YOUR META ADS FEEL MORE EXPENSIVE AND LESS EFFECTIVE. HERE'S WHY.

You set up the campaign. You checked the targeting twice. You wrote copy that felt good. Hit publish. Then finally, your budget drained with almost nothing to show for it.

You're not losing your mind. Meta ads really are harder to make work than they used to be, and the costs are going up along with every SAAS platform trying to charge you ‘$5 a week’ for nothing (sorry, I am sick to DEATH of that also!). The frustrating part? Most of the advice you'll find either blames you for doing something wrong or promises a secret hack that fixes everything.

I am warning you, there is no secret hack that fixes everything and there never will be. If anyone is selling you on this, they are trying to sucker you out of your hard-earned money, emotion and time to sell you an e-book, course or some new alternative method of BS.

What's actually happening is a combination of structural market shifts, platform-level changes, and some things that are within your control to improve. This isn't a definitive fix-it guide. Ad performance is a moving target that takes time, iteration, and real data from your specific account to crack. But understanding the full picture of what information is available now? That's where your journey restarts.

THE MARKET ITSELF GOT MORE EXPENSIVE

Before we even get into your account, it helps to understand what's happening at the platform level. A lot of advertisers are burning money troubleshooting their creative and targeting when the actual issue is the auction itself.

The numbers are pretty telling. In Q1 2025, Meta's average CPM (Cost per 1,000 Impressions) hit $10.88. The highest Q1 figure recorded since at least 2021. January alone came in at $10.48, up 26.8% from the same month in 2024. March pushed nearly to $12.

Those aren't small bumps. And according to benchmark data from Triple Whale tracking over 6,000 companies, the median CPM across Meta in 2025 landed at $13.48, with a median CPA of $38.17. I combed through thread after thread in Reddit with business owners and professional advertisers sharing the same frustrations in less PG words.

Some of that is simple supply and demand. More advertisers are competing for the same impressions every year, and Meta's auction system prices accordingly. During peak periods (Black Friday, Cyber Monday, Q4 in general) CPMs can spike 25–66% above annual averages. That's not a bug. It's just how auctions work when everyone wants the same eyeballs at the same time. So just a note, if you’re wanting to spend marketing dollars around this time, budget accordingly. As much as it hurts.

Part of it is also infrastructure cost being passed down. Meta has committed to spending between $135 billion in 2026 on AI and data centers. I know I don’t care for it, I don’t want it, but so be it. That build-out isn’t anything close to cheap, and advertisers are effectively footing part of the bill through rising ad prices. So, when your CPM creeps up and you can't point to a clear reason why. Here’s the big, fat and disgustingly ugly why.

None of this means Meta ads aren't worth it. Triple Whale data shows that brands still invest 68% of their total ad budget on Meta, and it continues to drive more conversions than most other platforms. But the era of cheap, easy Meta traffic is over, and the sooner you factor that into your expectations and budget planning, the better.

APPLE CHANGED THE RULES AND META IS STILL ADJUSTING

If you've been running Meta ads since before 2021, you probably felt a sharp performance drop at some point and couldn't fully explain it. That was Apple's App Tracking Transparency (ATT) update landing. And four years later, its effects are still very much present.

Before ATT, Meta's ad system thrived on cross-app tracking data. It knew what you browsed, what you almost bought, what you watched. That data fed its targeting and attribution engine in ways that made ads feel scarily precise. ATT gave iPhone users a pop-up that essentially said: do you want to be tracked? About 75% of iOS users opted out.

The downstream effects were significant. Meta's machine learning had less data to work with, which meant weaker targeting. Retargeting pools shrank. Attribution windows got shorter (now defaulted to seven days for clicks and one day for views). This means that conversions that happen outside that window simply don't show up in your reporting. A purchase made eight days after someone clicked your ad? Invisible. Research published by UCLA's Anderson School of Management found that firms highly dependent on Meta saw overall revenue drop by 37.1% following ATT's introduction, compared to those less reliant on the platform.

There's also a measurement problem that doesn't get talked about enough. In August 2025, a whistleblower case at the Central London Employment Tribunal (reported by the Financial Times) alleged that Meta inflated the performance of its Shops Ads by using gross revenue figures (including shipping and taxes) rather than net sales, reportedly inflating results by up to 19%. The case is still working through the courts, but it's a useful reminder that the numbers in Ads Manager don't always tell the complete story. Cross-referencing your Meta data with Google Analytics 4 and adding UTM parameters to every ad gives you a much more honest picture of what's actually working.

If you run ads, you also may notice that they are constantly testing new AI features for you and offering statistics like “Most advertisers like you saw a X% increase using Y feature. Are you sure you want to turn it off?” In some cases, they ask you why you aren’t using this feature as a mini-poll option. Do not let these things fool you. Do not give Meta full free will with these ad features. Choose and test strategically and again, let your own data do the talking. I also feel inclined to mention that Meta can, and will, turn these features on in the background without your consent. But these are the risks you run, and careful oversight is your best friend.

The short version: your ads might be performing better than Meta is showing you, or the results might be getting inflated. Either way, one dashboard isn't enough.

