The dangerous thing in digital marketing isn't what you don't know.
It's what you're sure you know — and that used to be right.
In 2020, these rules worked. They settled into your gut, became “just how it's done,” and nobody stopped to check whether they still held up. The catch: the market moved. Cookies died, attribution changed, the algorithms got smarter, and a few of your most basic assumptions quietly turned into liabilities.
They don't announce themselves. They don't crater the account in a single day. They just take a small cut of every conversion — a tax that compounds into months of wasted budget.
We pulled together seven of them. Not theory — seven assumptions we see running in real accounts, every week. For each one: why it was true, why it isn't anymore, and what's true in its place.
How many are alive in your account? Let's count.
Why old instinct is more dangerous than not knowing
Not knowing is honest. When you don't know something, you ask, you check, you proceed with care.
Outdated instinct does the opposite: it feels like knowledge. It hands you the confidence to decide fast — to approve a budget, pause a campaign, pick a keyword — without stopping. And when that confidence rests on a rule that was right six years ago, every “gut call” nudges the account a little further in the wrong direction.
That's the real contest in 2026. Not between people who know and people who don't — but between people who've updated their instinct and people still running on the 2020 version.
Here are the seven points where that gap costs the most.
Assumption 1 — “The last click tells the whole story”
Why it was true in 2020
Customer journeys were shorter, and cookies still worked. Last-click attribution was the default in every platform, and that was enough to know where conversions came from. You gave credit to the final touch, and it felt fair.
Why it hurts in 2026
Today the journey splits across devices, platforms, and days. Last-click hands 100% of the credit to the final touch — usually branded search or remarketing — and 0% to everything that created the demand in the first place. The result: you pause exactly the top-of-funnel campaigns that feed the bottom, because they “don't convert.” You reward the campaign that picked the fruit and starve the one that planted the tree.
What's true instead
A data-driven attribution model (Data-Driven Attribution, DDA) that distributes credit across the whole journey based on each touchpoint's real contribution. Not perfect — but far closer to reality than last-click. The right call is to see the entire funnel, not just the last line in it.
Assumption 2 — “More keywords = more results”
Why it was true in 2020
The more keywords you covered, the more impressions you captured. Broad was cheap, competition was thinner, and volume really did translate into traffic. A long keyword list felt like good coverage.
Why it hurts in 2026
Match types have expanded dramatically. Broad match now picks up variations, synonyms, and intents you never chose — and drags budget into searches that have nothing to do with you. A long, unfocused list isn't coverage; it's a leak. You're paying for clicks from people who never intended to buy.
What's true instead
Target by intent, not by volume. Fewer keywords, chosen for buying intent, with tight management of your search terms and a disciplined negative keyword list (negative keywords) that trims away the irrelevant. In a world of AI and broad match, your negative list is the steering wheel — not the list of keywords you bid on.
Assumption 3 — “Low CPL = success”
Why it was true in 2020
When the funnel was simple and leads were uniform, a low cost per lead (CPL) genuinely meant efficiency. Pay less per lead, get more leads on the same budget. A clean metric, easy to measure.
Why it hurts in 2026
The automated systems can get you cheap leads — easily. They'll simply find the people most likely to fill out a form, not necessarily the people likely to buy. A low CPL next to a low close rate isn't success; it's efficient waste. You're optimizing yourself deeper into bad leads.
What's true instead
Measure further down the funnel: cost per qualified lead, cost per acquisition (CPA), and ultimately customer lifetime value (LTV) against acquisition cost. A lead that costs more but closes — and is worth three times as much over time — isn't expensive. It's profitable. To get the system optimizing toward that, you have to feed it a quality signal (offline conversions / conversion values), not just “form submitted.”
Check yourself before you read on
The first three assumptions are alive in almost every account we open. If you spotted even one of them in yours, it's worth quantifying what it costs you per month before you start fixing things at random.
Spotted one in your account? We'll run a diagnostic on your account and put a number on the leak — no commitment, and an honest answer about whether we're the right fit.
Put a number on the leak in my accountAssumption 4 — “You can trust cookie-based tracking”
Why it was true in 2020
Third-party cookies were the backbone of measurement. They followed the user across sites, enabled precise remarketing, and powered cross-platform attribution. You leaned on them without a second thought — and it worked.
