Your Ads Are Getting Smarter. Your Audience Is Getting Smaller. That's Not a Coincidence.

Your Google Ads campaign is working. Leads are coming in. Cost per lead looks reasonable. The algorithm is doing its job.

And your business has stopped growing.

This isn't a coincidence. The same machine learning that finds your best converters is also shrinking your addressable market with every optimisation cycle. It's getting smarter at finding the people most likely to convert today. And it's getting worse at finding the people who could become customers tomorrow.

Sixty years of marketing science, a landmark analysis of 996 advertising campaigns, and decades of brand growth data all point to the same conclusion: the way ad platforms optimise campaigns is fundamentally at odds with how businesses actually grow. Understanding that tension is worth more than any bidding strategy or audience setting you'll ever change.

The Optimisation Loop That Shrinks Your Market

When you launch a campaign optimised for conversions, the platform's machine learning does something perfectly reasonable. It studies your first batch of converters. It identifies patterns in their demographics, browsing behaviour, device types, time of day. Then it shows your ad to more people who match those patterns.

Conversions come in. The algorithm's hypothesis is confirmed. So it doubles down, narrowing delivery further toward profiles that look like proven converters.

This is called overfitting in machine learning. Research from Cometly describes the pattern: overfitting occurs when the algorithm has found a small pocket of users that match your goal perfectly, but it isn't expanding beyond that group because the optimisation goal is too narrow.

For a small business spending $3,000 to $5,000 per month, the narrowing happens fast. There isn't enough conversion data to keep the algorithm exploring broadly. So it finds 1,000 to 3,000 high-probability converters and shows your ad to them. Again. And again.

Three things happen simultaneously:

• Your frequency climbs (the same people see your ad more often)
• Your reach flatlines (the algorithm stops looking for new people)
• Your cost per acquisition creeps up (you've saturated the only audience the algorithm is willing to serve)

And here's the part most businesses miss. Meta's own frequency documentation confirms that frequency is largely a function of audience size and spend. Lowering it requires expanding the audience or reducing the budget. You can't just toggle a setting. The narrowing is structural.

60 Years of Marketing Science Says Your Algorithm Has It Backwards

Here's where things get uncomfortable for anyone running conversion-focused campaigns.

Byron Sharp and the Ehrenberg-Bass Institute have studied brand growth across 130+ brands in 13+ product categories over six decades. Their conclusion, validated by both Nielsen and Bain & Company, is unambiguous: brands grow by acquiring new buyers, not by showing ads more often to existing ones.

The data is striking. Light buyers, people who purchase once or a few times a year, make up 70 to 80% of any brand's customer base. They're not your "best" customers on any spreadsheet. But they are your growth engine.

When Sharp analysed 880 IPA Effectiveness Award-winning campaigns, he found that 82% achieved growth through penetration (reaching new buyers). Only 2% grew primarily through loyalty strategies. We've explored why the loyalty trap keeps SME marketing focused on the wrong people before. The reach problem is the mechanical mirror of that strategic error.

Les Binet and Peter Field reinforced this through the IPA Effectiveness Databank, analysing 996 real-world campaigns. Their finding: broad-reach campaigns drive market share. Narrow, targeted campaigns drive efficiency metrics but not business growth.

Their most recent analysis, presented at the 2025 IPA Effectiveness Conference, delivered a statistic that should make every performance marketer pause: budget is eight times more important than ROI when driving effectiveness. ROI accounts for only 11% of the variation in campaign payback. Budget, which determines reach, accounts for 89%.

Read that again. The single factor that drives growth eight times more than return on investment is how many different people you reach.

The table below captures the fundamental conflict:

The Maths Problem Nobody Talks About

Let's make this concrete for a real business.

A plumber spending $3,000 per month on Google Ads at an average CPC of $5 gets roughly 600 clicks. Assume those clicks come from a pool of around 20,000 impressions in their service area.

