ЛОБ (Лабораторія Онлайн Бізнесу) ran online from Kyiv, Ukraine, on October 6-8, 2023; the speaker joined remotely from Manchester. Igor Ivitskiy presented the session in Ukrainian.
The talk, “Effective Google Advertising in 2023: How to Pay Less and Get More Customers”, connected three problems: the loss of literal control through keywords, the underuse of intent-rich audience signals and the damage caused when fake leads teach an automated system to find more fake leads. It reframed keyword research for Google Ads as one part of a wider intent and data-quality problem.
The real shift was from keywords to intent
A search can contain the product name and still express the wrong intent. Another search can describe a problem without naming the product, yet come from someone who is ready for the solution. That gap matters because Google’s keyword matching can connect ads with related queries rather than only literal repetitions of a keyword.
Igor’s strategic response was to stop treating the keyword list as a complete map of demand. Wider reach can expose the long tail, while evidence from the account is used to remove traffic that does not create value. This is the business logic behind combining broad match with carefully governed exclusions, including negative keywords.
The same principle extends beyond search terms. Audience signals help the system distinguish people by likely intent, while verified conversion outcomes tell it which apparent responses deserve to be repeated. Automation becomes useful when the signals describe the business result, not merely an easy action on a form.
Look past the literal query to the problem and commercial intent behind it.
Use signals of real interest instead of relying only on broad demographic labels.
Let real customer value, not the cheapest recorded action, guide optimization.
How I think about efficient reach
I do not expect a keyword to reveal a person’s full intention. I use wider signals to find demand, then judge the traffic by whether it produces a real business outcome. Reach is valuable only when the feedback given to the algorithm is trustworthy.
This way of thinking also shapes how our school teaches traffic quality and negative keywords: as parts of an economic system, not isolated interface settings. The free webinar introduces the same analytical perspective without promising that one setup fits every account.
Key insights
- The cost of the old playbook became visible. Average CPC in the account shown rose from UAH 3.41 in Q2 2020 to UAH 27 in Q1 2023, close to an eightfold increase.
- Intent can contradict the wording of a query. A product phrase can carry the wrong goal, while a problem-focused phrase can reveal a buyer who never names the product.
- The long tail contains most search traffic. Igor cited 70% of internet search traffic as rare queries, where competition can be lower than on the obvious top terms.
- Audience quality depends on useful signals. Generic interests describe almost everyone; signals connected to current intent can make automation more discriminating.
- A cheap conversion can be an expensive mistake. One campaign produced 22,000 recorded conversions at $8 each but no sales, after roughly $180,000 had been spent.
- Automation amplifies the feedback it receives. If fake registrations are treated as success, the system can learn to seek more of the same traffic.
Key moments from the talk
Why the keyword-first model stopped being enough
Igor opened with a number that explained why familiar advice felt less useful. In one of his largest accounts, average cost per click had moved from UAH 3.41 in the second quarter of 2020 to UAH 27 in the first quarter of 2023. Currency movement could explain only part of the change; in the account itself, the increase was close to eight times. If advertisers kept approaching the market with the same method, every old inefficiency became far more expensive. His criticism was not that keywords had become irrelevant. It was that a discipline built around finding the perfect literal phrase could no longer carry the entire job.
The talk illustrated the mismatch with a simple pair of searches. Someone searching for “wipers in Kyiv” could be looking for a cleaner rather than a car part, despite using the advertiser’s apparent keyword. Someone describing rain blocking the view from a car might need new wiper blades without naming them at all. The commercial signal therefore sits above the words. Igor stated the principle on stage in Ukrainian: “Секрет сучасної реклами: працювати не тільки на рівні ключових слів, але й на більш високому рівні, на рівні намірів.” That shift explains why literal relevance and buyer relevance are not always the same thing.
Where cheaper demand can hide
The obvious terms are expensive because every competitor can name them. Less obvious searches form a much larger and more fragmented field. Igor said that 70% of internet search traffic comes from the long tail: millions of queries that appear rarely on their own but matter in aggregate. The strategic opportunity is not to predict every rare phrase. It is to give the system room to discover demand while maintaining an evidence-based boundary around what the business will pay for. He compressed that idea into another line: “Я просто беру широкі загальні ключові запити і відсікаю все зайве.” The talk contains the detailed mechanics, but the publishable principle is broader: explore widely, then let economic evidence define waste.
Audience signals were the second part of the same argument. Age, gender and generic interests can produce a neat-looking audience while saying very little about what a person wants now. Google observes many signals that may indicate current intent, and Igor argued that advertisers should use that richer layer rather than settle for categories broad enough to include almost everyone. The point was not to collect more settings for their own sake. It was to move the optimization problem closer to the customer’s actual situation. Keywords, audiences and bidding then become different sources of evidence about the same question: who is likely to create value for the business?
When the conversion signal turns against the campaign
The final case showed why wider reach and stronger automation need trustworthy measurement. A proven project expected registrations at $16 each. The new campaign appeared to perform twice as well, producing 22,000 recorded conversions at $8. The team began planning to scale, but the sales department later reported that every lead was dead: no payments and no credible deals. About $180,000 had been spent in ten days. Igor’s concern went beyond the immediate loss. “Проблема насправді не в самому склікуванні. Проблема в іншому: Google дуже розумний.” When an algorithm sees fake registrations as cheap success, it can optimize toward sources that generate even more of them.
This is the boundary that ties the whole presentation together. Google can find patterns at a scale no manual keyword list can match, but it does not independently know which recorded action represents a genuine customer. The advertiser has to connect optimization with verified business quality early enough to prevent a false feedback loop. The right standard is not a low dashboard CPA on its own, just as the right search is not necessarily the one that repeats the product name. In both cases, the surface metric can mislead. Intent and downstream value are the stronger signals, and the detailed procedures for working with them remain in the full session.
The full recording of this talk is included in our bonus pack for students.
Key takeaways
Modern Google Ads performance depends on reading intent beyond literal keywords and giving automation richer audience signals.
The decisive safeguard is conversion quality: the system will scale whatever the advertiser labels as success, including false leads.