Why it hurts in 2026
The cookies are gone. Browsers block them, users decline them, and regulation keeps narrowing what's allowed. Every conversion you don't measure isn't just a missing row in a report — it's a faulty signal you're feeding into the AI systems at Google and Meta, which use it to decide where your budget goes. Partial data in 2026 costs you twice: once in your own wrong decision, and again in the machine learning the wrong lesson.
What's true instead
A measurement stack built on first-party data: server-side tagging that routes conversion events through your own server, and Consent Mode v2 that keeps measurement compliant even when a user declines consent. The teams that have moved to this stack are measuring better than ever — not less. (We go deeper on the full stack in a separate piece on post-cookie measurement.)
Assumption 5 — “High ROAS is always good”
Why it was true in 2020
ROAS (return on ad spend) was the metric that lined everyone up: how many dollars you brought in for every dollar you spent. High = good, low = bad. Simple, and right most of the time.
Why it hurts in 2026
High ROAS usually comes from an audience that would have converted anyway — branded search, remarketing to people already on your site. The system takes a bow for sales that would have happened with no ads at all. That's the absence of incrementality: you're paying for credit, not for impact. A ROAS of 8 on an existing audience can be less profitable than a ROAS of 3 that brings in genuinely new customers.
What's true instead
Measure incrementality — the added value the advertising actually created — through geo-tests, conversion-lift studies (conversion lift), or controlled budget holdbacks. And look at margin, not just ROAS: a profitable sale at a middling ROAS beats a thin-margin sale at an impressive one. The prettiest number in the report isn't always the money in the bank.
Assumption 6 — “Bigger budget = bigger results”
Why it was true in 2020
When a campaign was profitable and the market was wide, adding budget really did add results — more or less linearly. You saw a winner, poured more on it, and it grew. The logic held.
Why it hurts in 2026
Every campaign has a saturation curve. Past a certain point, each additional dollar buys a less relevant audience, CPA climbs, and ROAS drops — sometimes within days. Doubling the budget on a winner all at once also resets the algorithm's learning phase and damages the very performance you were trying to amplify. You break exactly what was working.
What's true instead
Scale in steps (often 15–20% every few days), watch your marginal CPA — the cost of the additional conversions, not the average — and expand sideways when a campaign saturates: new audiences, new platforms, new demand. Growth is a controlled process, not a faucet you open all the way.
Assumption 7 — “The prettiest ad wins”
Why it was true in 2020
Strong creative always helped, and in a less saturated market, a polished ad stood out. You invested in great design, and it was a real edge. “Beautiful” and “effective” felt like the same thing.
Why it hurts in 2026
Creative is now the number-one performance lever — but “beautiful” and “converting” are not the same thing. In the era of Advantage+ and Performance Max, creative is the main tool you still use to influence the machine. An ad that looks stunning to the boss and rests on a hunch — not a test — is a bet. And that bet repeats itself in every account that swaps creative based on “what looks good.”
What's true instead
Creative testing as a system, not an event. A steady stream of variations — hook, angle, format, proof — tested against a real metric, where the winners win in the data and not in the meeting. In a world where targeting is automated, creative is the last front where human skill genuinely moves the needle.
The checklist: how many of the 7 are alive in your account?
Go through your account and check the boxes. Every “yes” is a potential leak in your budget:
- Attribution: is the account still running on last-click (instead of DDA)?
- Keywords: long keyword list, but a thin negative list and search terms nobody reviews?
- CPL: measuring success by low CPL, with no eye on close rate / LTV?
- Measurement: still leaning on cookies, with no server-side tagging or Consent Mode v2?
- ROAS: judging campaigns on ROAS alone, with no test for incrementality?
- Budget: scaling winners in big jumps?
- Creative: picking ads by “what looks good,” with no systematic testing against a metric?
0–1 checked: your account is relatively up to date. Worth sharpening the edges.
2–4 checked: typical — and there's budget to recover here. Worth quantifying how much.
5+ checked: the account is being run on 2020 instinct. Every month like this costs real money.
No matter how many you checked, the next question is always the same one: what is this actually costing me? You can't guess that. You measure it.
