved a strong return on investment.
The task: speak only to your target audience
The task is extremely simple: we need to use ad copy and creative materials that resonate exclusively with our target audience.
Of course, there’s a huge difference between simple and easy. This is a significant challenge.
It’s a necessary but not sufficient condition for the ad to resonate with the target group. If it resonates with anyone besides the target group, it will mislead the algorithm.
We need to craft copy and visual/video content that not only grabs the attention of our potential clients but also actively repels everyone else.
Here, the principle of “one person’s trash is another’s treasure” applies in full force. For example, if you’re advertising an advanced Forex trading course, using a candlestick chart as the visual is a bullseye.
This is the kind of thing that makes a Forex trader’s eyes light up — they’re constantly looking at such charts — while almost everyone else finds it to be a boring I-do-not-care thing and scrolls right past it. And that’s exactly what we need.
By capturing this kind of exclusive attention with our ads, we feed the algorithm high-quality data.
From that, it will find the patterns we need more effectively than any human could.
This leads to significantly lower cost-per-conversion and a higher return on investment than what even the most precise manual targeting could achieve.
In summary – why it’s worth switching to A.I.
Using A.I. effectively in PPC requires a radically different approach than the manual methods we were used to before.
But it’s absolutely worth adapting, because we can achieve much better results this way.
If someone masters a 17th-century flintlock musket with great precision, it would clearly be a mistake to switch back to the musket after being handed a modern automatic rifle — just because learning to use the new weapon seems like too much work.
A.I. is the same.
It’s better, more efficient, faster.
You just need to know how to use it right.
Contact us!
The Digital Tailors team is constantly staying up to date with the latest innovations in digital marketing platforms.
If you’ll settle for nothing less than the most effective campaigns, don’t hesitate to reach out to us!
