Five months of Meta ads in 2026: What's Actually Working
If you’re running Meta ads yourself, paying an agency to, or thinking about either, you’ve been on a noisy ride this year.
January was a mess. Conversion numbers dropped overnight when Meta changed how attribution worked. Detailed targeting got hollowed out the same week. Forums lit up. Newsletters panicked.
That was January. It’s now May. We’ve run a full quarter of client work under the new rules, watched the dust settle, and formed actual opinions instead of guesses.
So this is the calm version, written from the other side. What we now know works in Meta ads in 2026, what’s worth ignoring, and what good ads should look like right now.
This is also how we do it at Base Brands.
A quick word on the January wobble
In January, Meta removed the 7-day view and 28-day view attribution windows. In March, they narrowed what counts as a click. The result was that reported conversions across most accounts dropped 15 to 40 percent overnight, even when nothing else changed.
A lot of advertisers panicked. Some cut budgets. Some sacked agencies. Some declared Meta broken and moved spend elsewhere.
Almost all of them were wrong. The actual sales were fine. The dashboard just got smaller.
That’s the only paragraph this article will spend on January. The rest is what we’ve learnt running through it.
Three real shifts we've now confirmed
After four months of running clients under the new rules, here are the three changes that have actually moved the needle.
1. The dashboard tells a smaller story now, and that’s permanent
The 7-day view and 28-day view attribution windows are gone for good. The 5-second engaged-view threshold is the new normal. Click-through means link clicks only.
What we’ve seen across our client accounts since January is consistent. Meta’s reported conversions track 20 to 35 percent lower. The gap is the attribution change, not the campaigns
What to do:
Get Conversions API running alongside your Pixel if you haven’t already. The Pixel will still work for you and track data, but it’s less reliable than ever before.
2. Detailed Targeting is functionally cosmetic now
Meta deleted dozens of interest categories on 15 January. The targeting that remains gets treated as a suggestion, not a filter. You tell Meta “women 30 to 50 who like running,” and Meta hears “start there, but find better buyers if you can.”
We’ve stopped using narrow detailed targeting on every account we run, with one exception (a niche B2B client where the interest genuinely is the audience). On every other account, broad targeting plus strong creative has either matched or beaten the targeted version since February.
What to do:
Spend energy on the data signals you feed Meta, the Pixel events, customer match lists, LTV data passed back through CAPI and on the creative. The system has the targeting handled. It needs you to teach it what a good buyer looks like once they convert, not who to show the ad to upfront.
3. Creative variation is the lever that moved most
This is the change that’s surprised us most. We knew it mattered going into the year. We didn’t expect it to be the single biggest performance driver for almost every account we’ve run since.
Through 2024 and most of 2025, creative testing on Meta meant “same ad, ten hooks.” Different opening lines, different on-screen text, same underlying concept. That worked because the algorithm could only handle so much variety in parallel.
Andromeda, Meta’s ad retrieval system, can hold a lot more in flight at once. It rewards genuinely different creative concepts. UGC testimonial against founder-to-camera against product demo against animated explainer. Different formats, different tones, different angles, all live at the same time. The system finds the right one for the right viewer.
We’ve moved every client account from “ten hook variants” to “three or four genuinely different concepts.” Cost per result has improved on every account where we’ve made that switch.
What to do:
Brief creative around concepts, not variations. If your last batch of ads was ten Reels with slightly different captions, that’s iteration. Variation is what the algorithm wants.
What good Meta ads look like in 2026
Whether you’re running ads yourself or paying someone else to, here’s the bar.
- A simple two-campaign structure. One for testing, one for scaling. Not five campaigns, not seven ad sets. Two. Andromeda needs consolidated budgets to learn quickly.
- Three to four genuinely different creative concepts in rotation at any time, with a defined kill rule. We use 48 to 72 hours or £100 spend, whichever comes first, with no signal.
- CAPI running alongside the Pixel.
- Broad targeting. Suggestions used sparingly. Lookalikes treated as inputs, not silos.
- A weekly review that looks at three things: cost per result against your benchmark, CPM trend (rising CPM means creative is fatiguing), and which concept is winning. Not what individual ad got the most clicks. Which concept.
- A 14-day creative testing cycle running in the background continuously, so the next batch of winners is always being prepared.
If your current setup looks roughly like that, you’re in good shape. If it doesn’t, that’s the gap.
Frequently asked questions
Do I still need detailed targeting? Treat it as a starting hint for Meta, not a constraint. Layer one or two broad interests if your offer is genuinely niche, and let the algorithm find buyers from there. Don’t stack ten interests hoping it narrows the audience. It won’t.
How much should I spend on Meta ads in 2026? Below £50 a day per campaign, the algorithm doesn’t get enough signal to optimise properly, but it will get there eventually. If you can’t sustain that though, Meta probably isn’t the right channel yet. Email, organic content, and warm referral all work better at lower budgets.
