Why Some Meta Ad Accounts Scale and Others Stall

Here is the short answer. Meta ad accounts that scale do three things stalled accounts do not. They test in a way that shows which ads genuinely lost, they vary the idea inside each ad rather than just the wording, and they put more than one face in front of the camera. Most of how to scale Meta ads for a lead-gen business comes back to those three habits. What does not carry over from the big accounts is the volume. A New Zealand business spending a few thousand dollars a month cannot run hundreds of ads, and should not try.

The list we are testing comes from Daniel Iles, who says he spends around a million dollars a month on Meta for his own business. In a recent video he names the mistakes he sees separating stuck accounts from ones that scale. It is a sharp list. It is also written from a budget most NZ service businesses will never see, so below we take the ad-account half of it and check each point against what our own published campaigns show. Several of his points are really about the business behind the ads, and our sister post on Lucid Media covers that side: why cheaper leads can break the sales process behind them.

1. Test so you can actually see a loser

Verdict: holds, but use a cheaper method than his.

When several ads share one budget, Meta decides which of them gets the money, and it often decides fast. One ad takes most of the spend, the rest get scraps. Campaign budget optimisation (CBO) does the same thing one level up, moving budget between ad sets. Iles’s complaint is that this leaves you with ads you cannot judge. An ad that got a sliver of spend did not fail. It was never tried.

His fix is one ad per ad set, each with its own budget, so every ad gets a fair run. He says moving his clients to that structure has saved them 30 to 50 percent. That is his figure from his client base, and we have no way to check it. The principle underneath it is right, though, and it matters more on a small budget than a large one, because a small budget can only afford a few tests.

Here is what it looks like in a real account. On a car detailing campaign we ran on Meta, about 69 percent of the budget and 71 percent of the leads went to one ad, the one that led with a “from” starting price. A before-and-after side panel ad added another 20 percent of leads. Those two ads brought in about nine in ten leads.

That tells you the price ad worked. Whether the other angles failed is a separate question. The social proof and same-day convenience ads shared what was left of the budget with the before-and-after ad, so some of them may never have had a fair run. The numbers cannot settle it either way, and that is exactly the problem Iles is describing.

Why strict one-ad-per-ad-set rarely fits an NZ budget

Meta’s own help centre says an ad set usually exits the learning phase after about 50 results in the week after its last significant edit, and the same page advises against creating many ads and ad sets because the system learns less about each one. At Iles’s spend, every single-ad ad set can clear that bar. At a budget that produces ten or twenty leads a week, eight single-ad ad sets means eight ad sets stuck in learning, none of them telling you much.

We covered the budget maths in detail in how many creatives to test at your daily spend. The short version: your testing volume is capped by your conversion volume.

What we run instead

• One main ad set carrying your proven ads. Broad targeting, most of the budget. For many accounts this is the only ad set that needs to exist.
• A small test ad set beside it, in the same campaign. Two or three new ads, each differing from the others on one thing. A vehicle finance campaign we launched runs this way: one campaign with a broad ad set, a regional ad set and a separate ad set for testing new creative angles.
• Keep known winners out of the test set. If a proven ad sits in the same ad set as the new ones, Meta will route the money to the ad it already trusts, and the new ones go untested.
• Sort every ad into winner, loser or untested. Our working rule: an ad that has spent two to three times your target cost per lead without producing one has had a fair go and has lost. An ad that has spent less than your target cost per lead has not been tested at all, whatever the dashboard colour says. Comparing cost per lead between two ads that are both producing is a different job, and needs more leads before the difference means anything.
• Give an untested ad you still believe in one more run on its own. This is where Iles’s single-ad ad set earns its place: as a targeted tool for an ad that keeps getting starved, not the default structure.
• When a test ad wins, leave it where it is. Raise the budget on the ad set it won in, then open a fresh test set beside it. We explain why in the post on cutting campaigns back: moving a winner to a new campaign leaves behind the learning that made it win.

2. Variety is the targeting now

Verdict: holds, and it is the most useful point for small budgets.

Iles argues that Meta reads the ad itself (the words said, the text on screen, the visuals) to decide who should see it, so a vague ad gets shown to everyone and converts almost no one. That fits how Meta’s delivery now works, which we unpacked in our guide to setting up campaigns for Andromeda. Interest targeting matters less. What the ad says matters more.

The practical consequence is that ten versions of the same ad with slightly different hooks is not variety. Meta is likely to treat them as one idea and show them to much the same people. Real variety comes from changing one of four things at a time:

1. Who it speaks to. Name the customer. An ad that calls out rental property owners and an ad that calls out first-home buyers are likely to reach different people, even with identical targeting settings.
2. Who is on screen. A different face reaches a different viewer. Section 3 covers this.
3. The outcome it promises. More bookings, fewer no-shows and better-quality jobs can all be true of the same service, and each one lands with a different buyer.
4. How the offer is packaged. A starting price, a free assessment, a guide or a direct booking are four different ads for the same service.

The fourth lever is the one most small accounts never touch, and it is where our clearest evidence sits. For an accounting firm, we ran four campaign variants side by side: a free guide through an instant form, a guide through a multi-step funnel, a website enquiry form, and a direct consultation booking. The free guide through an instant form produced the cheapest leads, 76 percent cheaper than the campaign that asked people to book a consultation. The consultation campaign still produced 70 percent of all leads, five times as many as the next-best variant. The service stayed the same. What changed was how it was offered and captured, and the cheapest leads did not come from the variant that won.

