Why your Meta ads stopped working — and how to fix it.

The pattern we see weekly

An SMB owner sends a panicked message. "Our Meta ads were making money for six months. Last month ROAS dropped from 3.2x to 1.8x. This month it's 1.1x. We boosted budget. Nothing changed."

The first instinct is to assume something inside Ads Manager broke — iOS update, algorithm change, increased competition. The actual cause is almost always a stack of 2–3 smaller problems, none fatal alone, compounding simultaneously. Attribution is usually one of them — which means the "1.1x ROAS" may itself be wrong.

What actually changed with iOS 14 — the technical explanation

Most people know iOS 14 "broke Facebook tracking." Few understand exactly what broke, what Meta built to fix it, and where the fix still falls short in 2026. This matters because the technical gap between what you think is happening and what is actually happening is the most common root cause of decaying Meta accounts.

ATT — App Tracking Transparency

Apple's App Tracking Transparency (ATT), launched in iOS 14.5 in April 2021, requires users to explicitly opt in to cross-app tracking. On average, 25–35% of iOS users opt in — meaning 65–75% of your iOS traffic is invisible to the Meta browser pixel for retargeting and conversion attribution purposes.

The Meta pixel works by reading the browser cookie set when a user clicks an ad, then firing a "Purchase" or "Lead" event when they convert. With ATT, the cookie is either not set or blocked from being read across apps. The conversion happens — the pixel just cannot see it.

SKAdNetwork — Apple's privacy-preserving attribution

Apple replaced cross-app tracking with SKAdNetwork (SKAN), a privacy-preserving attribution framework that:

• Reports conversions in aggregated, anonymized batches — no individual-user attribution.
• Delays reporting by 24–72 hours depending on conversion value tier.
• Limits conversion signal to a single 6-bit value (64 possible states), making revenue reporting coarse.
• Does not support retargeting — only acquisition campaigns.

For Meta advertisers, SKAN means that app-based campaigns lost most of their granularity. Web campaigns were less directly affected by SKAN but were heavily impacted by ATT's cookie-blocking.

The 7-day click vs 1-day view problem

Before iOS 14, Meta's default attribution window was 28-day click, 1-day view. After iOS 14, Meta moved to 7-day click, 1-day view as the default. This change alone reduced reported conversions by 15–30% in most accounts — not because fewer sales happened, but because conversions that occurred 8–28 days after the click were no longer attributed.

Many accounts have never had their attribution windows audited since the change. If your account was set up before mid-2021 and the settings were not manually reviewed, you may be comparing current performance against historical benchmarks that used different attribution windows — making the "drop" look larger than it is.

Server-side events — why CAPI + browser pixel beats either alone

Meta's Conversions API (CAPI) is a server-to-server integration that sends conversion events directly from your server to Meta's servers — bypassing the browser entirely. It does not depend on cookies, browser permissions, or ATT opt-in status.

The reason to run CAPI alongside the browser pixel — not instead of it — is deduplication. Each individual conversion should be sent via both channels and Meta deduplicates them using a shared event ID. Running both channels together:

• Recovers conversions that the browser pixel missed (ATT opt-outs).
• Preserves browser pixel data for users who did opt in (richer signal).
• Increases Event Match Quality (EMQ) by sending hashed email, phone, IP, and fbp/fbc parameters from server-side data.
• Provides a redundant signal path when one channel has a technical failure.

Internal data: CAPI vs non-CAPI account performance

How to set up CAPI correctly

There are three implementation paths. In order of reliability:

1. Direct server integration— your developer sends events from your purchase confirmation / lead submission endpoint using Meta's Business SDK (available for Node.js, Python, PHP, Ruby, Java). Passes hashed email, phone, IP address, user agent, fbp (Meta browser cookie value), fbc (click ID from URL parameter), and event ID for deduplication. This is the most reliable path and the one we implement for all client sites.
2. GTM server-side container — routes conversion events through a server-side Google Tag Manager container before forwarding to Meta. Works well for Shopify and other platforms where direct server access is limited. Adds latency (~200ms) but avoids custom development.
3. Meta's Partner Integrations — Shopify, WooCommerce, and Klaviyo have native CAPI integrations. Enable them in the Events Manager. They are reliable for purchase events but often miss custom events (account registration, trial start, specific form submissions).

Verify your setup with Event Match Quality

After implementing CAPI, go to Meta Events Manager → your pixel → Aggregated Event Measurement. Check the Event Match Quality (EMQ) score for your primary conversion event. Scoring:

• 8.0–10.0 — Excellent. You are passing rich customer data and Meta can match events to users reliably.
• 7.0–7.9 — Good. A few parameters missing; review whether you are passing hashed phone numbers.
• 6.0–6.9 — Marginal. Missing significant parameters. Usually means email is hashed but phone and fbp are absent.
• Below 6.0 — Poor. The algorithm is optimizing on incomplete signal. This is where most browser-only implementations land.

