Meta Ads Optimization: Why Your Ads Stopped Converting and How to Fix It

Start with the date results broke

Diagnosis starts with a date, not a theory. Break spend, purchases and CPA (cost per acquisition) down by day for the last 30-60 days. Mark the first day a line moved.

That date tells you which bucket you are in. Each bucket has different causes and fixes.

Next, check the date against your own change log. Look for theme updates, a new checkout or upsell app, a payment gateway switch, price or shipping fee changes, and the end of a sale. Teams often blame Meta for a break they caused on their own site the same day.

Meta's 2026 reporting changes can fake a performance drop

If the break lines up with 12 January 2026 or March 2026, part of the drop may be reporting, not buyers. Meta changed what it counts, so the same sales can show fewer conversions.

According to reporting on Meta's Ads Insights API change, the 1-day click, 7-day click, 28-day click, 1-day engaged view and 1-day view windows remain in the API. In Ads Manager, guides report that since March 2026 the default for website conversion campaigns is 7-day click, 1-day engage-through and 1-day view. Sources disagree on whether 28-day click is still shown in the standard interface, and availability seems to vary by account.

The same change limits API history. PPC Land and others, citing Meta's developer announcement, report that unique metrics are kept for 13 months and frequency data for 6 months. Looker Studio dashboards or data warehouse reports that pull from the API may quietly lose older rows.

Reconcile Meta with your backend before you edit anything

Your store, not Ads Manager, says whether sales actually fell. Put Meta purchases next to Shopify or WooCommerce orders by day, net of cancellations and returns.

If store orders held steady while Meta purchases fell, check tracking first. Confirm the Pixel and Conversions API (CAPI) both fire, deduplicate correctly, and send the right value and currency. Our Pixel and Conversions API setup guide covers each check. A new checkout app or payment gateway is a common reason the Purchase event stops firing.

COD and RTO: when a purchase is not a sale

In India, the Gulf and parts of South-East Asia, many orders are COD (cash on delivery). The Purchase event fires at checkout. If the parcel later comes back as RTO (return to origin), the customer never paid, but Meta has already learned from that buyer.

High RTO can train the algorithm toward people who order and refuse delivery. Options to test on one campaign:

• a prepaid incentive or partial advance on COD orders
• sending only confirmed COD orders, verified by call or WhatsApp, as purchases through CAPI

Fewer events can push ad sets into Learning limited, so test before you roll it out across the account.

Tax hiding inside your ROAS

If your Purchase event sends tax-inclusive order values, Meta's revenue includes GST in India, VAT in the UK, EU and Gulf, or sales tax in the US. ROAS (return on ad spend) looks higher than the money you keep.

Tax on your ad bill also adds cost unless you can reclaim it. Indian tax advisers say Meta's Indian invoices carry 18% GST, which registered businesses claim as input tax credit, or pay under reverse charge if billed from abroad. Tax advisers report that UAE accounts pay 5% VAT without a valid TRN (tax registration number), UK B2C advertisers usually pay 20% VAT, and some countries add location fees of up to 3%. Check your own Meta invoice.

Turning Meta revenue into real profit

Here is how a healthy-looking ROAS can still lose money once returns and tax come out.

Worked example: a brand spending ₹30 lakh a month

ScaleWin judges ads on this break-even view, after product cost, shipping, GST and RTO, not on platform ROAS.

Delivery failures that look like conversion problems

Sometimes conversions fell because ads stopped serving. Check delivery before anything else, because it takes minutes.

1. Billing: practitioner reports say an expired card or failed payment can pause the whole account without a clear alert.
2. Spending limits: an account spending limit set months ago can cap a growing account.
3. Rejections: ads can be rejected on re-review long after launch. Filter by delivery status.
4. Schedules and end dates: check for campaigns that ended with a sale or run on a schedule.
5. Silent ad sets: filter for ad sets marked Active with zero impressions since 15 January 2026.

Those silent ad sets are the deprecated-interest problem. This breakdown of the targeting changes explains that affected ad sets do not pause, error or email you. They simply stop spending, though other reports say Ads Manager showed audience warnings before the cutoff. Detailed-targeting exclusions had already been removed from ad sets in March 2025, according to third-party reports. Rebuild them on broad or Advantage+ audiences rather than hunting for replacement interests.

