Michael Mignogna

My Business Brain / Job 7 of 9

Optimization

Check every ad against the target cost per lead three times a week. Keep the winners, pause losing ads at the kill line, and refill the campaign.

For contractors and agency owners11 principles7 playbook steps

Optimization is a short check on a fixed schedule, run the same way every time. Each ad is judged on its own cost per lead against the target. The kill lines below are multiples of the target, so they work for any offer and any budget.

Every job also follows the 12 ground rules.

How to use this page: read the principles, or copy the starter prompt into ChatGPT or Claude. Other ways

Principles

  1. Judge each ad by its own cost per lead against the target.

    Why
    The campaign average hides a winner and a losing ad that cancel each other out.
    How to apply
    Pull spend and leads for every ad for the last seven days, including today. Divide spend by leads for each ad and compare it to the target. Compare ads only with ads for the same offer, because price report leads usually cost less than starting-at leads.

    GoodOne ad gets leads at half the target and one at three times the target. The first stays on and the second is paused.

    BadThe campaign is at target on average, so nobody notices one ad is spending most of the budget for one lead.

  2. Never pause an ad that is getting leads under the target cost per lead.

    Why
    An ad under target is doing its job. Pausing it because it is old, or because a new ad looks exciting, throws away leads you know you can get.
    How to apply
    Leave every ad under target on, no matter how long it has run. Add new ads beside it instead.

    GoodAn ad has run for four months at 80 percent of target. It stays on, and two new ads launch next to it.

    BadThe best ad is paused because it has been running a long time and someone wanted fresh ads.

  3. Pause losing ads at a kill line set as a multiple of the target.

    Why
    Without a fixed line, people pause ads on a feeling. They kill ads before they had a chance, or let losing ads spend for weeks. A multiple of the target works for every offer.
    How to apply
    Do not judge an ad until it has spent one and a half times the target cost per lead. Pause an ad with zero leads once it has spent two and a half times the target. Pause an ad with leads once its cost per lead is above two and a half times the target. When other ads in the same campaign already get leads under target, you may pause an ad whose cost per lead is above one and a half times the target.

    GoodThe target is 100 dollars. An ad with no leads is left alone at 140 dollars of spend and paused at 250. An ad with two leads on 600 dollars of spend costs 300 a lead, so it is paused.

    BadAn ad is paused after one day and 40 dollars of spend because it had no leads yet.

  4. Keep the only ad that is getting leads on, even when it is expensive.

    Why
    Pausing it leaves the campaign with nothing working while new ads learn.
    How to apply
    Leave it on and add new ads beside it. Pause it only after a new ad beats it.

    GoodA new campaign's only lead-getting ad runs at twice the target. It stays on while three new ads launch beside it.

    BadThe only ad with leads is paused, and the campaign gets no leads for a week.

  5. Do not react to daily swings.

    Why
    A local campaign may get only a few leads a day, so one day can swing 20 or 30 percent by chance. Chasing yesterday's numbers kills good ads.
    How to apply
    Check three times a week on a fixed schedule. Judge on the last seven days, not on yesterday.

    GoodCost per lead doubled yesterday, but the seven-day number is at target, so nothing changes.

    BadBudgets are cut every morning the cost per lead goes up.

  6. Refill empty slots the same day.

    Why
    A campaign with too few live ads stops finding new winners, and its cost per lead climbs.
    How to apply
    After pausing, count the live ads. Keep as many as three days of budget divided by the target cost per lead, rounded down. Fill the gap with a rested winner or with new approved ads.

    GoodTwo ads are paused in the morning check, and two approved ads go live in the same check.

    BadThree ads are paused, and the campaign runs on one ad for two weeks.

  7. Rest winners, then bring them back with their own copy.

    Why
    A winning ad can tire as the same people see it again. After a rest it often wins again. The image and the copy won together, so they come back together.
    How to apply
    Wait at least seven days before you turn a paused winner back on. Check that its price and offer still match the offer sheet. Never revive an ad that was paused for losing.

    GoodA winner rested for two weeks comes back with its original headline and primary text, and the starting price it names is still right.

