agentleFS
minhnv0807minhnv0807/ai-business-skills/skills/en/57-next-ads-plan-global/SKILL.md

57-next-ads-plan-global

Skill644 starsChanged 6 months agoMITView on GitHub

Use when the next period ADS plan must be built from last period results — read the report and the audit, keep and scale winners, stop or fix losers, set new test hypotheses, split budget, plan three budget scenarios, and schedule the creative pipeline. Trigger on 'next month ads plan', 'plan ads from the report', 'what should we run next month', 'Q4 ads plan', 'next quarter media plan', 'the month ended, now what'. Also use when a report and an audit both exist and no decisions have been made. Not for — the first media plan before any data, see `54-media-plan-global`; scoring the account, see `21-ads-audit-global`; the next period CONTENT plan, see `40-next-content-plan-global`.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is 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.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before writing a single line of plan, you need:

    line 36
  2. Distinctive:

    Data before opinion: Never plan from a feeling that something "seemed fine".

    line 47
  3. Distinctive:

    Plan three budget scenarios: Commit to the base case; keep prepared actions for the other two.

    line 51
  4. Very specific:

    If the target changed, or the offer, price, or market changed, recalculate with 10-reverse-kpi-global before continuing.

    line 77

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

Next Ads Plan (Global)

A plan without last period's data is guesswork, and guesswork is expensive. The order never changes: read the data, scale the winners, fix or stop the losers, then test something new. This is the last skill in the performance chain and the first skill of the next cycle.

Information gathering

Step 0 — always read the previous period first (mandatory input)

Before writing a single line of plan, you need:

  1. Previous period performance report from 07-marketing-report-global or the tracking dashboard: spend, CPA, ROAS, leads, orders, broken out by channel, campaign, creative, and market.
  2. Audit findings from 21-ads-audit-global or 03-performance-eval-global if the previous period had problems.
  3. KPI target for the new period — unchanged or changed? If changed, recalculate with 10-reverse-kpi-global.
  4. New period budget and any new offer or product entering the mix.

Missing the report means stop and either request the data or run 07-marketing-report-global first. If only part is missing, write the assumption explicitly into the plan.

Principles

  1. Data before opinion. Never plan from a feeling that something "seemed fine".
  2. Scale proven winners before funding new ideas. Capital goes first to what is already working; new tests get the 30% testing slice.
  3. Maximum 2-3 new hypotheses per period. More than that and no test reaches significance.
  4. Every keep/stop/fix decision carries a number. "Creative A at $42 CPA against a $30 target, frequency 3.1" — never "it was performing badly".
  5. Plan three budget scenarios. Commit to the base case; keep prepared actions for the other two.
  6. Adjust for the calendar. Per references/benchmarks-global.md, Q4 CPM runs about 26% above Q1, November peaks around 27% above average, and January is the cheapest month at roughly 21% below. A flat month-over-month plan across a seasonal boundary is already wrong.

Workflow

1. Review the previous period

Worked (KEEP)Did not work (STOP/FIX)Untested (ADD)
[winning creative/audience/channel with numbers][losing ad set with numbers and cause][new hypothesis from observation]

Extract the winning patterns:

  • Creative type: video, static, carousel, UGC — which produced the best CPA?
  • Hook type: pain, curiosity, social proof, demonstration — which produced the best CTR?
  • Audience: interest, broad, lookalike, Advantage+ — which was cheapest and highest quality?
  • Channel: which delivered the best ROAS, and which dragged the blended average down?
  • Market: compare each market against its own tier baseline in references/benchmarks-global.md, never against the blended account average.
  • Operational lessons: best posting times, creative refresh cadence, setup mistakes to avoid.

2. Direction for the new period

PeriodTotal budgetObjectiveKPI targetNew offer or campaign
[month/quarter][amount][lead/conversion]CPA [x] / ROAS [x] / volume [x][if any]

If the target changed, or the offer, price, or market changed, recalculate with 10-reverse-kpi-global before continuing.

3. Multiply the winners

WinnerPrevious CPA/ROASHow it multipliesWhat changesNew budgetChannel
[Creative A][numbers]Keep as is, +20-30% budgetNothing[amount][x]
[Audience B][numbers]Duplicate with new creative2 new variants[amount][x]
[Hook C][numbers]Port to a new channelFormat change[amount][x]
[Market D][numbers]Expand to an adjacent marketNew geo, same creative[amount][x]

Apply the 55-scaling-ads-global rules: +20-30% per step, never double, do not edit the winning ad set.