You may have restarted learning yourself

Frequent edits are a common cause of a performance drop the team created. Guides quoting Meta's Help Center say ad sets usually exit learning after about 50 results in the week after the last significant edit, so every significant edit restarts the count.

Meta's Help Center, as quoted by Scalemate, defines a significant edit as pausing the ad set or changing the optimization event, audience or creative. Budget and bid changes may count, depending on size. Guides citing Meta's list add that adding a new ad or pausing for 7 days or more also restarts learning.

How to check it in Ads Manager

• Add the Last significant edit column at ad set level.
• Compare that date with the date performance changed.
• Switch the date preset to since last significant edit to see only post-edit data.

Learning limited is not a penalty. It means the event is too rare or the budget is split across too many ad sets. Consolidate first. US brands running Shops ads should know that guides quoting Meta's Help Center give them their own threshold: 17 purchases through your website and 5 through Meta after 7 days. Our guide to exiting the learning phase faster covers the fixes in depth.

Creative concentration: why 20 ads behave like 3

If Meta spends on only a few of your ads, it may be treating many of them as one. Andromeda, the ad retrieval system that industry coverage says replaced Meta's older ad selection between late 2024 and October 2025, is reported to group creatives that look alike. We could not confirm the timing from Meta.

Separate official metrics from agency talk. "Entity ID" is shorthand for Meta's internal similarity grouping, not an official term. Practitioners report creative metrics such as a Creative Similarity Score and a Creative Fatigue Score in Ads Manager. Check what your own account shows. The popular 60% and 40% thresholds come from third parties, not Meta.

Diversify concepts, not variants

New headlines, colour swaps and AI variations of one winner may count as the same ad. Change at least one of these:

• the hook or opening problem
• the format, such as UGC video, static, carousel or demo
• the creator or face on screen
• the setting and visual style

To spot fatigue, follow each creative's CTR (click-through rate) and CPA over time. Our guide to creative fatigue shows how to read those trends.

What most guides get wrong about ads that stop converting

• "The default window is 28-day click, 1-day view." Since March 2026 it is reported as 7-day click, 1-day engage-through and 1-day view. Reports say the longest view window left is 1 day.
• "Any 2026 drop is fatigue." Drops near 12 January or in March 2026 may be measurement changes.
• "Get 50 conversions a week." Meta's Help Center wording, as quoted in secondary guides, is about 50 results in the week after the last significant edit.
• "Duplicate the campaign to reset it." Practitioners report this no longer resets performance the way it used to. Treat it as unreliable.
• "Meta lowered your quality score." Meta reports relevance diagnostics: quality, engagement rate and conversion rate rankings, not one score.
• "Refresh when frequency passes 2.5." Fixed frequency cut-offs are agency heuristics with no Meta source. Use them only alongside creative-level CTR and CPA trends.

A 15-minute, 1-hour, 2-week triage plan

Work in this order and stop when you find the cause. Most teams skip to the slowest step first.

1. 15 minutes, delivery: billing, spending limit, rejected ads, end dates and Active ad sets with zero impressions.
2. 1 hour, measurement: Events Manager, Pixel and CAPI deduplication, daily store orders net of RTO, and the 2026 change calendar.
3. 1 hour, learning: Last significant edit column against the break date.
4. 2 weeks, performance: one significant edit at a time, new creative concepts, then judge after enough events.
5. Alongside: check offer and page. Price rises, higher shipping fees, removed COD, UPI checkout failures in India and slow pages all cut conversion rate in ways ads cannot fix.

Before suggesting a change, ScaleWin's assistant pulls the last 7-10 days and checks learning status and landing-page views against link clicks to spot page drop-off. For full account structure, see Meta Ads Optimization: The Complete 2026 Guide.

Meta ads optimization: frequently asked questions

Sources

• Meta restricts attribution windows and data retention in Ads Insights API (PPC Land)
• Meta attribution changes and engage-through (Good Morning Co)
• 2026 Meta targeting changes and deprecated interests (Webblend)
• Meta ads learning phase reset and significant edits (Scalemate)

Ajay Kumar

Founder, ScaleWin · 12+ years in performance marketing

Ajay is the founder of ScaleWin. He works hands-on with Meta Ads, Google Ads and SEO for Indian D2C and service brands, and built ScaleWin so that the daily work of a performance marketer - reading numbers, pausing losers, scaling winners - happens automatically.

Drafted with AI assistance. How these articles are written