The car detailing campaign above is the same lever from the other direction: putting a “from” price in the ad is a packaging decision. That ad carried about 69 percent of the spend and brought in 71 percent of the leads, so Meta backed it and it held up.

Run the winning idea through every format

Once an idea wins, most accounts go hunting for a new one. Iles’s approach is to keep the idea and rebuild it in every format: talking-head video, short looping text over footage, carousel, static image. Different people consume different formats, so the same proven message reaches more of them.

For a small account, the order matters. A plain static card is the cheapest way to find which idea works, which is why we recommend starting with the 3-line static ad. Once one idea has earned its budget, that is when it is worth paying to turn it into video.

3. More faces, done at small-business scale

Verdict: holds, and it costs less than he makes it look.

If the owner is the only person in your ads, your reach is capped at the people who respond to the owner. Iles puts it simply: “Different people trust different people.” He says that when he added female UGC creators reading scripts he had already proven, his costs dropped almost 30 percent, and that he then had more creators read the same winning script. Again, those are his results on his account, not something we can verify.

An NZ lead-gen business can get most of this from two or three different people reading a script that already works. Three practical sources:

• Your own staff. The technician, the receptionist, the second-year apprentice. They know the work and they sound like the people your customers deal with. A phone on a tripod is enough.
• Real customers, with written permission. A thirty-second phone video of a customer saying what the job was and why they called is often the most believable ad in the account. Get the permission in writing before it runs.
• Local creators. Pay one or two creators for a few takes of your proven script. If they agree, Meta’s partnership ads let the ad run under their name as well as yours.

Two rules keep this clean. Only put new faces on scripts that have already won, so the face is the one thing that changed and you can read the result. And match the face to the customer the ad calls out, using lever 1 from the last section: the person on screen should look like someone the viewer would recognise as one of their own. If you would rather not make these in-house, this is the work our ad creative service does.

4. What does not carry over when you scale Meta ads in NZ

Verdict: keep the principles, cut the volumes down to size.

This is our opinion, formed from running Meta for NZ service businesses rather than from a data set, so read it as a view.

A thousand active ads. Iles says he scaled to over a thousand active ads. That volume sits on top of the conversion volume his spend produces. An NZ account producing a few dozen leads a week has nowhere near enough results to judge a thousand ads, or a hundred. Run fewer ads with more real variety between them.

Distrusting Meta’s delivery entirely. Iles is openly scathing about Meta’s automation. At his budget, overruling it by hand is affordable. At yours, Meta’s delivery is doing most of the work, and fighting it everywhere splits your data until nothing learns. Overrule it in one place, the test set, and let it run the rest.

A big-country audience. New Zealand is small, and a local service area is smaller. We would expect the same people to see your ads more often here, so fatigue should arrive sooner. That is an argument for the variety in section 2 mattering more in NZ, not less, even while the ad count stays low.

Judging fast. Big accounts get enough results to call a test in days. Small accounts get fooled by early numbers. On the vehicle finance campaign, 27 percent of the first month’s enquiries arrived in the first three days, at a much lower cost than the rest of the month, and the cost rose from there. Judged on its first few days, the account would have looked far cheaper than it really was. Give a test ad set a fixed review date and judge it on the whole period.

The part of Iles’s list that carries over best is the least glamorous one: know which of your ads lost and which were never tried. Almost everything else on the ad-account side follows from getting that right.

Frequently asked questions

Is CBO or ABO better for testing Meta ads?

For testing, ad set budgets (ABO) give you more control, because you decide how much each test ad set gets rather than letting Meta move budget toward whatever it already favours. For small NZ accounts, though, the bigger issue is not the budget setting. It is keeping proven ads out of the test ad set so new ads actually get spend.

Should I run one ad per ad set on Meta?

Not as your default structure on an NZ lead-gen budget. Meta says an ad set needs about 50 results in a week to exit learning, and many single-ad ad sets on a small budget will not get there. Use a single-ad ad set as a targeted tool when an ad you believe in keeps getting starved of spend.

How do I know if a Meta ad failed or just did not get enough spend?

Check spend by ad, not just results. As a working rule, an ad that has spent two to three times your target cost per lead with no leads has had a fair test. An ad that has spent less than your target cost per lead has not been tested yet, so it has not failed.

How do you scale Meta ads on a small NZ budget?

Find one ad that wins in a small test ad set, leave it where it won and raise that ad set’s budget in steps, then feed in new versions of the same idea with a different audience call-out, face, promise or offer package. Our guide to scaling again after a failed attempt covers how big each step should be.

How many ads should a small business run on Meta?

Fewer than the big-account advice suggests. The number is capped by how many leads your budget produces each week. Our creative testing guide gives a count for NZ$50, $200 and $1,000 a day.

Do I need UGC creators for Meta lead ads?

You need more than one face. Paid creators are optional. Staff and real customers reading or talking about a proven message will reach people the owner alone does not. Paid local creators are worth adding once you have a script that has already won.

Find out which of your ads were never tested

When a Meta account stalls, the gap is rarely a shortage of ads. It is usually a shortage of clear answers about which ads lost and which never got a chance.

If you want your Meta account set up so every test produces a result you can act on, book a strategy call and we will go through your ad-level spend and show you what your budget can realistically test.