Facebook Ads Not Working? Find Which of 4 Problems You Have

Facebook ads not working falls into four stages, and the stage where your numbers stop names the problem. In Ads Manager, Meta's tool for building and monitoring ads, no impressions means a delivery block. Impressions without clicks means the ad earns no attention. Clicks without sales means a broken step after the click. Good results that faded mean a learning reset, a small budget, or fatigue.

Work through the table from top to bottom. An ad that is not delivering tells you nothing about its creative, and a store that cannot record a sale makes every ad look bad.

An ad that is switched on but not delivering is blocked by something outside the creative. Ads Manager prints the reason in smaller text under the delivery status, so read that line first. Then run five checks, each with a clear pass or fail, using Meta's delivery troubleshooting guidance as the reference.

If all five pass and impressions are still zero, the next suspect is an audience so small that the system can find almost nobody. Widen the location or interest targeting before touching anything else.

A CTR near 1% or lower after about 4,000 to 7,000 impressions shows the ad is failing to earn the click. That threshold is our recommendation. One vendor's 2026 averages, from Lebesgue's March 2026 benchmarks, run from 1.94% (dogs, cats and pets) to 2.92% (art and home decor) across 11 industry categories. Those figures come from an undisclosed sample, so use them only as a rough yardstick.

How much spend gives a fair reading? The same vendor's cost per 1,000 impressions (CPM) runs from $6.96 to $12.46. At those CPMs, $50 buys about 4,000 to 7,000 impressions ($50 ÷ $12.46 × 1,000 = 4,013; $50 ÷ $6.96 × 1,000 = 7,184, rounded). At a 2% CTR, 4,000 impressions produce about 80 clicks. Below that, a few clicks more or less move the rate a lot.

When CTR is the problem, change one thing: the first line of copy or the first seconds of video, with the offer held constant. Changing the hook, offer and format together tells you nothing about which one mattered. Our page on Meta hook testing: hook signals versus conversion data explains how far an early hook signal can be trusted. Do not touch the audience or budget in the same edit, because those changes can restart learning (see the decline section).

Compare your clicks with the number at which zero sales becomes unlikely. For 332 clicks at a 0.90% conversion rate, the chance of zero sales is under 5%. If you are past that number, check tracking, checkout and the landing page, in that order.

The odds check. Conversion rate here means purchases divided by clicks. Use your own rate from the last 30 days. If your true rate is r, the chance of zero sales in n clicks is (1 − r)^n, and the number of clicks after which zero sales drops below 5% is ln(0.05) ÷ ln(1 − r), rounded up. This is our model. It assumes each click is independent, so treat it as a guide, not a verdict.

The table below uses Lebesgue's March 2026 conversion rates as illustrative inputs. They are one vendor's averages, the sample size is undisclosed, and the definition of conversion rate may differ from yours. Your own rate replaces them. Suppose a store gets 600 clicks over five days and no sales:

In clothing, 600 clicks and no sale is a 1-in-227 event, so something is broken. In hardware, the same result is about a 1-in-9 event, so more data is needed first.

A second check uses spend. Lebesgue's March 2026 customer acquisition cost (CAC, the average cost of winning one customer) runs from $20.47 to $39.03 across its categories, and we use it here as a stand-in for cost per purchase. Suppose a store spent $800 on Facebook ads in a month. At those costs it would expect roughly 20 to 39 purchases ($800 ÷ $39.03 = 20.5; $800 ÷ $20.47 = 39.1). A month with none points to a fault, not a weak ad. Your own cost per purchase is a better input than any benchmark.

If the odds say "broken," check three things in this order:

A drop after a few good days usually comes from one of three causes: an edit that restarted learning, a budget too small to finish learning, or an audience that has seen the ad too often. Check them in that order, starting with the date of the last significant edit in the ad set.

As of October 2026, Meta's learning phase guidance ties the end of learning to about 50 optimization events in 7 days. Meta labels an ad set Learning limited when it does not expect to reach that count. The ad set still delivers, but unevenly. A significant edit can restart learning: pausing the ad set, or changing the optimization event, audience or creative. Budget or bid changes can too, depending on their size.

The budget formula. Daily budget per ad set = 50 purchases × cost per purchase ÷ 7 days. This is our model. Use your own cost per purchase from the last 30 days as the lead input. The table below shows the result when Lebesgue's March 2026 CAC, an undisclosed-sample vendor average, stands in for cost per purchase. Read it as an illustration of the range, not as a target for your account. For the wider question of what a full month of Meta spend and testing needs, see our guide to Meta ads cost and creative testing.

Suppose a store with a $34.29 cost per purchase spends $25 a day. That buys about 0.73 purchases a day, or 5.1 a week ($25 ÷ $34.29 × 7), about 10% of the 50 needed. That ad set will sit in Learning limited. The same math gives about 4 purchases a week at $20 a day and about 1 at $5.

If you cannot fund the formula, you have three options, and these are our recommendations. Run fewer ad sets, because the requirement applies to each one separately. Optimize for a higher-funnel event such as add to cart, which happens more often than purchase. Or hold every setting still for a full 7 days and judge on the data you have.

Zero add-to-carts alongside zero sales points to tracking or the site, not creative. Suppose a store gets 3 purchases in its first 3 days, then 4 days with none. At the first three days' pace of one a day, four empty days would happen about 1.8% of the time (e^−4). That figure rests on a three-day sample, so it only says "check tracking first."

If the last edit and budget check out, look at fatigue. Compare CTR and frequency, the average number of times each person has seen the ad, week over week. Frequency rising while CTR falls means the audience has seen the ad too often. Our post on why ad creatives lose performance and how to fix creative fatigue covers the fix.

Once delivery, tracking and budget pass, test one change, chosen from the stage that failed, and hold everything else still for 7 days. That is our recommended method:

Our Creative Intelligence platform combines ad-performance data, competitor activity, customer signals and past ad learnings into a ranked weekly test plan, delivered every Monday, with execution-ready creative briefs. Deepsolv is sold through a personalized demo with our founding team, where you can ask about pricing: book a demo.