On average, that's a monthly frequency of 3 to 4 per unique user. But the algorithm doesn't distribute evenly. It shows the ad disproportionately to people it predicts will click. A core group of maybe 2,000 to 4,000 people sees it 10 to 15 times per month. Everyone else in the market barely sees it once.

Meanwhile, customer acquisition costs have risen 222% over the past eight years, with an 18.4% jump in 2025 alone. The algorithm's response to rising costs? Show ads to even more "guaranteed" converters. Which makes the audience narrower. Which pushes frequency higher. Which makes costs rise further.

This is a death spiral for growth. And it's invisible from your dashboard because the CPA still looks acceptable. You're just not reaching anyone new.

As we explored in why your ads get more expensive every year, cost escalation isn't purely a platform pricing problem. Part of the issue is structural: your own campaign is competing against itself for the attention of an audience that's already oversaturated.

The Difference Between Showing Up and Stalking

The solution isn't to abandon conversion optimisation. You need leads. But you need to understand the difference between sufficient frequency and wasted frequency.

Avinash Kaushik's research on brand marketing measurement identifies 3 exposures per unique user per week as the threshold where advertising starts to genuinely move preference. Below that, you're building vague familiarity. At that level, you're building the kind of mental availability that makes someone call you first when the need arises.

The Ehrenberg-Bass Institute's work on the mere exposure effect supports this. Sharp notes that "people prefer things they have been exposed to more" and that even "fleeting exposures can do good when pointed at distinctive brand assets." You don't need someone to watch your entire video or read your whole ad. You need them to recognise you quickly, repeatedly, across their week.

But here's the crucial distinction: 3 exposures per week to 15,000 people builds a brand. 15 exposures per week to 2,000 people builds resentment.

Rory Sutherland's costly signalling theory) explains why this matters psychologically. A business that shows up consistently across multiple contexts, in search results, on display, in social feeds, in review sites, signals confidence and scale. The broad presence is itself the message: this is a business that invests because it believes in what it sells. It's the commercial equivalent of a flower investing energy in bright petals. The investment signals value.

A business that follows the same 2,000 people around the internet with the same ad doesn't signal confidence. It signals that nobody else is paying attention.

The 95/5 rule makes this concrete. At any given moment, only about 5% of your potential market is actively looking to buy. The other 95% are future buyers. If your algorithm only targets the 5% who are ready now, you're competing at maximum cost for minimum volume, while your competitors build mental availability with the 95% who will buy next month, next quarter, next year.

Sam Tomlinson's research on Meta's Andromeda algorithm adds a practical lever. Truly diverse, audience-centric creative outperforms incremental variations because different creative reaches different audience segments. If you're running one ad to your narrowing audience, you're accelerating the saturation problem. Running five genuinely different ads forces the platform to find different pockets of people, naturally expanding your reach without changing a single targeting setting.

What This Means for Your Business

Understanding the conflict between platform optimisation and growth science doesn't mean ditching Google Ads or Meta. It means managing the tension deliberately rather than letting the algorithm make the growth decision for you.

The platforms will always optimise for what's measurable in the short term. Your job is to optimise for what actually drives growth in the long term. That means deliberately fighting the algorithm's natural tendency to narrow, and reaching the people who haven't heard of you yet.

Because the maths is clear. You cannot grow a business by showing ads to the same people, no matter how smart the algorithm gets at finding them.

Further Reading

• Ehrenberg-Bass Institute Research - Six decades of evidence on penetration, reach, and why light buyers drive brand growth
• The Long and the Short of It by Les Binet and Peter Field - IPA Effectiveness Databank analysis of 996 campaigns on the balance between brand and activation
• IPA: Balance Efficiency and Effectiveness or Risk a Marketing Death Spiral - Latest 2025 IPA research showing budget is 8x more important than ROI for effectiveness
• Brand Marketing Measurement by Avinash Kaushik - Kaushik's framework for frequency thresholds and brand impact measurement
• Deep Marketing: Light Buyers and Brand Growth - Practical breakdown of how light buyer acquisition drives growth, updated for 2026