Most Facebook advertisers are making ad kill decisions the wrong way. They see a high cost, feel a flash of anxiety, and pause the ad. Or they leave a clearly dead ad running because it "might turn around." Neither approach is a strategy. Both cost you money.
The real problem is not that the data is hard to find. Ads Manager shows you everything. The problem is knowing which numbers actually matter, in what order to check them, and what thresholds separate a struggling ad from a genuinely failing one.
This guide gives you a repeatable decision process you can run through every week. Not a vague framework with phrases like "it depends." An actual step-by-step filter that tells you, for each active ad, whether to turn it off, put it on a watchlist, or scale it up.
Here is what makes this hard for most advertisers: the signals are often contradictory. An ad can have a great click-through rate and terrible conversion numbers. Another ad might look expensive on the surface but be your most profitable when you trace it all the way to revenue. Without a structured approach, you end up making emotional decisions on incomplete data.
The six steps below walk you through the full evaluation process, from setting your benchmarks before you open Ads Manager to building a weekly review routine that keeps your budget working on winners. Whether you are running three campaigns or three hundred ad sets, the logic is the same: let the data make the call, not your instincts.
Step 1: Set Your Performance Benchmarks Before You Judge Anything
Before you can identify a failing ad, you need to define what failure looks like for your specific business. This sounds obvious, but most advertisers skip it entirely. They open Ads Manager, see a high CPA, and start pausing things without ever establishing what their target CPA actually is.
Start with your break-even CPA. Take your average order value, subtract your cost of goods and any fixed overhead allocated to that sale, and what remains is your maximum allowable acquisition cost. If your product sells for $100 and costs you $40 to produce and fulfill, your gross margin is $60. That $60 is the ceiling for what you can spend to acquire a customer before you break even on the first transaction.
If you factor in customer lifetime value, that ceiling rises. A customer who buys three times over 12 months changes the math significantly. Know which number you are optimizing against: first purchase profitability or long-term customer value. Both are valid, but they produce very different CPA targets.
Next, set your target ROAS. This is simply the revenue return per dollar of ad spend you need to hit your margin goals. If you need to generate $3 in revenue for every $1 spent to remain profitable, your break-even ROAS is 3x. Your target ROAS should sit above that to account for overhead and growth goals.
Finally, establish a minimum acceptable CTR. Here is an important note: do not use generic industry benchmarks for this. Use your own account history. Pull your top 20% of past ads by conversion volume and calculate the average link click-through rate across that group. That number becomes your internal benchmark. It reflects your audience, your creative style, and your offer, which is far more relevant than any external average.
Common pitfall: Advertisers turn off ads before they have defined what "bad" actually means for their business. If you do not have a written CPA target, ROAS target, and minimum CTR sitting next to you when you open Ads Manager, you are not ready to make a kill decision.
Success indicator: You have three numbers written down before you evaluate a single ad: your target CPA, your target ROAS, and your minimum acceptable CTR based on your own account history.
Step 2: Confirm the Ad Has Had a Fair Evaluation Window
Once your benchmarks are set, the next question is whether the ad has actually had enough time and spend to be judged fairly. This is where a lot of money gets wasted, not on bad ads, but on good ads that get killed before they have a chance to prove themselves.
Meta's algorithm goes through a learning phase every time you launch a new ad or make a significant edit to an existing one. During this period, the system is actively testing delivery across different audiences, placements, and times of day to figure out who is most likely to take your desired action. Performance during the learning phase is inherently unstable. Costs fluctuate, results are inconsistent, and the data does not yet reflect how the ad will actually perform at scale.
You can see the learning status directly in Ads Manager under the Delivery column. If it says "Learning," the algorithm has not yet gathered enough data to optimize reliably. Meta's own guidance suggests that an ad set needs roughly 50 optimization events before the learning phase completes. For most advertisers running conversion campaigns, that means 50 purchases, leads, or whatever event you are optimizing for.
The practical implication: do not make a kill decision based on early data. A general rule of thumb used by many performance marketers is to wait until an ad set has spent at least one to two times your target CPA before drawing any conclusions. If your target CPA is $30, you want to see at least $30 to $60 in spend before evaluating the ad's performance.
Audience size also matters here. Smaller, more targeted audiences tend to stabilize faster because the algorithm has fewer variables to test. Broad audiences or interest-based audiences with millions of potential users can take longer to settle into consistent delivery patterns.
Tip: Set a calendar reminder when you launch a new ad to review it after it hits your minimum spend threshold. Not on day two. Not because you are anxious about the early numbers. After the threshold.
Common pitfall: Pausing ads during the learning phase and then restarting them resets the learning process entirely. Every time you do this, you are starting from zero and forcing the algorithm to re-learn delivery from scratch. This is one of the most expensive habits in paid social advertising.
Success indicator: You only evaluate ads that have cleared your minimum spend threshold and are no longer showing "Learning" status in Ads Manager.
Step 3: Pull the Four Metrics That Actually Tell You an Ad Is Failing
Now that you have your benchmarks and you know the ad has had a fair window, it is time to look at the actual performance data. There are four metrics that matter most for this decision, and the key is reading them together rather than in isolation.