MEET ANDROMEDA — META'S AD ENGINE BASICALLY GOT REBUILT

Here's where things get interesting, and also where a lot of advertisers are unknowingly leaving performance on the table.

In late 2024 and into early 2025, Meta rolled out a system called Andromeda–its next-generation ad retrieval engine. By October 2025, it was fully deployed across all active advertisers. This isn't a minor feature update. According to Meta's own engineering blog, Andromeda represents a 10,000x increase in the complexity of models used for ad retrieval. It’s the first step in the process where Meta narrows tens of millions of available ads down to the few thousand it actually considers showing to any given person.

What that means practically: the old system had a bottleneck. It couldn't efficiently process the explosion of ad variations created by AI tools and Advantage+ automation. Andromeda fixes that bottleneck using NVIDIA chips and Meta's own proprietary AI hardware, enabling it to process far more creative options, map users and ads in a shared mathematical space, and make better real-time matches between who's on the platform and which ad is most likely to land.

Meta's reported results from the rollout: a 6% improvement in retrieval recall, an 8% improvement in ad quality on selected segments, and up to a 22% higher ROAS for campaigns using Advantage+ Creative.

But here's the part that actually changes how you should be running ads.

CREATIVE IS NOW THE TARGETING

Under the old system, you told Meta who to reach through interest stacks, behavioral targeting, and tightly defined custom audiences. The audience was the lever. Under Andromeda, that lever has shifted to creative. The algorithm is now sophisticated enough to determine who your audience is based on the signals in your ad — the visuals, the hook, the language, the offer — and find those people itself.

As Jon Loomer, one of the most cited independent Meta advertising analysts, wrote after the Andromeda rollout: the shift is from "niche targeting to creative diversification as the best lever to find the most relevant audiences." Manual interest stacks and lookalike audiences haven't disappeared, but they increasingly act as starting suggestions rather than hard constraints. Andromeda will expand beyond them if the creative gives it reason to.

There's also a technical layer to this called Entity IDs. Andromeda groups ads it perceives as similar into the same retrieval cluster, meaning if you upload ten variations of the same ad with minor copy tweaks or slightly different backgrounds, the system may treat all ten as a single ad. One ticket to the auction, not ten. The system is pattern-matching visuals, not just text, so ads with the same creator, same environment, and same general look can get grouped even if the words are different.

The strategic implication: creative diversity now means conceptual diversity. Different benefit angles, different personas, different formats, different emotional hooks. Meta's own research found that different creative motivators unlock new audiences 89% of the time. That figure isn't a reason to panic. It’s useful, actionable information. It means a single strong creative concept, executed well, can reach a lot of people. But if your whole ad account is running variations of the same message, you're leaving a lot of potential reach untapped.

AD FATIGUE IS HITTING FASTER THAN IT USED TO

This one isn't just anecdotal. Nielsen's 2025 Digital Ad Effectiveness Report found that ads lose impact up to 35% sooner in algorithm-driven campaigns than in more manually structured ones. When Meta optimizes aggressively for performance (which is what it does to get you out of the learning phase) it tends to show your best-performing ad to the most receptive people quickly and repeatedly. The speed of the win accelerates the burnout.

What makes this harder to catch is that the early warning signs have changed. Frequency numbers can look perfectly normal right up until performance drops. According to Meta's own updated advertiser guidance from late 2025, degradation increasingly shows up in declining conversion efficiency before it shows up in click-through rate. So, your CTR holds steady, your frequency looks fine, and then your CPA quietly climbs until suddenly you're paying way too much for way too little.

WARC's 2025 research on creative effectiveness  showed that ads optimized purely for short-term performance underperform by up to 40% over longer horizons compared to campaigns built around multiple rotating creative concepts. Which is a pretty compelling argument for having a creative refresh system, not just a winning ad.

THE METRICS THAT ACTUALLY TELL YOU WHAT'S GOING ON

Before jumping to fixes, you need to know what you're actually looking at. A lot of people use Meta Ads Manager as their single source of truth — which, given everything above, is risky. Here's how to read the signals more accurately.

CPM (COST PER 1,000 IMPRESSIONS)

This tells you how competitive the auction is for your audience. A rising CPM doesn't always mean something is wrong with your ads. It might just mean more advertisers are targeting a similar audience. In the US, average CPMs in 2025 are running around $13–20 depending on industry, with health and wellness at the high end. CPM is a contextual metric. Don't obsess over it in isolation, but do watch it for sudden spikes that don't correlate with seasonal periods. That's often a sign your targeting is competing too narrowly.

CTR (CLICK-THROUGH RATE)

A platform-wide median CTR for Meta in 2025 sits around 2.19% according to Triple Whale data. CTR is primarily a creative feedback signal. It’s going to tell you whether your ad is stopping people or not. A CTR below 1% generally means the ad isn't resonating at the surface level. That might be the hook, the image, or a mismatch between the creative and the audience seeing it. High CTR with poor downstream performance points to a different problem: the ad is interesting, but the landing page or offer isn't following through.