Is it worth running ads while I figure all this out, or should I wait? Don’t wait. The platform rewards consistency. Start with a simple two-campaign structure, three or four different concepts, broad targeting, and CAPI running. Refine from there.
How long until I see results? For a new account, 14 to 30 days before you have enough data to make decisions. Anything faster than that is guesswork dressed up as analysis.
Want this run properly?
If you’re running Meta ads yourself and the structure above doesn’t match what you’re doing, get in touch at basebrands.co.uk and we can talk through what’s missing.
If you’re already with an agency, we can do a calm second-opinion review of your account. No hard sell, no campaign trash-talk. Just an honest read on whether your setup matches what good looks like in 2026.
Either way, we’d love to hear from you.
The Conversation To Have With Whoever Runs Your Ads
If someone else is running your Meta ads, whether that’s a freelancer, an in-house hire, or an agency, this is the part worth raising with them.
Not as a test. As a proper conversation.
Questions worth asking:
- Where are we sitting against the 50 conversions per ad set per week threshold?
- If we’re under it, what’s our plan? Are we optimising for the real goal and managing expectations, or are we moving up the funnel and tracking the real goal separately?
- How much week-on-week variance should I expect as normal given our structure?
- What cadence of budget changes do you think is appropriate for where we’re sitting?
- If we wanted to scale, how would we do it without destabilising what’s working?
You’re not looking for a gotcha. You’re looking for a conversation that feels considered. The right answer isn’t “we’re hitting 50, we’re fine.” It isn’t “we’re not hitting 50, we’re stuck.”
The right answer sounds more like “we’re at X conversions a week, we’ve made the call to optimise for Y because of Z, here’s what we expect performance to look like, and here’s how we’d scale when you’re ready.”
If that conversation feels one-way, or if the answers feel vague, it’s worth pushing. Not because anyone’s doing a bad job. Because this is exactly the kind of framing conversation that should happen before the ads go live, and a lot of the time it doesn’t.
If you’re running the ads yourself, the same questions apply. Ask them to the spreadsheet. Be honest about the answers.
How We Do This At Base Brands
When we take on a new paid social client, the first thing we do is work out what their realistic conversion volume looks like before we touch the ad account.
We look at the cost of acquiring a customer. We look at the budget they can sensibly commit to. We do the sum. If 50 a week is realistic, great, we plan around that. If it’s not, we have the conversation up front about which lever we want to pull and what we want the first few months to actually look like.
This is the bit I want to be loudest about. The most valuable thing a paid social partner does is set expectations before spend goes live, not after.
We tell clients what to expect in concrete numbers. Weekly variance of X%. Budget changes at 15 to 20% maximum for the first quarter. Three-week performance windows, not one-week ones. Written down. Agreed in advance.
None of it is clever. It’s just calibrating expectations to match structure, which is the thing everyone thinks they’re doing and most people aren’t.
Frequently Asked Questions
Can I ignore the Learning Limited status if my ads are profitable?
Yes. Learning Limited is a diagnostic, not a verdict. If your ads are delivering the results you need at a cost you can live with, the status is essentially cosmetic. Plenty of profitable campaigns run in Learning Limited for months.
Should I always move up the funnel if I can’t hit 50?
No. It’s one option, not the default. Moving up the funnel means the algorithm learns faster but on a less direct signal. If your real conversion event is firing at all, and you can accept slower scaling, there’s a strong case for staying optimised on it. Decide based on your product and your timeline, not on what’s easier.
What’s the minimum budget for Meta ads to work at all?
There’s no universal minimum. The right question is whether your budget generates enough conversions for the algorithm to do useful work. If your cost per customer is £20, a modest daily budget and a single ad set can work fine. If your cost per customer is £200, that same budget won’t get off the ground. The budget has to match the economics of what you’re selling.
How long should I run a new campaign before judging it?
Three weeks minimum if you’re sub-50, and longer if volume is very low. Weekly judgements on low-volume campaigns are mostly noise. Look at 14 to 21 day rolling averages for cost per result before deciding whether something is working.
Does Advantage+ Audience fix the sub-50 problem?
No, but it helps a bit. Advantage+ lets Meta’s AI do more of the targeting work, which can surface conversions you wouldn’t have reached with narrower manual audiences. It doesn’t change the maths of needing signal volume to learn, but it does broaden where that signal comes from. Most small businesses should be using it either way in 2026.
Want This Done Properly For Your Business?
If any of this is making you wonder whether the conversations you’ve been having about your ads are the right ones, that’s a useful thing to notice.
We help business owners at Base Brands work out what their ads should actually be doing given their budget, their product, and their realistic conversion volume.
We set expectations before anything goes live, run the campaigns with a full-funnel view, and talk to our clients like partners, not like they should be grateful for screenshots.
If you’d like to talk about what that might look like for your business, we’d love to hear from you.