CPA (Cost Per Acquisition): This is your primary efficiency signal. Compare your actual CPA directly against the target you set in Step 1. If your CPA is at or below target, the ad is working. If it is running above target, you need to understand why before deciding what to do. CPA alone tells you whether the ad is profitable, but it does not tell you where the problem is.
ROAS (Return on Ad Spend): ROAS gives you the revenue picture. An ad might technically have an acceptable CPA but still be generating below your break-even ROAS if average order values are lower than expected. Check ROAS against your break-even threshold. If the ad has been running for 7 or more days and ROAS is consistently below break-even, that is a serious flag.
CTR (Link Click-Through Rate): CTR tells you how well your creative is resonating with the audience seeing it. A strong CTR means the ad is stopping the scroll and generating interest. A declining CTR over time, especially when paired with rising frequency, is a reliable signal of creative fatigue. Compare CTR against your internal benchmark from Step 1.
Frequency: Frequency measures how many times the average person in your audience has seen your ad. As frequency climbs, CTR typically drops and CPMs often rise because you are reaching the same people repeatedly. There is no single frequency number that works for every account, but the pattern to watch is rising frequency paired with declining CTR and worsening CPA. That combination tells you the audience is exhausted.
Here is how to read these metrics together. A high CTR but poor CPA is a classic signal that the ad itself is compelling but something downstream is breaking down, usually the landing page. The ad is doing its job; the conversion path is not. In that case, turning off the ad is the wrong move. Fixing the landing page is the right one.
Conversely, a low CTR with high frequency and a rising CPA points to creative fatigue. The audience has seen the ad too many times and is tuning it out. That ad needs to be replaced with fresh creative.
Tip: Sort your ad sets by CPA descending in Ads Manager to instantly surface your worst performers at the top of the list. This saves time and keeps your review focused on the ads that need attention most.
Common pitfall: Focusing only on CTR or only on spend without connecting to downstream conversion data. CTR is a creative metric. It tells you about attention, not revenue. Always trace the full funnel before making a call.
Success indicator: You can look at any active ad and immediately identify which of the four metrics is the primary failure signal, and what that signal suggests about where the breakdown is occurring.
Step 4: Apply the Turn-Off Decision Criteria to Each Ad
You have your benchmarks. You know the ad has had a fair window. You have pulled your four core metrics. Now comes the actual decision. This is where a clear decision framework eliminates the guesswork.
Think of it as a tiered system with three buckets: turn off, watchlist, or scale.
Turn off immediately when an ad meets these criteria: CPA is more than 2x your target AND the ad has cleared your minimum spend threshold. This is your clearest signal. If you need a $30 CPA and the ad is delivering at $65 after $60 in spend, there is no scenario where waiting longer makes sense. The ad is not finding its audience at an efficient cost.
Additional criteria for immediate turnoff: ROAS has been below your break-even threshold for 7 or more consecutive days, CTR has dropped below your minimum benchmark after clearing the learning phase, or frequency has climbed above 4 with a simultaneous decline in CTR and worsening CPA. Any one of these, combined with cleared minimum spend, is sufficient reason to turn the ad off.
Watchlist for 48 hours when an ad is underperforming but has not clearly crossed your thresholds. Maybe the CPA is 1.5x your target and frequency is rising but not yet critical. Put it on a watchlist, set a reminder for 48 hours, and check again before acting. Some ads recover as the algorithm continues to optimize. Others slide further. Giving it 48 hours with a clear review date prevents both premature kills and prolonged waste.
Safe to scale when CPA is at or below target, ROAS is above break-even, CTR is holding at or above your internal benchmark, and frequency is still in a healthy range. These are your winners. They deserve more budget, not more scrutiny.
One important nuance: distinguish between pausing and turning off. Pausing an ad keeps its data and history intact in Ads Manager, which is useful when you want to review it further or potentially reactivate it with edits. Turning it off permanently is appropriate when you have diagnosed a clear failure and have no intention of running that specific creative or targeting configuration again.
Tip: Never make a kill decision during a sale, a major seasonal event, or immediately after a significant audience shift. External factors distort performance data in both directions. An ad that looks terrible during a slow holiday week might be perfectly healthy in normal conditions. Context matters.
Common pitfall: Applying blanket rules without accounting for campaign objective. An awareness campaign should never be evaluated on CPA. A conversion campaign should never be judged primarily on reach or impressions. Match your evaluation criteria to what the campaign was actually designed to accomplish.
Success indicator: Every active ad in your account falls clearly into one of three buckets: off, watchlist, or scale. No ambiguity, no ads running on autopilot without a status.
Step 5: Diagnose Why the Ad Failed Before You Replace It
Turning off a failing ad is the easy part. Understanding why it failed is what separates advertisers who keep improving from those who repeat the same mistakes with different creative.
Before you cut an ad, take 60 seconds to document the failure reason. This single habit will make every future campaign smarter because you are building a record of what does not work for your specific audience and offer.
There are three primary failure categories to work through.
