You get $100 a day and five edits a week for 28 days. I would rather launch under that constraint than follow another schedule that says to wait two weeks and then start optimizing.

I used to give clients that advice. I was wrong. Waiting does not protect a new account; an edit budget does. I have watched accounts suffer from too many well-meant changes, including a SaaS client that changed its offer every quarter while I tried to keep Smart Bidding pointed at a target that no longer existed. Each reset cost us three weeks. This is the 28-day plan I would run instead: make a change only when a number earns it, and leave the rest alone.

The rules: $2,800 maximum, five saves per week

At $100 a day, the 28-day spend cap is $2,800. You have five edits in each seven-day launch week. An edit is anything you change inside the account after launch: a bid, budget, keyword, ad, negative, or conversion goal. Opening the Search Terms report costs nothing. Debating a change costs nothing. Clicking save costs one edit. Unused edits expire at the end of the week.

Five is deliberately restrictive. Smart Bidding can take a week up to a month to calibrate, with volatility during exploration. Too many changes at once obscure what worked and can send a strategy back into learning. Five edits give you room to fix real problems, but not enough room to rearrange the account every time a chart looks ugly.

Here is the order. Each step has a gate, a cost against the limit, and a job to do. No gate, no edit.

1. Day 0: verify the signal before you spend

Cost: zero weekly edits. Buys: a conversion signal worth trusting. Nothing is live yet, so do the setup work before the edit counter starts.

  1. Verify conversion tracking the boring way: use Tag Assistant, complete a test conversion as a customer would, and confirm the Ads tag fires on the right trigger. Check that it does not fire on the wrong pages or on a page reload. If a button click counts as a lead every time someone presses it, Smart Bidding can spend your $100 a day learning to buy button clicks. I have watched that go unnoticed for weeks.
  2. Freeze the landing page. Tell everyone who can change a headline, form, or URL to leave it alone through day 28 unless something is broken. A page swap changes conversion rate while you are trying to learn whether the campaign works.
  3. Seed 30 to 50 obvious negatives: jobs, free, DIY, locations you do not serve, and competitor names you have decided not to pursue. This is not the final list. It keeps predictable waste out of the auction while you collect the first search terms.

The gate is absolute: no verified conversion, no launch. Delay rather than spend the first week teaching Google the wrong definition of success. What gets cut is the urge to launch on schedule and sort out tracking later.

2. Week 1: buy evidence, not efficiency

Cost: zero to two of five edits. Buys: the first clean read on queries and conversions. Launch one Search campaign at $100 a day with a tight set of exact and phrase keywords. Use Maximize Conversions without a target. That gives the strategy room to seek verified conversions within the budget while it builds predictions from the auctions it sees.

Calibration depends on data, not on the date in your launch plan. The rough guide is up to 50 conversion events or three conversion cycles, with usable history potentially speeding things up. At $100 a day, you might have only 8 conversions after week one. You may still be in learning when week two begins. That is not, by itself, a reason to intervene.

Keep your two available edits for these gates, in this order:

  1. Tracking is broken: correct it. A campaign optimizing toward the wrong action does not deserve another dollar of undisturbed learning.
  2. A query is plainly irrelevant and spending: add the negative that blocks it. Do not spend an edit because an expensive but relevant query has not converted after a handful of clicks.

If neither gate opens, spend zero edits. Do not change the target, budget, keyword set, and ad copy to make the first weekly report look tidier. Strategy, budget, and structural changes can disrupt stabilization. Small accounts have less data to absorb that disruption: at 10 leads a month, reaching 50 conversions can take around five months; at 25 a month, around two.

Week one buys a trustworthy baseline. Clean tracking and controlled waste count as progress even if the CPA graph is not yet fit for a slide deck.

3. Week 2: spend edits on waste you can name

Cost: plan for three of five edits; leave two unused. Buys: less spend on demonstrably wrong searches. Start with search term mining: open the Search Terms report, sort by cost, and inspect the expensive terms first. Looking is free. Saving a change is not.

Before you touch a term, calculate the most you can afford per conversion. Say a job produces $400 in gross profit and you need half of that left after ads. Your break-even ad cost is $200 per conversion. If week one brought 8 conversions at $87.50 each, do not tear up the campaign. Block proven irrelevant traffic and keep gathering data from the intent that worked. If those conversions cost $260 each, do not respond by adding keywords or pushing bids. Find the waste first.

Use the edits in this order:

  1. First edit: block the highest-cost theme that has no chance of converting, such as job-seeker queries for a customer campaign.
  2. Second edit: block the next distinct, irrelevant theme if the report shows one. Do not manufacture a second theme to fill the slot.
  3. Third edit: pause a keyword only if its spend and search terms show that it is bringing in the wrong intent. Otherwise, use another warranted negative or save the edit.

Count each negative theme or keyword pause as an edit. Do not turn this into a full-list rewrite because you found three bad queries. The point is to remove the clearest leaks without changing so much that next week’s results become impossible to read. An unused edit beats an unearned edit.

Know which changes carry the bigger cost

The five-edit limit counts saves, but not every save disrupts learning equally. A new or reactivated bid strategy, a budget or tCPA/tROAS target change, and adding or removing campaigns, ad groups, or keywords can push a campaign back toward stabilization. Changing which conversions count as primary may look like housekeeping. To the bidder, it changes the job. That distinction also matters across Search, Shopping, and Performance Max.

