Our Daily Meta Ads SOP: The 6-Step Checklist We Run on Every Client Account

A while back I watched a brand torch a good account in about four days, and the whole thing came down to one bad habit: they managed off today's numbers.

Here's what happened. They'd scaled a campaign up to roughly A$3k/day and it was working. Then one Tuesday the cost per acquisition spiked, mid-afternoon, on maybe six purchases of data. They panicked and cut the budget back to A$800. Wednesday looked worse, so they cut again. By Friday they'd nuked a winning campaign down to nothing, sat there confused about why performance had "fallen apart", and started blaming the creative.

The creative was fine. The account was fine. The problem was that nobody had a system. Every decision was a fresh emotional reaction to whatever the dashboard happened to say in that moment.

So here's my take, and it's the thing this whole post is about: disciplined account management is boring on purpose. It's the same short checklist, run in the same order, every single day, so that no single bad afternoon ever gets to make a decision. This is the exact six-step routine we run on every client account. It takes about 15 to 20 minutes per account once it's a habit.

1. Record yesterday's numbers before you touch anything

First thing, before you open Ads Manager and before you have an opinion, you write down what already happened.

We log it in a plain spreadsheet. Nothing fancy. Total spend, total revenue, blended MER, new customer revenue, new customer orders, and the blended CPA, for yesterday and for the trailing 3, 7 and 30 days. That's it.

The reason this goes first is psychological as much as practical. If you open the ad account cold, the first big red number you see sets your mood for the whole session, and you start managing to feel better rather than managing to the data. Writing the actuals down first forces you to look at the trend, not the twitch. A single day at a 1.4 MER means nothing if the 7-day sits at 2.6. The log is what lets you see that.

2. Read the metrics in a fixed order, on the right window

Now you open the account. But you read it in a set order, every time, so you're not just chasing whatever's flashing.

The order we use: spend first (where's the money actually going), then purchases and cost per purchase, then ROAS, then the early signals like click-through rate and cost per click, then frequency to check we're not just hammering the same people.

Two rules that matter more than the order itself.

• Read on a 7-day click, 1-day view attribution window, not the default. Meta is very good at sliding in front of someone right before they were going to buy anyway and claiming a view-through. Optimising to click data keeps you honest about which campaigns are genuinely doing the work.
• Never make a call on today's figures. Today is half-baked and statistically meaningless until it closes. Every decision below is based on yesterday and the trailing windows. If a number scares you at 2pm, you write it down and you wait.

3. Set budgets by the rule, not by the mood

This is where the brand at the top went wrong, so it's worth being precise about the actual rules we follow.

For each scaling campaign we ask two questions in sequence.

Was yesterday's new-customer CPA at target or below? If yes, increase the budget by about 20%. Not double, not "send it", just 20%. You're nudging, not yanking.

If it was above target, you don't immediately cut. You ask: has it been at least three full days since the last change? If it hasn't, you leave it alone and let it settle, because we just changed it and the algorithm needs room to optimise to the new level. Only if it's been three-plus days of genuine underperformance do you scale down by 20%.

And we hold a hard floor on every account, a spend level we won't drop below regardless of a rough patch, because dropping to almost nothing makes it impossible to keep testing your way back out. So if a campaign that scaled to A$3k/day has a bad week, we might ease it back toward A$1k and hold there, rather than slamming it to A$100 and killing the account's ability to learn.

The point of all of this: 20% steps, a three-day patience rule, and a hard floor. That alone would have saved the brand I opened with.

4. Launch the new creative that's ready to go

With budgets handled, you check whether there's anything new to put live.

We run each account off a creative pipeline, so on any given day there's a clear "ready to launch" queue or there isn't. If there is, the buyer's job here is genuinely just to build a new test and upload it. We launch new concepts as fresh tests and then leave them alone to run for about 7 days, because that's roughly what it takes to see whether a concept can pull real spend against the incumbents.

Some days there's nothing to launch. That's completely fine. A surprising share of days, four or five out of seven on a stable account, you check the queue, see it's empty, and move on. Launching for the sake of launching is not discipline, it's fidgeting.

5. Prune the losers, protect the winners

Now the careful part. Turning things off.

The rule for new tests is simple. If a concept has run its ~7 days and made basically no dent, you cut it. I mean a real test, not a tiny one. If a new ad has spent A$300 against a campaign spending A$15k, it has a rounding-error effect on the business whatever its CPA reads, so you stop it and move on. Be honest that most concepts lose. We win on something like 1 in 5 of the ideas we test, and the whole model only works because the winners win big enough to carry the misses.

The bit people get wrong is pruning the winners. The instinct, when a high-spending ad's ROAS dips, is to pause it. Resist that. Pausing your top one or two ads forces Meta to suddenly redistribute thousands of dollars a day across weaker stuff, which can wobble the entire campaign far more than the dip you were worried about. The one fair reason to pause a high-spender is a clear structural signal, say its click-through rate is way out of line with the account and it's spending hard while not converting. Short of that, you leave the engine running.

6. Log the result and send a one-line client update

Last step, and it's the one most people skip because the day already feels done.

When you turn an ad off, you record it as a winner or a loser in your creative tracker. That log is gold over time. It's how you spot that, say, your "founder talking to camera" concepts have a much higher hit rate than your fancy edited ones, which is the kind of pattern that actually improves the next round of briefs.

Then you tell the client what changed, in a line or two. Raised or dropped a budget, launched two new tests, found a new winner. No essay. The point is a clean trail of what you did and why, so a bad week is a story you can read back, not a mystery.

Why the order is the whole point

None of these six steps is clever on its own. Record, read, budget, launch, prune, log. The discipline is doing them in that sequence, every day, so you're acting on yesterday's closed data instead of reacting to today's noise. The brand I opened with had all the right instincts and no system, and the system is the entire job.

If you've got the discipline but you're not sure your account is set up to reward it, that's exactly what our Signal/Noise Audit is for. We go through your account structure, your creative history and your unit economics and show you where the routine's quietly leaking money, before you change a single budget. Want me to take a look at yours?