What do you need before you open the account?
Four things, and gathering them first turns setup into a twenty-minute job rather than a week of half-finished attempts.
A business entity and billing details that match it. One landing page that does one job, with the offer above the fold. Working conversion tracking — the single thing most first campaigns skip and the reason most first campaigns teach nobody anything. And one sentence describing what a successful outcome is, agreed before you spend, so you are not deciding what "working" means after seeing the numbers.
If tracking is not in place, stop and fix that first. Reporting on this channel is thinner than Google's, so an untracked campaign leaves you guessing from click counts. You will spend the money and end up knowing nothing.
Choosing the objective
This is the decision with the longest shadow, and reach is the wrong default for almost everyone.
If you need enquiries, buy clicks and optimise towards the enquiry. If you sell online and your tracking is genuinely solid, conversion-optimised bidding is available — but it needs conversion data to learn from, so it is a second-phase choice rather than a first-week one. Cost per thousand impressions is for brand campaigns with a brand budget and a brand measurement plan. Most businesses reading this do not have one, and should not pretend otherwise.
Structuring the account so you can read it later
The most common structural mistake is one campaign carrying every audience and every ad, which produces a number at the end that cannot be attributed to anything.
Split by the thing you want to learn about. One campaign per offer or service line. Within it, separate ad groups for genuinely different audiences or intents — not fifteen near-identical splits, which just divides your data until none of it is significant. Name things so a stranger could read the account in six months.
A useful test: before launch, ask what question each campaign will answer. If two campaigns answer the same question, merge them. If one answers none, delete it.
Writing the creative
You are writing for someone mid-conversation who has already explained their problem. That is a different job from writing a headline against a keyword, and it changes the register.
Lead with the specific thing you do, in their words. Name the situation rather than the demographic. Say what happens next — a call, a quote, a demo — because vague calls to action underperform reliably. And resist writing to sound impressive: the reader is midway through solving something and will skip anything that reads like a brochure.
Generate more variants than you need and cut hard. Our ad copy generator is a reasonable way to get past the blank page, but the version you run should be the one you edited.
The first fortnight
Resist the urge to optimise on day two. Early numbers on a young account are noise, and reacting to noise is how people talk themselves out of a channel that would have worked.
- Days 1–3: check delivery and tracking only. Are impressions serving, are clicks arriving, do conversions register end to end? Fix plumbing, change nothing else.
- Days 4–10: let it run. Look daily, act rarely. You are accumulating enough data for a difference to mean something.
- Days 11–14: first real read. Which creative earned its spend, what your cost per outcome actually is, and whether that number works against your margin.
- Only then: scale what clears the number, stop what does not, and write down what you learned so the next test starts further along.
When to hand it over
If the honest answer to "who is going to look at this every week" is nobody, a managed setup will outperform a self-run one regardless of who built it — the discipline is the job, not the setup.
That is what our ChatGPT ads service is for, and the first conversation is an assessment of whether the channel suits your offer at all rather than a proposal.
Common questions
What do I need before launching a campaign?
A business entity with matching billing details, one landing page that does one job, working conversion tracking, and a written definition of what success means. The tracking is the one people skip, and skipping it is why most first campaigns teach nobody anything.
Which objective should I start with?
Clicks, optimised towards an enquiry, for most businesses. Conversion-optimised bidding needs existing conversion data to learn from, so it is a second-phase choice. Impressions are for brand campaigns with a brand measurement plan.
How long before I can judge the results?
Around a fortnight before the first honest read, and resist optimising before then. Early numbers on a new account are mostly noise, and reacting to noise is how people abandon a channel that would have worked.