Why did my ads stop working?

Is the drop even real?

Before you diagnose anything, check whether the platform is actually measuring correctly.

A reporting drop and a real performance drop look identical in the dashboard. The number goes down and the account signals a problem. But sometimes the number went down because the platform stopped counting correctly, not because fewer people bought. A broken measurement system produces a fake crisis.

Three checks take about ten minutes and answer this before you touch anything else.

Compare platform conversions against your store or CRM for the same time window. Pull the number of purchases your ad platform reported for the last 30 days. Pull the actual orders from your store backend or CRM, your customer database, for the same window. If the platform shows substantially fewer sales than actually happened, you have a measurement gap. The ads may be fine. The tracking is not.

Check that the attribution window has not changed between the two periods you are comparing. An attribution window is the number of days after a click or view that the platform will still credit a sale to that ad. If you are comparing a period when the window was set to seven days against a period when it was set to one day, you are not comparing the same thing. The narrower window produces a smaller reported number without any change in actual sales.

Check that your pixel still fires after any recent site update. A pixel is the small tracking code on your site that tells the ad platform when a purchase happens. Theme updates, plugin changes, and checkout redesigns can break the purchase event silently. The platform stops receiving the signal. Sales keep happening. The reported number drops.

Platform behavior also shifts without notice on your end. Meta's Andromeda, a change to how Meta's system decides which ads to show people, announced in late 2024 and described in a March 2025 Meta engineering post, made that selection lean much harder on machine-learning models. Industry analyses since then link some unexplained delivery shifts to it. The point is not to diagnose Andromeda specifically. It is that the platform you are measuring on is itself changing, which is one more reason to confirm the measurement before blaming the account.

No one outside your account can look at your dashboard and tell you whether the drop is a measurement problem or a real performance problem. That distinction lives inside your data. The comparison check above answers it in ten minutes.

Was it creative fatigue?

Creative fatigue is real, but it is narrower than most people think. It means your audience has seen the same ad enough times that they have stopped registering it. The signal to watch is frequency, the average number of times one person has seen a given ad.

There is no universal threshold where fatigue kicks in, and Meta does not publish one. The Trade Desk's published guidance on ideal frequency says the cap is campaign-specific: you find the point where serving the ad more times stops adding results, and that point is different for every campaign and audience.

What matters more than any threshold is the pattern fatigue produces. Your click-through rate (the percentage of people who click after seeing the ad) falls. Your cost per click rises. And CPM, the cost to show your ad to a thousand people, may stay flat or drift upward as the platform reads weaker engagement.

If your CPM has spiked but your click-through rate is holding, fatigue is not the explanation. Something else changed in the auction. The platform is paying more to reach your audience, or your audience has shifted, or a structural problem further down is making the whole system look like a creative problem from the top.

Swapping creative addresses fatigue specifically. It does not address any of the other reasons performance declines. This is why new creative sometimes helps for a few days and then the numbers fall again. The creative was not the root cause.

What actually changes when nothing changed?

This is the harder question. You did not touch anything. The ads look the same. The audience is the same. So why did performance drop?

Four things can shift without you doing anything.

The demand level in your category changes. Demand for most products is not flat. It moves with seasons, with news cycles, with what people are currently worried about or excited about. A skincare product that rode a trend in one quarter may face a genuinely smaller pool of eager buyers six months later. The product did not change. The size of the currently-ready audience did. Your CPM rises because fewer people in your target audience are in the right mental state to respond.

The competitive noise floor rises. Other brands enter or increase spend in your category. They run ads aimed at the same audience, making the same kinds of promises. The buyer has now heard your type of message more times, from more sources. This is called market sophistication: how worn out a buyer is from hearing similar claims in a category. When sophistication rises, the same headline that used to stop the scroll becomes invisible. The ad still runs. The message just no longer registers as new. NCSolutions (Nielsen Catalina Solutions), in its long-running ad effectiveness research, found that the creative itself drives roughly 49% of a campaign's incremental sales, more than targeting or reach. When the noise floor rises and your message is no longer differentiated, that advantage quietly goes to zero even though the physical ad did not change.

The offer drifts out of match. This one is subtle. Your offer is everything the buyer is actually being asked to say yes to: the price, the promise, the guarantee, the friction of the checkout process, and whether the stated outcome still matches what buyers currently want. Markets move. A promise that felt compelling eighteen months ago may now feel ordinary because competitors have made the same promise, or because buyer priorities shifted. The offer did not change. The context around it did.

The path from click to purchase leaks. Your landing page, your checkout flow, your post-click sequence. These can degrade without anyone touching them. Page speed worsens after a site update. A plugin breaks a form field on mobile. A price test left a price inconsistency between the ad and the page. Conversion rate (the percentage of people who take the desired action after clicking) drops while everything upstream looks fine. From the ad account, this looks like the ad stopped working.

How do I tell if it is the creative or something underneath?

You look at which number broke first.

Ad accounts have three main handoff points you can read directly. CPM is what you pay to show the ad to a thousand people: the auction layer, mostly about demand and competition. CTR (click-through rate) is what percentage of people who saw the ad clicked it: the creative and message layer. CVR (conversion rate) is what percentage of people who clicked actually bought or completed the desired action: the offer and post-click layer.