    BadA winning image is paired with another ad's headline, and an ad that lost badly is turned back on to fill a slot.

  8. Use cost per lead as the scoreboard, and say which number you used.

    Why
    Booking data is often late or missing. A made-up cost per appointment leads to wrong decisions.
    How to apply
    Use cost per booked appointment only when you can see real bookings. Count first conversions, not all conversions, so one person is not counted twice.

    GoodThe report says cost per lead, because bookings were not in the CRM this week.

    BadThe report guesses that half the leads booked and shows a cost per appointment nobody measured.

  9. Stop and report a campaign that spends a week with zero leads.

    Why
    Zero leads after real spend usually means a broken form, broken tracking, or a wrong offer. New ads will not fix it.
    How to apply
    Do not refill it. Tell the owner, then check the landing page, the form, the tracking, and the offer.

    GoodA campaign spends its full budget for a week with no leads. The check finds the form stopped sending after a page edit.

    BadThe campaign gets five new ads every check for two weeks while the form is broken.

  10. Pause the ads and the calendar together around time off.

    Why
    Leads that come in while nobody can call or meet them go cold, and booked appointments get cancelled.
    How to apply
    Get the dates in writing. Pause the ads two days before the owner leaves, turn them back on one day before they return, and block the calendar for every day away. Do both, or do neither.

    GoodThe owner is away from the 10th to the 17th. Ads pause on the 8th and restart on the 16th, and the calendar is blocked from the 10th to the 17th.

    BadThe ads are paused, but the calendar stays open, and homeowners book days nobody can come.

  11. Update ads that name a season or a year.

    Why
    An ad that names last year or a season that ended looks out of date and loses trust.
    How to apply
    Turn off seasonal sale ads when the season ends. Keep ads that name a season only as a problem, such as winter drafts through old windows. In the first week of a new year, pause every ad that names last year, and replace the winners with new versions the same week.

    GoodOn the first workday of January, ads that name last year are paused, and new versions of the winners launch that week.

    BadA spring sale ad is still running in July.

Agent playbook

Steps an AI agent can run in order. Each step names what it needs and what it hands to the next step.

  1. Pull the numbers.
    Inputs
    Spend and leads per ad for the last seven days, including today, and the target cost per lead.
    Output
    A table with one row per ad: spend, leads, and cost per lead.
  2. Mark each ad.
    Inputs
    The table and the kill lines.
    Output
    Each ad marked keep, too early, or pause. Too early means it has spent less than one and a half times the target. Pause means zero leads after two and a half times the target, or a cost per lead above two and a half times the target. An ad above one and a half times the target may also be paused when other ads in the campaign get leads under target.
  3. Apply the exceptions.
    Inputs
    The marked table and any holds the owner set.
    Output
    The final list of ads to pause. The only lead-getting ad in a campaign stays on, and every hold is respected.
  4. Count the empty slots.
    Inputs
    The daily budget, the target cost per lead, and the ads still live.
    Output
    Three times the daily budget, divided by the target, rounded down, minus the live ads.
  5. Refill the slots.
    Inputs
    Rested winners and approved ads waiting to launch.
    Output
    A list of ads to turn on, with a rested winner first when one is ready.
  6. Brief new ads if nothing is ready.
    Inputs
    What won, what lost, and the offer sheet.
    Output
    A short brief for new ads that names the winning ideas to build on.
  7. Write the report.
    Inputs
    Everything above.
    Output
    A short summary in plain words: what was paused and why, what was added, and what is waiting.

Final output

  • A table of every ad with spend, leads, cost per lead, and the decision
  • Each paused ad and the kill line it crossed
  • The ads added
  • Slots still empty
  • A plain-language summary
Starter prompt

Read https://byminyona.com/brain/playbook.md first and follow its Optimization rules and its ground rules. Review these ad results and tell me which ads to pause using the kill lines. My target cost per lead is <target cost per lead> and my daily budget is <daily budget>. Results: <paste the last 7 days of spend and leads per ad>

If your AI cannot open links, paste this page's principles into the chat with the prompt. You can also paste the whole playbook or download AGENTS.md.

Next job: Lead Follow-up and Booking