Campaign work where the hard part is not the copy. It is the floors that decide whether a result means anything, the published caps and match rules nothing in the interface enforces for you, and the difference between a documented number and a repeated one. How we test →
Written, reviewed and free to take. No run behind them, so no claim about what they do to an output. What that means.
Most paid advertising advice in circulation is copywriting advice with a budget attached. It tells you to lead with a benefit, to test three headlines, to speak to the pain. All of that is true and none of it is the reason campaigns fail.
Campaigns fail on arithmetic that nobody does. A budget too thin to produce enough results for any cost-per-result reading to mean anything, so a verdict gets made on noise. An audience below the size floor at which the platform's own optimisation behaves sensibly. A landing page converting at a rate that guarantees a loss before the first impression is served, because cold traffic will never convert better than people already searching for you by name. A fourth creative variant added out of thoroughness, which divides the same fixed budget four ways instead of three and drops every variant further below the threshold at which any of them could be judged. A mid-flight edit to a live ad set, made on day two out of impatience, which resets the delivery system's learning phase and destroys the very test it was meant to improve.
None of those are copy problems and no amount of better copy fixes any of them.
So the skills here are built around the numbers and the specifications rather than around the words alone. One of them computes what a campaign needs before it can produce a readable answer and ends in a numeric branch: at this threshold you scale, at that one you kill, and in this middle band you have genuinely not learned anything yet and the correct action is to wait rather than to act.
The rest work the second problem, which is that most published paid advertising numbers are either folklore or were true about an interface that has since been rebuilt. Both platforms replaced their ad formats wholesale inside the last four years, so a large share of what circulates describes objects that were removed. These files separate what the platform documents from what the field repeats, quote the fields, caps and match rules as the platform states them, mark their own rules of thumb as their own, and date every figure they carry, which is the only way a number in this subject stays honest for more than a season.
Two kinds of number live in these files and they are not interchangeable. A published limit, 30 characters in a Google headline, 50 search themes in an asset group, is a fact about the tool and your own history cannot overrule it, only a change to the tool can. A threshold, a floor or a starting budget is a rule for an account with no history of its own, and once you have ninety days of your own delivery data, your own cost per thousand impressions and your own cost per acquisition should replace it. A starting rule exists to stop you flying blind, not to override evidence you have since collected.
One honest limit across the whole category. None of these skills can see your ad account. They cannot read your delivery data, verify your tracking, or tell you whether an event is firing correctly. A confident-looking campaign optimising against a broken conversion event is a real and common failure, and diagnosing it is an engineering job rather than a marketing one.
Measured means the skill was given a realistic task on real material, then the identical task was run again with the skill removed, five runs each way. Each output was graded alone, against a rubric written by someone who had never seen the skill, by a session that was not told the other arm existed. Whether it passed was decided by a rule written down before any run executed. Those pages carry the worst case, the median, the p value, and what the skill costs you as well as what it buys.
Not yet measured means exactly that. It is written, it has been read, it is free to take, and we have run no experiment on it, so we make no claim about what it does to an output. It is not a skill that failed. Skills that failed are not published at all, in either state, and their numbers are in the results table.
Measuring one skill properly costs roughly twenty model sessions. We are working down the queue and moving skills from the second group into the first. Read the full method, or go back to all skills.