Case Study

​

01

The setup

​

The client is a small, founder-run home decor and lifestyle brand featuring hand-designed pieces rooted in cultural heritage. The brand had been running Meta ads for nearly three years across a rotation of small monthly campaigns, none producing sustained results.

The pattern was familiar for small ecommerce: some months better than others, no reliable trajectory, cost per acquisition climbing over time. June closed at a $94 CPA on 13 purchases. July finished slightly better at $84 CPA on 16 purchases. Both months operated at a net loss on paid acquisition.

​

The obvious response was to do more. Test new creative. Rotate audiences. Add campaigns. Optimize aggressively. What actually worked was the opposite.

​

02

The diagnosis

​

Two months of hands-on management surfaced a pattern worth naming: changes were happening too frequently. Every audience adjustment, every creative refresh, every budget swing reset Meta's learning phase — and reset the algorithm's ability to actually optimize.

The account's best historical months (a $32 CPA month in one high-performing period, $34 CPA in another, $37 CPA in a third) shared one thing in common: they were periods when the campaign ran with minimal intervention. When it was left alone, it worked. When it was actively managed, it degraded. The intervention wasn't the strategy. The intervention was the problem.

​

03

The approach

The August plan was intentionally boring:

Refine audience signals once by adding a lookalike audience built from the brand's actual customer list. Hold the campaign structure through end of month. Skip creative refreshes unless frequency crossed the fatigue threshold. Check performance twice weekly instead of daily. Do the harder work of client communication instead of ad-side tinkering.

Two touches in thirty-one days. Everything else — hands off.

​

04

The results

August closed with dramatically better performance across every meaningful metric.

Purchases32↑ 2.5× June volume

Cost Per Acquisition$48↓ 49% from June

Ad Spend$1,535Same budget bracket

ROAS2.4×↑ From 1.2× in June

Compared to July: double the purchase volume at 43% lower cost per acquisition. Compared to June: 2.5× the volume at nearly half the cost. Cost per landing page view dropped from $1.72 to $1.04. Ad quality ranking held at "Above Average" throughout the entire month. Frequency stayed healthy at 2.53.

The best thing to give the campaign was a configuration to optimize against.

​

05

What made the difference

​

Three principles carried this account:

​

1. Stable inputs beat clever inputs.

Meta's algorithm rewards consistency. Every change resets learning. The best thing a campaign can be given is a configuration to optimize against — and the discipline to leave that configuration alone.

​

2. Better audience beats broader audience.

A single lookalike built from actual purchasers moved the campaign more than three months of audience experiments combined. Quality of seed data matters more than quantity of tests.

​

3. The right work isn't always on the platform.

The bottleneck was rarely in Ads Manager. The real levers — average order value, checkout friction, repeat purchase capture — sit outside the ad account. Recognizing that changed the strategic recommendations to the client, not the ad-side execution.

​

06

Reflection

The month proved something worth internalizing: paid media practitioners often confuse activity with strategy. Making changes feels productive. Watching a campaign run untouched feels like nothing is happening. But the discipline of doing less — when the data supports it — is often the most valuable skill a media buyer brings to an account.

​

The best month in three years did not come from a tactical breakthrough. It came from restraint.

​​