Your First 90 Days With a Meta Ads Agency: The Ecommerce Brand’s Benchmark Guide

Most ecommerce brands fire their Meta ads agency in month 2.

Not because the agency did something wrong. Because the brand had no framework for what month 2 is supposed to look like — and when results didn't match a vague expectation of "things should be improving by now," they pulled the plug.

The first 90 days of any agency relationship is a distinct phase with its own benchmarks, its own milestones, and its own set of things that should and shouldn't be happening. Treating it like a regular month — where results are the primary signal — is the fastest way to make a bad decision in either direction.

This guide breaks down what the first 90 days actually look like across fashion/swimwear and health/natural ecommerce accounts, what your agency should be doing at each stage, and what signals actually matter versus which ones are noise.

Key Takeaways

→ Months 1–2 will almost always underperform your steady-state ROAS target. That's not a problem — it's how Meta's algorithm works.

→ The learning phase requires a minimum of 50 optimization events per ad set per week. Well-structured ecommerce accounts typically exit it within 14–21 days from launch.

→ Creative testing volume in the first 60 days tells you more about your agency's capabilities than any ROAS number.

→ At day 90, you should have a real performance baseline — not just a number, but a structural understanding of what's working, what's not, and why.

→ Based on Dash Activate Online's managed ecommerce accounts: agencies that hit green or yellow on five or more of the 90-day milestones below consistently reach the brand's target ROAS by month 5–6. Agencies that hit red on three or more rarely recover without a structural reset.

Why the First 90 Days Are a Different Category Entirely

The mistake most brands make: they evaluate month 1 of an agency relationship the same way they'd evaluate month 12.

That's not how Meta's algorithm works.

When a new agency inherits an account — or restructures an existing one — Meta's delivery system has to relearn your audience. New campaigns enter a learning phase where the algorithm tests audiences, placements, and creative combinations to find efficiency. During this phase, your cost per purchase will be higher and your ROAS will be lower. That's not the agency failing. That's the algorithm doing exactly what it's supposed to do.

The learning phase, defined: A period during which Meta's delivery system has not yet gathered enough data to optimally serve your ads. An ad set exits learning once it collects at least 50 optimization events (usually purchases) in a seven-day window. Until then, performance is inherently volatile.

The question isn't whether performance dips in the first 30–60 days. It almost always does. The question is: is your agency building something that compounds, or are they just spending your budget while the learning runs out the clock?

Here's how to tell the difference, phase by phase.

Month 1 (Days 1–30): Foundation, Audit, and the Learning Phase

What your agency should be doing in month 1

The first month is foundation work. An agency that launches campaigns on day 1 without spending time on setup first is taking a shortcut that will cost you later. Here's what a legitimate onboarding looks like:

→ Full account audit: pixel health, attribution setup, historical ROAS by campaign type, audience exclusions, conversion tracking verification

→ CAPI verification: Conversions API setup ensures purchase signals reach Meta even where browser-based tracking fails (especially important for health/natural product brands with high iOS buyer populations)

→ Campaign architecture: building or restructuring the prospecting vs. retargeting separation, aligning campaign objectives to where your buyers actually are in the funnel

→ Audience segmentation: existing customers, 30/90/180-day website visitors, email list uploads — all installed before significant spend goes live

→ Creative intake: reviewing what's been tested, what hasn't, what's worked, and what needs to be built

If your agency skipped straight to running ads without this foundation work, you'll spend months chasing problems that should have been solved on day 3.

What the numbers will look like — and why that's OK

During month 1, ROAS will likely run 20–40% below your target. CPM may spike as new campaign structures enter the auction. Cost per purchase will feel high.

This is normal.

Each time you make a meaningful change to a campaign structure — new ad sets, new objectives, new audiences — the learning phase resets. An account that exits month 1 with clean structures, verified attribution, and a tested creative set is in a far better position than one that spent the month chasing early ROAS by layering changes on top of changes.

What you need to do in month 1

Feed your agency creative. This is the phase where creative inputs matter most. If you're slow to produce or approve, you'll burn 2–4 weeks of the learning window waiting on assets. Your agency can't optimize what they don't have.