CPA (COST PER ACQUISITION) AND ROAS

These are the business metrics. CPA tells you what it's costing to get one conversion. ROAS tells you what you're getting back for every dollar spent. The 2025 median CPA across industries is $38.17. Newsflash: that number is largely meaningless without context. A $38 CPA is incredible if you're selling a $300 product and a disaster if you're selling a $40 one. ROAS benchmarks typically aim for 2:1 to 3:1 for most businesses, though e-commerce with thin margins often needs higher. One thing to remember: because of the attribution issues mentioned above, your actual ROAS is probably a bit better than Meta is reporting. That's not a license to ignore the numbers. It is another reason to have more than one reporting source.

HOOK RATE AND HOLD RATE (FOR VIDEO)

If you're running video ads (under Andromeda you really should be testing them) these two custom metrics are worth setting up manually. Hook rate is three-second video plays divided by impressions, and the benchmark to aim for is around 25%. Hold rate (how many people watch past 25% of the video) also benchmarks around 25%. Low hook rate = weak opening. Low hold rate = the ad started okay but lost people. Both are fixable with creative adjustments, and they tell you exactly where to look.

WHAT TO ACTUALLY DO ABOUT IT

None of what follows is a guarantee. These are directional moves that make your account better positioned to work with the current system. Do not think of this as a formula that bypasses the testing and iteration that paid advertising genuinely requires. Time, patience, and budget to do so is your best friend here.

RETHINK WHAT 'CREATIVE VARIETY' MEANS

Stop thinking of creative variations as the same ad with a different headline or button color. Under Andromeda, those are the same ad. True creative diversity means different angles.

For example:

• One ad addresses a pain point

• One ad leads with social proof

• One ad explains the product with a founder's voice

• One ad uses humor

Depending on your budget, length and target market, aim for 3-4 genuinely distinct creative concepts per campaign, not 10 variations of one concept. The goal is to give the algorithm different tickets to the auction, not copies of the same one.

SIMPLIFY YOUR ACCOUNT STRUCTURE

The old playbook was:

• Lots of ad sets

• Tight targeting

• One or two winning ads per set

Andromeda is built for the opposite. Multiple overlapping ad sets with narrow interest targeting fragments the data the algorithm needs to learn. Consolidated campaigns with broader targeting and diverse creative perform better because they give the system more to work with. If you're running 10+ ad sets with narrow interest stacks, that's likely a structural issue worth addressing.

GIVE THE LEARNING PHASE ROOM

This one's hard but important: resist the urge to make changes every few days. Frequent edits reset Meta's learning phase and interrupt the pattern recognition Andromeda relies on. The Search Engine Land reporting on Meta's combined Andromeda and GEM systems (GEM is a separate ranking layer that runs alongside Andromeda and is reportedly 4x more efficient at driving ad performance gains than previous models) recommends setting a minimum no-touch window. At least a week, or 50–75 conversions, whichever comes first. Early volatility is normal. It doesn't always signal failure.

START BUILDING FIRST-PARTY DATA

Given the ongoing signal loss from iOS, your own customer data is increasingly valuable. Examples of first-party customer data:

• Email lists

• CRM data

• Purchase history

These can be uploaded to Meta to create custom audiences that aren't dependent on pixel tracking. Meta's Conversions API (CAPI) is also worth setting up if you haven't already. It sends conversion data directly from your server rather than relying on the browser-based pixel, which dramatically improves tracking accuracy for users who've opted out of iOS tracking.

DON'T READ YOUR RESULTS FROM ONE PLACE

Add UTM parameters to every ad and run them through Google Analytics 4 alongside your Ads Manager data. This triangulation approach: Meta reporting, your analytics and your actual backend sales data gives you a far more accurate read on what campaigns are driving. The goal is to stop relying on one platform's version of what's working.

THE PART NO ONE WANTS TO HEAR

There's no single fix that's going to make your Meta ads instantly cheaper and more effective. The people who are seeing strong results right now (and they do exist) are advertisers who've adapted to Andromeda. Reportedly, they’re seeing 8–17% conversion increases while cutting costs by up to 16% according to Infinity Nation's analysis. These businesses not running a secret playbook. They've put in the reps. They've tested creative concepts across different angles. They've simplified their account structure. They've given campaigns room to learn. They've stopped treating Ads Manager as the whole truth.

The mindset shift that moves the needle is going from "why isn't this working" to "what is this data telling me to try next."

CPM spiked for no seasonal reason? Your audience targeting might be too narrow.

CTR is solid but CPA is too high? The problem is probably post-click, which means it’s either your landing page or your offer.

Multiple creative versions tanking at the same time? You might be dealing with creative fatigue and need new angles, not new variations.

Meta advertising in 2026 is less of a plug-and-play channel and more of an ongoing practice. I don’t view it as a bad thing; it just is a shifted approach. Brand marketers who treat it like a craft tend to outperform the ones who expect it to be automatic.

If Meta advertising feels like a puzzle you're tired of solving alone, that's what I’m here for. Whether you're starting from scratch or trying to figure out why your current campaigns aren't pulling their weight, reach out. We can dig into it together. Head to spill-social.com to get in touch.