We've watched teams spend three weeks adjusting bids, tweaking assets, and restructuring budget allocations on a campaign that was always going to underperform, because the original setup made success structurally hard.
Optimization is what most marketers think they get paid to do. It's also what they spend the most time on. But here's the unflattering truth. In the campaigns we've audited, somewhere between 60 and 75% of the eventual outcome is locked in before the campaign goes live.
This post is the pre-launch checklist we use, why each question matters, and how to actually run it without turning your team into a planning bureaucracy.
Setup vs. optimization
Setup is the four hours of decisions before you publish:
- Who specifically is this for? (audience definition)
- What user intent are we matching? (search intent, life-stage, problem-solving, browsing)
- What's the offer? (hook, value, proof, CTA)
- What does the journey from click to convert look like?
- How will we know this is working in 2 weeks? In 6 weeks?
Optimization is everything you can change after launch: bid strategy, asset mix, schedule, audience tweaks, budget reallocation. It's the knobs the platform exposes.
The platform makes the optimization knobs feel high-leverage. They aren't. They're high-frequency. There are more of them, you touch them more often, so they fill up your day. But the magnitude of impact per touch is small compared to the setup-stage decisions you only make once.
The six questions
Before any campaign goes live, we ask the same six questions. If two or more come back fuzzy, we don't launch. We go back and sharpen the setup.
1. Intent mapping
What is the user thinking when they encounter this ad?
Not "what do they care about" in general. What are they doing in the thirty seconds before the ad appears? Are they searching for a solution, browsing for inspiration, killing time on YouTube, comparing two specific products? Each of those is a different mental state, and the ad has to meet that state, not the abstraction of it.
The fastest way to surface fuzzy intent: write the user's literal search query (or the literal scenario they're in) on a sticky note. If you can't, the campaign isn't ready.
2. Segmentation
Is this a campaign other accounts couldn't run?
If yes, you have a real angle. If anyone in this category could run this exact ad, you're entering on platform default rules and competing on bid. Most accounts that are "optimizing for ROAS" are really just paying more for clicks because their angle isn't differentiated.
The test: imagine three competitors saw your ad creative without your logo. Could one of them run it tomorrow? If yes, sharpen.
3. Message coherence
Does the keyword or audience signal, the ad copy, and the landing page tell the same story without effort?
Read all three out loud, in sequence. If you have to explain the connection, the user won't make it either. The handoff between intent → ad → page is where most campaigns leak. We've seen accounts with 8% click-through and 0.4% conversion rates because the page didn't deliver on what the ad implied.
A useful frame: write a single sentence that connects the three. "Person searches X, sees ad about Y, lands on page selling Z." If X, Y, and Z don't sound like the same thing, you have a coherence problem.
4. Friction audit
Is the post-click experience built for this specific intent, or is it the generic homepage doing double duty?
"We'll fix the landing page later" is the most expensive sentence in the category. We've watched teams spend $40,000 in paid clicks driving traffic to a page that loads in 6 seconds, has the wrong headline above the fold, and asks for a phone number before the user knows what they're buying. Optimizing the campaign while ignoring the page is like sharpening one knife in a dull set.
The audit: open the landing page on the slowest phone you have, on a flaky connection. If you'd close the tab in five seconds, your users do too.
5. Measurement plan
What's the leading metric we'd watch in week one? What's the lagging metric we'd watch by week six?
If they're the same metric, you don't have a measurement plan, you have a reflex. Leading metrics tell you the system is moving (CTR, qualified clicks, cost per qualified click). Lagging metrics tell you the system is working (CAC, blended ROAS, retention of the cohort acquired). You need both. Confusing them is how teams pause campaigns that were about to work and ride dead campaigns into the ground.
6. Kill criteria
At what point does this campaign get paused, and based on what?
This matters more than people realize. Campaigns that don't have explicit kill criteria tend to live forever, accumulating spend and noise. Write the kill rule before launch. Make it specific: "If after $X spend the qualified-action rate is below Y%, pause." It's much harder to argue your way out of a written rule than out of a feeling.
How to run this without slowing the team down
The honest objection: "we don't have four hours per campaign." Fair. The checklist isn't designed to add four hours; it's designed to reveal whether the four hours of clarity already exist somewhere, or whether the campaign is being launched on instinct.
The actual practice we use: run the six questions in a 25-minute review, not as a planning session. If five out of six are answered crisply, ship. If two or more are fuzzy, fix the brief, not the bidding strategy. Most "fixes" take 10 to 20 minutes, not hours.
The campaigns where the fuzziness can't be resolved in a 25-minute review are the campaigns that wouldn't have worked anyway. Those are exactly the ones worth pausing before they spend.
Why this is hard
The honest reason most teams skip the setup discipline: it doesn't feel like work.
Sitting and thinking for two hours about audience definition feels like procrastination. Tweaking bid strategy in the platform UI feels productive. There's a UI, there are buttons, things change.
But "productive-feeling work" and "high-leverage work" are different things. The setup work is high-leverage and feels slow. The optimization work is low-leverage and feels fast. Most teams optimize for the feeling.
We did this ourselves for a long time before we caught it. The teams that finally turn it around are the ones who get comfortable with two hours of "nothing happening" looking like work, because what's actually happening is the campaign quietly becoming twice as likely to succeed.
Before asking how to improve performance, ask whether the campaign was designed to learn anything useful in the first place. The answer is more often "not really" than anyone wants to admit.
That's the work.
KaiNet · Operator lessons