Stop shopping for an ecommerce Meta agency. The niche on its homepage will not move your CPA.

I used to sell that badge myself. When I managed accounts by hand, I walked prospects through interest stacks and lookalikes I had built for brands like theirs. They nodded because it sounded like insider knowledge. I was wrong to sell it, and they were wrong to buy it. Meta changed the machine underneath us. Much of that old audience knowledge expired.

Meta finds your buyers from your creative and your conversion signal, not from your agency’s client list. If you want lower CPA, ask how many new ads ship each week, whether the purchase or trial event is clean, and whether the manager knows your break-even number. The vertical badge is mostly positioning.

Why the ecommerce badge used to earn its fee

In the detailed-targeting era, that badge meant something real. I spent nights building interest stacks for a skincare brand: layering engaged shoppers over beauty interests, excluding past purchasers, and seeding 1% lookalikes off add-to-carts because the client did not have enough purchases to seed anything better. It was fiddly work. If an agency had already done it for ten stores like yours, it started faster. That head start was worth paying for.

Meta took much of that head start away. Advantage+ Audience is now the default and treats interest inputs as suggestions, not rules. Location and minimum age remain hard constraints. Detailed exclusions are gone, and Meta consolidated or removed dozens of interest categories in January 2026. My old stack-building skill is now a poor reason to choose a manager.

That does not mean the product, customer or offer stopped mattering. It means an agency cannot hand Meta a list of past clients and expect better delivery. The ad has to speak to the right buyer, and the conversion signal has to tell Meta what a valuable buyer did. A familiar vertical is not a substitute for either one.

The first lever: new creative, not an inherited audience map

Meta’s own benchmarks put switching to Advantage+ at up to 32% lower CPA in ecommerce and lead gen, with 11 to 15% higher CTR. The mechanism matters more than the headline number. When you stop constraining delivery with narrow interests, the system can test your ad against more people. The creative gives it a reason to find the people who stop scrolling. As one guide puts it, creative is the targeting signal.

That reframes the manager’s job. Only about 5 to 10% of creatives meaningfully outperform, so you need roughly 15 tests to find one winner on average. A manager shipping two variations and waiting for a clever audience setting to rescue them is not running enough experiments.

The rough production pace I look for depends on spend:

  • At $5k to $15k a month, ship 5 to 10 new creatives a week.
  • At $15k to $50k a month, ship 10 to 25 a week.
  • Put $50 to $200 behind each test before judging it.

Those are working rules, not a license to make 25 near-identical ads. Test different hooks in the first two seconds. Put UGC next to studio work, and offer-led messages next to problem-led ones. Then kill the losers. I used to tell clients audience research was the work. I was wrong. Shipping volume and variety is the work now.

The second lever: the event Meta learns to find

Creative gets an ad in front of people. The conversion event tells Meta which response to pursue next. My rule is to optimize to the event closest to revenue that still clears about 50 events per ad set per week. For ecommerce, start with Purchase. Only step back to AddToCart or InitiateCheckout if purchases are too sparse for the system to learn; fewer than 25 to 50 purchases a week is the warning range I use.

That fallback has a cost. Optimize to AddToCart because it looks cheaper, and you train Meta to find people who add things to carts. Some of them buy; some abandon them. The report can fill with cheap results while revenue stays thin. Then everyone asks whether the agency has enough ecommerce experience, when the more useful question is why it chose that signal and what it will take to move back to Purchase.

Do not confuse a cheaper event with a cheaper customer. Make the manager explain the event choice before the campaign goes live, not after the results disappoint.

The third lever: margin, not someone else’s ROAS benchmark

A 4.0 ROAS tells me almost nothing until I know contribution margin. Break-even ROAS is one divided by contribution margin expressed as a percentage, with that margin calculated after product, fulfillment, shipping, payment fees and returns. One worked example puts a 38.75% margin at roughly 2.6 break-even on net revenue, or 3.1 in platform gross terms once VAT sits in the number.

The same 4.0 ROAS can lose money at a 22% margin and make money at 48%. That is why a vertical case study can be true and still be useless to you. It shows what happened under somebody else’s economics, not what your account can afford to acquire.

Ask any agency quoting a benchmark to calculate your break-even with you. If the manager cannot do that math, the logo wall does not matter. A profitable-looking dashboard is not a profitable account.

The SaaS trap: cheap trials that never become customers

SaaS founders make the same mistake with a different label. They hire a SaaS Meta agency. The agency optimizes to trial start because trial starts arrive in volume. CPL drops to $9. Everyone celebrates. Three months later, sales says those trials are not activating.

I have seen the split documented: two ad sets optimized on trial start, one driving 220 trials at $9 and the other 140 at $15. The cheap set produced 4 paid conversions. The expensive set produced 19. On the trial-start report, the first set wins. On the number the business cares about, it does not. Cheap trials taught Meta to find tourists.

The fix was not a more impressive list of SaaS clients. It was moving the conversion event down to activation, defined there as connecting a data source and completing a report, once that event cleared about 50 a week. That is the ecommerce lesson with a longer fuse: Meta learns from the event you pay it to find. Pay it to find raw trial starts, and raw trial starts are what you can expect to measure.

