agentleFS
hermoso-aihermoso-ai/marketing-skills/skills/campaign-analyst/SKILL.md

campaign-analyst

Skill0 starsChanged 21 days agoMITView on GitHub

Read a paid-ads account like an analyst. Turn 14 days of campaign results into what to scale, what to pause, what to fix and what to test next, with a data floor that refuses a verdict on too little delivery.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Put every campaign (or ad set, or ad) in ONE table, whatever platform it came from, with the same columns: spend · impressions · clicks · …

    line 13
  2. Distinctive:

    Write the platform's own conversion definition next to its rows.

    line 21
  3. Distinctive:

    Days are the account's own days: Every platform cuts its window on its own midnights.

    line 44

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Campaign analyst: scale, pause, fix, test next

Most ad reporting answers "what happened". The only useful question is "what do I change on Monday". This is the method, and the rules that stop it producing confident nonsense.

The table comes first

Put every campaign (or ad set, or ad) in ONE table, whatever platform it came from, with the same columns: spend · impressions · clicks · CTR · CPC · conversions · CPA · ROAS · frequency. Two rules make the table honest:

  1. Recompute every ratio from its parts. Platforms disagree about whether CTR is a fraction or a percentage and whether spend is in whole currency or millionths. CPA is spend ÷ conversions, computed by you, every time.
  2. A metric the platform does not publish is UNMEASURED, never 0. Write it as unmeasured and say so in the summary. A zero in a column nobody measured is the single most common way an ad report lies.

Write the platform's own conversion definition next to its rows. "48 conversions" means something different on each platform, and a cross-platform CPA comparison is only honest when both definitions are on the page.

The data floor: the rule that separates analysis from noise

Decide the minimums BEFORE you look, state them in the answer, and hold to them. A workable set:

  • 1,000 impressions, 50 clicks and a floor of real spend before a row can be judged at all.
  • 10 conversions before you compare its CPA to anything.
  • A row with zero conversions is only a loser once it has spent about two account CPAs; a row with a few conversions, once it has spent about two account CPAs for each conversion it produced, plus one.

Everything below the floor goes in an insufficient data bucket with what it lacks and what to wait for. It is never "pause it": killing an under-delivered row is how a winner gets buried before it ever gets a chance.

The asymmetry that matters: the evidence needed to STOP spending is not the evidence needed to spend MORE. A row that has burned five account CPAs for one conversion is a measured loser at any conversion count. The same row can never be called a winner on that data.

The other four rules

  • A platform you could not read is named as unreadable. Its numbers are missing, not zero, and no budget moves away from it on the strength of an absence.
  • Learning phase. A Meta ad set still calibrating is not a verdict, and editing it restarts the learning. Leave it alone and say when to look again.
  • Days are the account's own days. Every platform cuts its window on its own midnights. Never present two platforms' windows as hour-aligned.
  • Do not pause what is already paused, and if the best row in the table is switched off, say plainly that the recommendation is to switch it back on.

What a recommendation must contain

The row, its own numbers as evidence, the reason in one or two sentences, and the concrete next move with the number in it ("raise the daily budget from 25 to 40 and re-read CPA in 3 days"). A recommendation that cites a number not in the table is the failure mode to watch for, so check each one back against the row before you send it.

What a test must contain

One variable (creative, audience, bid, budget, placement, landing page or offer), a hypothesis the numbers already support, a success metric WITH a number, a budget and a duration. "Test new creative" is not a test.

Before you send it

Re-read every number in the answer against the table. Every recommendation that moves money (a budget change, a pause, switching something back on) is a proposal: show it to the person who owns the account and get an explicit yes before anyone applies it.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.