Creative is now the targeting — the 2026 reality

This is the principle most agencies resist because it changes their value proposition. In 2024–2026, Meta's algorithm has become sophisticated enough at audience discovery that the creative asset itself performs the targeting function.

When you run a broad audience (18–65, no interest stacks, no demographic restrictions) with Advantage+ Placements enabled, Meta's system reads the creative, infers the intended audience from the visual and copy signals, and serves it to the users most likely to respond. The creative is literally the targeting parameter.

The practical implication: interests-based targeting and narrow demographic boxes do not improve performance as reliably as they did in 2019–2021. They constrain the algorithm's audience discovery without adding proportional signal quality.

What this means for your creative strategy

If creative is the targeting, hook quality is everything. The first 2–3 seconds of a video or the first 6 words of a static ad determine whether Meta serves it to more of the right people or suppresses it in favor of a different asset.

The hooks that consistently outperform in 2026:

• Problem-first:"If your Meta ads stopped working last month, this is why." Immediately filters for the audience experiencing the problem.
• Specific outcome:"How we got a 2.8x ROAS from a $6,000/month budget." Specific numbers signal credibility and filter for buyers at the right stage.
• Counter-narrative:"Everyone says to boost budget when ROAS drops. That is wrong." Pattern interrupts trained ad-blindness.
• Social proof lead:"312 SMBs used this campaign structure last quarter." Triggers FOMO and authority simultaneously.

Campaign structure for 2026 — Advantage+ vs manual, budget consolidation, creative testing

When to use Advantage+ Shopping Campaigns (ASC)

Advantage+ Shopping Campaigns consolidate prospecting and retargeting into a single campaign and give the algorithm full control over budget allocation, placement, and audience. When it works, it outperforms manual campaigns significantly. The conditions required:

• 50+ weekly purchase events (the learning threshold).
• A tested creative library of 6–10 assets (images, video, carousel).
• CAPI properly configured with EMQ above 7.0.
• Product catalog uploaded and synced (for e-commerce).
• A Pixel that has been firing for 6+ months with consistent data.

If any of these conditions are not met, start with manual campaigns until they are. Running ASC on an account without these foundations produces erratic results and burns budget on audience exploration before the algorithm has enough signal to optimize.

Manual campaign structure for SMBs in 2026

For accounts under 50 weekly conversions or under $8,000/month, the structure that consistently outperforms:

• 1 prospecting campaign — broad audience (no interest stacks), Advantage+ Placements enabled, CBO (Campaign Budget Optimization). 2–3 ad sets by creative angle (not by audience). 4–6 ads per ad set.
• 1 retargeting campaign — engaged custom audience (video views 50%+, website visitors 30 days, engagement 60 days). Manual budget. 3–4 creative variants with a stronger offer.
• No more than 2 campaigns active simultaneously at sub-$5k/month budgets. More campaigns at this spend level fragment conversion data and prevent any single campaign from reaching the 50-conversion learning threshold.

Creative testing structure

The creative testing process that produces reliable winners:

1. Hook test: 4–6 videos with different first 3 seconds, identical body. Run at $20–30/day each for 7 days. Kill anything with Hook Rate (3-second video plays / impressions) below 25%.
2. Body test: Take the winning hook. Test 2–3 different value propositions in the remaining 15–30 seconds. Run for 7 days. Optimize for cost per click to landing page, not conversions (budget too small to get stat significance on conversions at this stage).
3. Offer test: Take the winning hook + body. Test 2–3 different CTAs or offers. Now optimize for conversions. This is the ad that goes into production spend.

This 3-stage structure finds winning creatives in 3–4 weeks while keeping test budgets under $3,000 total — compared to the common mistake of running 10 completely different ads simultaneously with no systematic isolation of the variable being tested.

Budget consolidation — why less is more

The most common structural mistake in SMB Meta accounts: too many campaigns, too many ad sets, too little budget per unit. An account with $5,000/month split across 4 campaigns and 8 ad sets has no single unit reaching the 50-conversion threshold. Every ad set is in permanent learning phase. Performance is unstable by design.

The fix is counterintuitive: consolidate to fewer campaigns with more budget concentration. One prospecting campaign, one retargeting campaign, CBO at the campaign level, 3–4 ad sets maximum. Let the algorithm allocate within campaigns rather than forcing budget allocation through manual ad set distribution.

The full diagnostic order

We run these in sequence. Each step takes 30–60 minutes on a properly-instrumented account. Attribution first — always — because without accurate measurement, every other diagnostic is reading a broken instrument.