Read them in order.

If CPM rose but CTR held: the problem is in the auction. Demand shifted, competition increased, or your targeting now reaches people who are less ready than they were. New creative will not fix this.

If CPM is flat but CTR fell: now creative is the suspect. The message is not cutting through. This is where testing new angles, new hooks, or new proof points is the right move.

If CPM and CTR are both fine but CVR dropped: the problem is downstream. The landing page, the offer, or the checkout path broke. No amount of ad-side work will fix a post-click problem.

Most accounts show a messy version of all three happening at once, which is why a single "swap the creative" response rarely solves the whole problem. You are treating one variable when two or three have moved.

For context on what normal looks like: WordStream's 2024 Facebook Ads benchmark report puts the average conversion rate across industries at 8.78%, with wide variance by vertical. But a sudden drop below your own historical baseline matters far more than any industry average. Your baseline is the benchmark worth chasing.

Why does new creative sometimes work for a week?

Because novelty has its own short-lived effect, and because the algorithm rewards new creative with an initial exploration phase.

When you launch a new ad, the platform does not immediately know who responds to it. It spends the first few days testing the ad against a broader slice of your audience than it would normally reach. This exploration period can produce better-than-normal numbers for a short time, lower CPMs and higher engagement, simply because the system is casting a wider net before it narrows.

Once the exploration phase ends and the algorithm settles into its learned audience, the structural problems reappear. If the underlying issue was demand, positioning, offer, or the post-click path, none of those changed during the creative swap. The brief improvement was real, but it was telling you something: the account has a deeper problem the creative swap didn't reach.

This is also why running the same ad to a new, cold audience sometimes re-ignites performance temporarily. You have moved the ad into a pool of people who have not developed fatigue. But if what the ad says no longer fits what that market wants to hear, the cold audience will also wear out, usually faster than the first one did, because you are not attracting a qualified response. You are just reaching more people less efficiently.

What should I check before the next test?

Before you build anything new, get clear on what actually moved.

Pull your account-level data for the 30 days before the drop and the 30 days after. Look at CPM, CTR, and CVR as separate lines, not just cost per result. Note which one broke first and by how much.

Check whether a real change happened outside the account. Did a competitor launch a major campaign or a heavily discounted offer in your category? Did you change the landing page, even a small update? Did your product pricing or shipping terms change? These are questions anyone running ads should be able to answer from their own records.

Check the landing page manually, on a real mobile device, on the connection your buyer actually uses. Forms break. Pages load slowly. Mobile layouts collapse. Things your desktop testing never caught.

Check whether the audience you are targeting has genuinely shifted. Interest-based audiences and lookalike audiences drift as the platform's underlying data changes. An audience that was tight and responsive six months ago may now include people who matched the signal for reasons that no longer hold.

None of these checks require a tool or a consultant. They require honesty about what you actually know versus what you assumed was stable.

Frequently asked questions

Why did my ads stop working after two weeks?

Two weeks is roughly when the initial exploration phase ends and when early frequency effects start to compound. If performance was strong in week one and fell in week two, the most likely causes are: the algorithm finished exploring and narrowed to a smaller, more expensive audience; frequency built quickly in a small audience; or the offer or post-click path had a structural problem that the initial novelty masked. Check CPM and CTR separately across the two periods to see which layer moved.

Do Facebook ads just stop working eventually?

Not in the way the question implies. The auction, the audience, and the competitive environment all change continuously. An ad that worked a year ago can fail today because the market has heard that message many more times since then, not because ads as a channel run dry. Ads are a match between a message and a moment in the market. When the moment changes, the message needs to change too.

Should I restart my ad campaign?

Restarting has a narrow use case: when you need the algorithm to re-learn the audience from scratch because targeting or creative changed substantially. It does not fix structural problems. If the offer is off, restarting shows the off-offer to a fresh pool. If the landing page is broken, restarting sends more traffic into a broken path. Restart after you have identified and fixed the underlying issue, not before.

Why are my ads working for some audiences but not others?

Different audiences are at different stages of wear. An audience that has seen many ads in your category is harder to move with the same message than one that has not. The message that works for a cold audience often fails for a more experienced one, not because of targeting mechanics but because the same words carry a different weight depending on how many times the person has already heard that type of promise.

My ads platform shows fewer sales than my store. Which is right?

Your store or CRM is right. The ad platform only sees sales it can attribute to a click or view within its attribution window, the number of days it looks back from a purchase to find an ad interaction. If that window is short, or if the tracking pixel broke after a site update, the platform undercounts. Your store records every completed order regardless of where the buyer came from. Start by checking whether the pixel still fires and whether the attribution window changed recently.

My cost per result doubled but I did not change anything. What happened?

A CPM spike with no account-side changes almost always points to the auction. Either competition for your audience increased, or the audience pool shrank, making each impression more expensive. The response is not to bid more. It is to reassess whether the audience you are targeting is still the right one, or whether your message needs to be different enough to compete at the new price.

Related reading: Why your ROAS problem isn't a creative problem · What the Realignment Hub teaches