I wrote more about cheap Meta trials training your funnel to find tourists. Read it before signing an agency that brags about cost per trial without putting trial-to-paid beside it. The practical question is not whether its homepage says SaaS. It is whether the manager can name the point where a trial becomes worth acquiring.

Where vertical experience actually earns its fee

I will give the badge two exceptions. Neither is a generic claim that an agency “knows your customer.” Both involve work that changes what the manager can do inside the account.

First: regulated categories. Housing, employment, credit and social issue ads face Special Ad Category restrictions that limit targeting: no gender or ZIP-code targeting, a 15-mile minimum radius, an age range locked to 18 through 65-plus, and no detailed interests, lookalikes or exclusions. Meta can apply limits based on the creative itself, not just the category checkbox.

An agency that has operated inside those guardrails may know how to write compliant copy that still converts and how to structure broad geography without bleeding budget. That is useful experience. Ask what it changes in the work, not how many regulated logos fit on a slide.

Second: long sales cycles that need offline signal. A January lead that becomes a September opportunity will not fit neatly inside Meta’s conversion window. Without pipeline data coming back through Conversions API, Meta can keep optimizing toward cheap form fills rather than MQLs, SQLs, opportunities or closed-won customers.

I ran SaaS accounts where CPL looked great and lead-to-SQL was a disaster. No audience trick fixed that. The plumbing mattered: sending the later-stage signal back so the account was not judged by the easiest form to fill. Ask how many accounts an agency has wired to offline conversions. The answer tells you more about its usefulness than a wall of SaaS logos.

Vertical experience earns its fee when it solves a constraint in your account. A category label by itself solves nothing.

Three questions that break the niche pitch

When an agency says it specializes in businesses like yours, I do not ask for another case study. I ask these three questions. Each forces the conversation away from who the agency sold to and toward what it will do.

  1. How many new ads did you ship per client per week last month, and who made them? I want a number and a name, not a process deck. The answer should make it clear whether new ideas can reach the account every week. If it amounts to two boosted posts and a Canva freelancer, I walk away.
  2. What event will you optimize to, and what gets us to 50 of them a week? For ecommerce, start the conversation with Purchase. For SaaS, ask about activation or a qualified trial rather than raw trial start. Make the manager name the fallback if the preferred event lacks volume, and explain how the account gets back to the event that matters. “We’ll monitor performance” is not an answer.
  3. What is our break-even ROAS, and what breaks if margin moves? Bring contribution margin after shipping, fees and returns. A good manager works through the math with you on the call. A badge manager reaches for somebody else’s benchmark. If your economics change, the target should change with them.

Ask those questions and the vertical pitch collapses fast. I have watched agencies with twenty DTC logos stumble on question two, still optimizing to AddToCart because Purchase volume looked thin while leaving the feed or offer problem untouched. I have watched generalists with no SaaS logos nail all three because they knew how to wire offline conversions and say no to cheap trials.

If you run ecommerce and want the longer version of that structure choice, compare Advantage+ Sales versus hand-built campaigns before letting anyone rebuild your account from scratch. A rebuild is not proof of management. Ask what it fixes.

Hire for the loop. Check it in your own account.

A vertical badge suggests that someone else’s account will teach Meta how to run yours. It will not. What transfers between accounts is process: the cadence of new creative, the discipline to optimize toward revenue, and the habit of checking performance against margin. That is what I look at in an audit now.

I open the ad account and count the live variations shipped in the last 30 days. I open Events Manager and check which event carries the volume and whether deduplication through Conversions API is working. I open the P&L and calculate break-even ROAS from contribution margin, not Shopify gross. Those checks tell me whether the manager is running a loop or presenting a story about one.

An agency that runs that loop every week will beat an ecommerce specialist that runs it once a month. The logo wall tells you who the specialist sold to. It does not tell you what changed in an account on Tuesday, what was killed on Wednesday or whether anyone noticed the event was rewarding the wrong customer.

Do not take any of this on trust, including from me. Give the shortlist the same paid test inside your account: same budget, same offer, same two-week window. Look for a dated log of action: ads killed, budgets moved, signal fixed and reasoning written down. If a manager cannot show you what changed and why, they did not manage anything. A deck about vertical playbooks fades fast when real spend is on the line.

That is why I run Meta the way groas runs it now: Meta Ads worked around the clock for one flat $1,499 a month, with audiences, budgets and ads tested continuously, new ads shipped, winners scaled, and actions logged with the reasoning beside them. No percentage of spend, no setup fee, live inside your accounts within 24 hours of onboarding, and a named manager who owns the number. Judge that offer by the same questions. A flat fee is not the point if the work does not happen.

Keep shopping for the niche badge and you will get what that badge was built to sell: a higher retainer for old targeting skill, cheap trial and add-to-cart volume that flatters the report, and another quarter explaining why Meta does not match Shopify. Hire for the loop and you can see the work that is supposed to lower CPA.

Ignore this and you will keep paying for the badge. Another quarter of narrow audiences choking Advantage+, another report celebrating cheap add-to-carts and $9 trials that never activate, another benchmark deck that never mentions your margin. Meta does not care who else your agency ran. It cares what you feed it this week: fresh creative, a revenue event with volume, and a manager who knows the break-even. Hire for that loop or keep funding someone else’s logo wall.