1. Verify CAPI + Event Match Quality + AEM configuration. Confirm EMQ is above 7.0 for primary conversion event. Confirm 7-day click attribution window is set consistently. Confirm AEM event prioritization is correct (primary conversion in slot 1).
2. Pull frequency, CTR, CPM by ad set — frequency above 3.5 and CTR dropped 25%+ from baseline = creative fatigue.
3. Pull audience overlap report — 60%+ overlap between ad sets = saturation. Consolidate.
4. Run Lighthouse + GA4 week-by-week conversion rate trend. Identify the week the decay started. Correlate to any site changes: redesign, new apps installed, price changes.
5. Competitor + offer audit. Has market pricing moved? Has a new alternative entered? Is seasonal demand down in your category?
6. Campaign structure audit. Count active campaigns, ad sets, and weekly conversions per unit. Identify if any unit is near or above the 50-conversion threshold — those are the ones to protect. Everything else gets consolidated.

Suspect 1: Creative fatigue

The most common cause of Meta decay for SMB accounts. Symptoms:

• Frequency above 3.5 per audience.
• CTR dropping 25%+ from the campaign's historical baseline.
• CPM rising even though audience size has not changed.
• The same 2–3 creatives running for more than 6 weeks.

The fix: ship 4–8 new creative concepts in the next 7 days. Not variations of the same hook — genuinely different angles, formats, and value propositions. Test them at low budget against the existing winner. Whatever beats the winner replaces it.

Suspect 2: Audience saturation

For SMBs targeting a niche audience (a city, a specific demographic, a product category), Meta can exhaust profitable impressions faster than expected. Symptoms:

• Reach growth has flattened despite stable spend.
• Custom audience overlap shows 60%+ overlap across ad sets.
• Cost per result rising while click volume holds steady.

The fix: consolidate ad sets to reduce overlap, expand audience by 30–50% (broader interests, larger geo, looser age band), or shift spend up-funnel to video views and engagement campaigns to rebuild the retargeting pool before returning to conversion campaigns.

Suspect 3: Attribution drift

Most SMB accounts have one of three broken setups:

• Browser-only pixel, no CAPI. Losing 30–60% of conversion signal from iOS traffic. Reported ROAS is wrong; the algorithm is optimizing on incomplete data.
• CAPI installed but EMQ below 7.0. Meta cannot deduplicate browser + server events properly. Reported ROAS double-counts some conversions, misses others.
• AEM configured wrong. Wrong event prioritization means the algorithm optimizes for a secondary event (add-to-cart) rather than the primary conversion (purchase). Often set up once and never reviewed.

Suspect 4: Landing page degradation

The Meta side looks healthy, but the destination got worse. This is invisible inside Ads Manager. Common causes:

• A redesign shipped. New page has different hierarchy, weaker social proof placement, slower mobile load. Converts 30%+ worse.
• A Shopify app was installed. Page weight doubled. Mobile bounce rate up, conversion down.
• Price increased. Cart abandonment up. ROAS down — correctly.
• A third-party script (chat, popup, video) slows time-to-interactive.

The fix: run a Lighthouse audit on the destination page. Pull GA4 conversion rate by week for the last 90 days. The week conversion rate broke is almost always the week something changed on the page — not in Ads Manager. For a rebuild path, see our notes on landing page builds.

Suspect 5: Offer staleness

The ad and page work fine — but the market has moved. Symptoms:

• Competitor pricing has shifted and your offer is no longer competitive.
• Industry standard terms have changed (free returns are now expected; you charge for them).
• A new alternative entered the market at a better unit-economic position.
• Seasonal demand dropped. The account is fine; the category is soft.

The fix: step out of Ads Manager. Run a competitor positioning audit. Update the offer — price, bundle, guarantee, urgency mechanism — and relaunch with fresh creative reflecting the change. Sometimes the problem is the strategy, not the ads. For the strategic level see marketing strategy.

What we will not recommend

• Turn everything off and reset. You lose 6 months of conversion signal — the most valuable asset in the account. Surgical fixes beat resets in 95% of cases.
• Switch to Advantage+ to "let the AI handle it." Advantage+ campaigns are excellent when the account is healthy. On a broken account without sufficient conversion volume or CAPI data, switching to ASC accelerates the decay — the algorithm has more budget and less signal to work with.
• Move budget to Google Ads as the fix. Google Ads is a strong channel, but if your Meta funnel is broken, your Google funnel has the same downstream problems (landing page, offer, tracking). Fix the funnel; channel choice is secondary.
• Add more campaigns to "test different approaches." Budget fragmentation is almost always the wrong move at SMB spend levels. Consolidation and patience beats proliferation.

Next step

Grant us agency access to your Meta Business Manager. We run the 6-step diagnostic over 5 business days — attribution audit, creative fatigue analysis, audience overlap report, landing page Lighthouse, CAPI configuration review, and campaign structure assessment — and send you the written findings with prioritized fixes. Billed as a one-time audit, not a retainer commitment.