Don't touch the campaigns. Every structural change to a running campaign — new audiences added, budgets shifted significantly, creative swapped mid-flight — resets the learning phase counter. Resist the urge.

Set a day-1 baseline snapshot. Ask your agency for a documented starting state: current ROAS, CPM, CPC, CAC, click-through rates. You'll need these numbers in 60 days when you want to know if anything has actually moved.

Month 1 red flags

Month 2 (Days 31–60): Creative Testing Windows Open

Month 2 is where you start to see what your agency is actually made of.

The learning phase is settling. You should have stable campaign structures running. The work shifts from foundation to iteration — specifically, creative testing.

This is the most diagnostic phase of the first 90 days for ecommerce brands on Meta. Here's why: Meta's delivery algorithm treats creative as the primary targeting signal. Advantage+ has largely flattened audience-level targeting. What your agency does with creative — how many concepts they're testing, how they structure variants, what they learn from each test — is the primary lever they have over performance.

What good creative testing looks like in month 2

A legitimate Meta ads agency running an ecommerce account should be cycling through 3–5 new creative concepts per month at minimum, with multiple format variants per concept (static, video, carousel). At that volume, you start generating signal on what actually resonates versus what you assumed would.

By day 60, that testing process should have produced:

→ A clear read on which creative angles are generating the most efficient cost per purchase

→ Initial data on audience segment performance broken down by campaign type (new vs. returning, broad vs. interest-based)

→ At least one concrete learning — something specific your agency now knows about your buyer that wasn't knowable on day 1

That last one is the real test. Ask your agency directly: "What have you learned about our customer in the last 30 days?" If they can answer with specifics — "The angle around [problem X] is outperforming the [benefit Y] angle by 40% on CPC, which tells us the buyer is more aware than we assumed" — you have a team that's actually running experiments. If the answer is vague, you have a team managing spend without learning.

The ROAS "trough" — what it is and when to worry

Most ecommerce brands experience a ROAS trough somewhere in weeks 4–7. New structures are stabilizing, creative is being tested, and the algorithm is still optimizing. This is normal and expected.

What's not normal: a ROAS trough that extends past day 60 with no upward movement and no clear explanation from your agency of what's being tested and why.

The distinction matters. A ROAS trough with active learning behind it is a process. A ROAS trough with no discernible testing activity behind it is a problem.

Month 2 red flags

Month 3 (Days 61–90): Establishing the Baseline

By month 3, the learning phase is behind you. You have 60-plus days of real data. Your agency has made decisions, tested creative, and adjusted based on what it found.

This is the phase where you set your true performance baseline — and where you can make a genuinely informed judgment about the agency relationship for the first time.

What you should have at day 90

→ A stable new customer CAC by campaign type (prospecting cold audiences vs. retargeting warm ones)

→ A clear ROAS trajectory — not just a number, but a direction. Is it improving month over month, flat, or declining?

→ Creative performance data: which concepts won, which were cut, and why

→ A forward-looking testing roadmap: what comes next in months 4–6

→ Honest attribution context: the gap between your in-platform Meta ROAS and your blended ROAS calculated from backend revenue (Shopify or equivalent)

Related: How to Evaluate Your Meta Ads Agency: The KPIs Ecommerce Brands Must Track

Vertical-specific ROAS baseline expectations at day 90

Most benchmark guides give you one universal ROAS target. That's not useful when ROAS performance at 90 days looks meaningfully different depending on your vertical.

Fashion and Swimwear (ecommerce, $500K–$5M revenue):

The fashion vertical benefits from broader audiences and higher purchase frequency, which means the algorithm gathers signal faster. Campaigns typically exit learning more cleanly, and prospecting campaigns can reach efficiency earlier.

Blended prospecting ROAS at day 90 for fashion/swimwear accounts generally lands in the 1.5x–3.0x range, depending on average order value (AOV) and contribution margins. If you launched heading into an off-season for your product category, expect the lower end of that range — the algorithm needs demand to learn from, and thin seasonal demand compresses the window.