4. Stop or fix the losers

Campaign / ad setProblem (with numbers)DecisionIf fixing: what changes
[name]CPA above 3.0x target with spend >= 3x target CPA (house heuristic ladder, 03-performance-eval-global)STOP (propose pause via rule 7)—
[name]CPA above 1.5x target for 7 daysFIXChange creative or audience
[name]CTR far below the median bandFIXNew creative and hook, keep the audience
[name]Frequency above 2.5 (heuristic), CTR fallingFIXRefresh creative, widen the audience
[name]Good CTR, poor CVRFIXLanding page or offer, not the ad

Use the diagnostic order from references/benchmarks-global.md: weak CTR is a creative problem; good CTR with weak CVR is a landing page or offer problem; good CVR with weak ROAS is a margin or AOV problem. Every stopped item gets a logged reason so nobody retests the same thing next period.

5. Budget split for the new period

CampaignPurposeDaily budget% of total
Scale winnersMaintain and grow[amount]50%
Testing (new hypotheses)Find the next winner[amount]30%
RetargetingConvert warm and hot[amount]15%
LookalikeOpen new pools from fresh seeds[amount]5%
Total100%

6. Three budget scenarios

ScenarioAssumptionBudgetExpected volumeAction if it happens
DownsideCPA up 30% (seasonal peak, creative fatigue, auction pressure)[amount][number]Cut testing to 20%, concentrate on winners and retargeting
BaseCPA holds at last period[amount][number]Run this plan as written
UpsideCPA down 20% (a new winner lands)[amount][number]Push scaling +20% per 72h, open a new channel or market

If the new period crosses into Q4, build the downside scenario as the default assumption rather than the exception — the seasonal CPM lift is documented, not speculative. Commit to the base scenario with stakeholders; keep prepared actions for the other two so nobody is caught flat-footed.

7. New hypotheses (maximum 2-3)

#Hypothesis ("If X then Y because Z")VariablePlatformMetricTimeline
1If we use UGC testimonial creative, CPA drops because trust is higherCreative format[x]CPA7 days
2If we open a 3% lookalike, volume rises while CPA holdsAudience[x]CPA, volume7 days
3If we test a second market, blended CPA improvesGeo[x]CPA by market14 days

Set up and conclude these according to 19-ab-test-setup-global: one variable at a time, at least 7 days, enough conversions to decide.

8. Creative pipeline

Creative neededTypeHookChannelDeadlineUsed for
[name]VideoPainMeta[date]Testing
[name]StaticSocial proofMeta[date]MOFU / retargeting
[name]UGCTestimonialTikTok[date]Scaling

Have 2-3 reserve creatives ready before the period starts. Do not wait for fatigue to brief.

9. Weekly schedule and acceptance thresholds

WeekFocusActionCumulative % of period KPI
1Launch winner clonesBuild, launch, verify tracking green20%
2Launch testsRun hypotheses, read day-3 signal45%
3OptimizeScale winners, propose pausing losers per the house ladder (rule 7)70%
4Review and preparePull data, write the report, draft the next plan100%

At the end of each week, compare cumulative actual against the threshold. A shortfall above 15% triggers the downside scenario immediately — do not wait for the period to end.

Output structure

File name: next-ads-plan-[product]-[YYYYMMDD].md — nine sections matching the nine steps above: review, direction, multiply winners, stop losers, budget split, three scenarios, hypotheses, creative pipeline, weekly schedule.

Related skills

  • 07-marketing-report-global: the primary data source — run it first.
  • 21-ads-audit-global and 03-performance-eval-global: audit before planning if the previous period had problems.
  • 10-reverse-kpi-global: recalculate whenever the target, budget, offer, or market changes.
  • 54-media-plan-global: when the new period involves a major change (new channel, new offer, new market), build a full media plan instead of an increment.
  • 19-ab-test-setup-global: how to set up and conclude each hypothesis.
  • 52-account-structure-global: build and launch the campaigns this plan defines.

Quality checklist

  • Previous period report and audit were read — the plan is not written from impression
  • Review table has all three columns (keep / stop-fix / untested) with numbers attached
  • Market performance compared against its own tier baseline, not the blended average
  • Winners are multiplied first, following the +20-30% scaling rule
  • Every stopped item has a logged reason
  • Budget split follows 30/50/15/5 and totals 100%
  • Three scenarios present, each with a prepared action
  • Seasonality applied where the period crosses a Q4 or January boundary
  • Maximum 2-3 hypotheses, each with a variable, metric, and timeline
  • Creative pipeline has deadlines and 2-3 reserve creatives
  • Weekly schedule has cumulative acceptance thresholds of 20/45/70/100%

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.