I once moved a SaaS client from counting demo bookings to counting qualified demos in week three. The reason was defensible. The bidder still had to relearn the conversion signal, and we saw 19 days of volatility. I would make that correction again if the old goal was wrong, but I would not pretend it was a harmless tidy-up.

Reading reports and checking learning status cost no edits. Relevant negative keywords are narrower interventions than changing the campaign’s bidding goal. That is why week two concentrates on them. If you want to change the conversion goal, budget, and keyword set in the same week, choose the correction tied to the worst number. Leave the rest alone.

4. Week 3: earn a target CPA or stay put

Cost: one edit if the tCPA gate opens; otherwise, at most one warranted waste cut. Buys: a bid constraint backed by conversion data, or another week of usable evidence. Week three is when a target starts to feel overdue. Feeling overdue is not a metric.

My gate is 30 conversions in the last 30 days before switching from Maximize Conversions to tCPA, with 50 giving more confidence. Maximize Conversions uses the $100 budget to seek verified conversion volume. A tCPA asks the strategy to be more selective. Set a tight target with too little data and it may pass over auctions that could have taught it where buyers are. Spend can dry up before you have learned much.

Say day 21 shows $2,100 spent on 14 conversions at a $150 average CPA. No 30, no target. Keep Maximize Conversions. If the search terms justify one negative, spend one edit there; otherwise, bank all five.

If you have at least 30 conversions and the recent results are clean, the switch costs one edit. Set the initial tCPA 5 to 10% above the observed average, run it for a full week, then consider stepping it down. Do not set $55 merely because finance wants $55 when the account has been producing conversions at a $75 average. That target can shut out the auctions behind the volume you already have.

I made that mistake in 2019. A healthy brand campaign went from 11 conversions a week to three after I set a target the data could not support. Same keywords. Same ads. Keep the daily budget at $100 during the switch, too. Changing bid logic and budget together uses two edits and leaves you with an unreadable test. The other four edits stay untouched unless something breaks.

5. Week 4: cut losers only when they cross a gate

Cost: up to five edits. Buys: fewer moving parts and a cleaner account for the next decision. By now you have 21 days and roughly $2,100 of spend behind you if the campaign has used its daily budget. That is enough to inspect specific losers, not a license to rebuild everything.

Work down this kill list. Stop when you run out of qualifying candidates or edits:

  1. Pause an ad group that spent more than 1.5 times your break-even CPA with zero conversions. At a $200 break-even, more than $300 with nothing back meets the gate. Cost: one edit per pause.
  2. Pause a keyword that spent a full break-even CPA with zero conversions and more than 30 clicks. Cost: one edit per pause.
  3. Consolidate overlapping intent if two ad groups are splitting the same work into thin pools of data. Cost: two edits for the merge. Do it only if the overlap is clear and you have two edits left.

I would cut the clever long-tail buildout people want in week four, the second campaign for a new offer angle, and the ad rewrite that fixes neither a disapproval nor a broken landing-page promise. Those are changes you cannot judge cleanly yet. If the kill list warrants three edits, spend three, not five.

What survives should be boring on purpose: one campaign, three to six winners, a negative list you have built from actual queries, and a CPA average you can interpret because the conversion definition stayed put. If you need to separate ordinary volatility from a setup mistake, this breakdown of the learning phase is worth reading before you spend another edit. Day 28 should leave you with fewer moving parts than day one.

6. Day 29: lift the spend cap, not the discipline

Cost against the 28-day constraint: none; the four weeks are over. Buys: a controlled path to more volume. If day 28 shows a CPA below break-even and at least 30 conversions banked, lift the $100 daily cap one step at a time. Keep the five-edit discipline for decisions you make yourself.

  • Raise daily spend 20%, hold for seven days, then check whether average CPA stayed within 10% of the prior week before raising it again.
  • Add one new intent cluster per fortnight, funded from the loser spend you cut in week four rather than from new money.
  • If you have set tCPA, step it down 10% every two weeks only while volume holds and impression share lost to rank stays under 30%.

The order matters. First establish that the existing intent still converts as spend rises. Then test another cluster. Do not raise the budget, add an offer, and tighten the target on the same Tuesday and call the result a lesson. One raise and one addition per fortnight use two of your old five edits. Leave the others available for problems with a number attached.

Paper budget ledger with five paper clips beside a coffee mug

Who should skip this plan

Skip it if you close two deals a month and need both to be perfect. At $100 a day, you will not bank 30 conversions in 30 days, so the tCPA gate will not open. That business needs tighter geo, a call-only structure, and manual review of every click rather than this calibration plan.

Skip it if your catalog runs through Shopping or Performance Max with 500 SKUs. That is a feed problem first and a bid problem second. Skip it if you already spend $20k a month and have account history. You are restructuring, not launching; start with an audit, not day zero.

When the constraint lifts, do not go back to daily human tweaks. Give the routine work to a system that can act within guardrails without treating every bad Tuesday as a reason to change direction. That is why I now run launches inside groas: its models handle search-term cuts, bid pacing, and budget shifts continuously, while a named strategist owns the goals and the calls that change them.

My five-edit rule was a way to keep my hands off changes the evidence had not earned. Keep that standard after day 28, even when the work no longer depends on you clicking save.