Health and Natural Products:

Meta's policy environment is meaningfully more restrictive for health and natural products — and that restriction has tightened since the platform's July 2026 shift toward claims-based enforcement. Creative that makes implied transformation claims (before/after, comparative "without/with" framing) faces higher rejection rates, which disrupts the learning phase mid-cycle.

Prospecting ROAS at day 90 for health/natural product accounts typically runs 10–20% lower than fashion equivalents at the same spend level. The constraint isn't demand or audience — it's available creative surface. Your agency needs to be proactively auditing health claims in every asset before launch. If they're not, you'll spend weeks recovering from preventable rejections.

Related: Meta's "Implied Transformation" Rule: Why Health & Supplement Ads Are Getting Banned

Month 3 red flags

The 90-Day Snapshot: A Benchmark Checklist

Use this at the end of month 3 to assess your agency relationship objectively. Based on Dash Activate Online's managed ecommerce client data (aggregated, anonymized), agencies that score green or yellow on five or more of these milestones consistently reach the brand's target ROAS by month 5–6. Agencies that score red on three or more rarely recover without a structural reset of the account and campaign architecture.

What to Do at the 90-Day Mark

The 90-day mark is a natural checkpoint, not a verdict. Here's the framework for reading what you've seen:

If results are trending in the right direction: Stay consistent. The biggest mistake brands make at this stage is introducing new variables — restructured campaigns, new creative briefs, budget cuts — right as the account is building momentum. Give it another 60 days before making structural changes.

If results are flat but you're seeing active learning: Have the deeper conversation. Ask your agency what the hypothesis is for months 4–6. If they have a clear thesis and the foundation is solid, flat months 1–3 can precede strong months 4–6. Flat results with an active testing process are not the same as flat results with no activity behind them.

If results are flat and you're not seeing evidence of learning: This is the signal to push harder. Ask directly: what have you tested in 90 days, what were the results, and what does that tell you about our customer? If the answers are vague or circular, you have a competence problem.

If results are declining and explanations are absent: At this point, the conversation shifts from performance review to accountability review. A declining ROAS with no clear explanation of what changed and what's being done about it is not a learning phase — it's an agency not doing the work.

Related: When to Hire a Meta Ads Agency for Your Ecommerce Brand

The first 90 days of a Meta ads agency relationship should feel like building infrastructure, not chasing a number. Brands that understand what phase they're in — and what their agency should be delivering at each stage — make better decisions, avoid pulling the plug too early, and catch real problems before they compound.

If you're inside a 90-day window right now and want an honest read on whether your account is on the right trajectory, we can walk through the numbers with you.

Book a strategy call — no pitch, no fluff. Just clarity on whether your first 90 days are building toward something real.

Frequently Asked Questions

How long does the Meta ads learning phase last for ecommerce?

The learning phase ends once an ad set accumulates at least 50 optimization events (typically purchases) in a seven-day window. For most ecommerce accounts running purchase conversion campaigns with a healthy account structure, this takes 14–21 days from launch. Accounts with lower daily spend or narrow audiences can take longer. Making structural edits to a campaign — changing audiences, objectives, or budgets significantly — resets the clock.

What ROAS should I expect in the first 30 days with a new Meta ads agency?

Expect your ROAS to run 20–40% below your target during the first 30 days. This is the normal result of new campaign structures entering the learning phase. It is not a reliable signal of your agency's competence in either direction. A better signal at 30 days: has your agency built a clean account structure, verified attribution, and started testing creative?

How many creative tests should a Meta ads agency run in the first 90 days?

A minimum of 6–8 distinct creative concepts across the 90-day window, with multiple format variants per concept. Agencies running fewer than 3–4 concepts in 90 days are managing spend, not testing. Creative velocity is the single best proxy for agency competence during the first 90 days.

When should I fire my Meta ads agency?

Not in the first 60 days based on ROAS alone. If by day 90 you can't get a clear answer to "what have you learned about our customer in the last 90 days" — and you're not seeing a coherent forward plan — that's a legitimate signal. The right time to make a change is after you have real data, not during the learning phase.