Creative fatigue is the most common. The signals are rising frequency, declining CTR, and worsening CPA over time despite no changes to targeting or budget. The audience has seen the ad too many times and stopped responding. The fix is fresh creative, not a new audience. If the original creative worked before fatigue set in, the concept is sound. Test a new visual or hook while keeping the underlying message.
Audience mismatch shows up differently. Here, CTR is low from the start, not declining over time. The ad never resonated because it was shown to the wrong people. You might also see low engagement and poor relevance signals in your ad feedback scores. The fix is audience refinement, not necessarily new creative. Take the same ad and test it against a different interest group, lookalike audience, or demographic segment before writing off the concept entirely.
Landing page or offer problems are the trickiest to catch because the ad itself looks healthy. High CTR, reasonable frequency, but conversion rate falls off a cliff after the click. The ad is doing exactly what it should. Something between the click and the conversion is breaking down. This could be page load speed, a misaligned headline between the ad and the landing page, a weak offer, or a confusing checkout flow. Turning off the ad here would be a mistake. The ad is not the problem.
Tip: Before you turn off any ad, take a screenshot of the metrics and tag the failure reason in a simple tracking sheet. Even a basic spreadsheet with columns for ad name, failure date, failure category, and notes is enough. Over time, this document becomes one of your most valuable assets because it shows you patterns in what consistently fails for your account.
Common pitfall: Replacing a failed ad with a nearly identical creative and expecting different results. If the original ad failed due to audience mismatch, a new visual with the same targeting will likely fail for the same reason. Diagnose first, then decide what to change.
Success indicator: Every ad you turn off has a documented failure reason. You are not just cutting ads; you are building institutional knowledge about what works and what does not for your specific business.
Step 6: Build a Repeatable Weekly Review Routine
A decision framework only works if you actually use it consistently. The biggest mistake advertisers make is reviewing ads reactively, after a bad week, after the budget spikes, after someone panics about ROAS. By then, you have already lost money you did not need to lose.
The fix is simple: schedule your ad review the same way you would schedule any other recurring business task. Pick one day each week, block 30 to 60 minutes on your calendar, and run every active ad through the four-metric filter you built in Step 3. Doing this on the same day each week also means you are always comparing consistent time windows, which makes trend identification much easier.
Inside Ads Manager, set up custom columns so your four key metrics are visible the moment you open the dashboard. Add columns for CPA, ROAS, link CTR, and frequency. Remove the default columns that clutter the view with vanity metrics. Save this as a custom column preset so it loads automatically every time. This alone cuts review time significantly because you are not hunting for the numbers that matter.
Between manual reviews, Meta's automated rules can serve as an early warning system. You can set rules to send you a notification, pause an ad, or adjust a budget when a specific threshold is crossed. For example, a rule that notifies you when any ad set's CPA exceeds 2x your target over a 3-day window gives you a signal to investigate without waiting for your next scheduled review. The key is setting accurate thresholds. Automated rules with poorly calibrated numbers create more noise than signal.
For teams managing larger volumes of campaigns, manual review at this level of detail becomes genuinely difficult to sustain. This is where tools like AdStellar change the equation. AdStellar's AI Insights continuously monitors every creative, audience, headline, and landing page against your actual performance benchmarks, ranking them by ROAS, CPA, and CTR in real time. Instead of spending your weekly review hunting for underperformers, you open a leaderboard that has already done the sorting for you. The Winners Hub surfaces your best performers so you can instantly pull them into your next campaign. The AI Campaign Builder uses that performance history to build new campaigns around what has already proven to work.
Tip: Do your weekly review at the start of the week rather than the end. This gives you the full week ahead to act on what you find, test replacements for ads you turned off, and let new ads begin accumulating data before your next review cycle.
Common pitfall: Doing ad reviews reactively after a bad week instead of proactively on a fixed schedule. Reactive reviews are emotional reviews. You are looking at the data while already feeling the pressure of poor results, which makes it harder to apply a calm, criteria-based process.
Success indicator: You have a recurring calendar block for ad review, a saved custom column preset in Ads Manager, and a documented process that anyone on your team can follow in your absence.
Putting It All Together
Knowing which Facebook ads to turn off comes down to running every active ad through the same filter, every week, without exception. Here is the full checklist in sequence.
Define your CPA target, ROAS target, and minimum CTR benchmark before you evaluate a single ad. Confirm the ad has cleared your minimum spend threshold and is no longer in the learning phase. Check CPA, ROAS, CTR, and frequency together, not in isolation. Apply your decision criteria to place every ad into one of three buckets: off, watchlist, or scale. Document the failure reason before you cut any ad. Repeat this on a fixed weekly schedule.
That process, done consistently, will stop the slow budget bleed that most advertisers never fully trace back to its source.
If running this process manually across dozens of active campaigns sounds like a full-time job, that is because it can be. AdStellar was built specifically to handle this at scale. The AI Insights feature ranks every creative, audience, and headline by real metrics like ROAS and CPA, scores them against the benchmarks you define, and surfaces your winners automatically. You always know what to cut and what to scale without spending hours in Ads Manager every week.
Start Free Trial With AdStellar and put your ad evaluation on autopilot. Stop guessing which ads to kill. Start running your budget on data.
