TL;DR
Most Meta ads best practices fall into three buckets: genuinely useful, Meta upsells dressed as advice, and outdated rituals from 2019. The useful ones: consolidate your structure, run CAPI properly, feed the machine creative volume and variety, give changes time to work, and measure blended results instead of worshipping platform ROAS. The upsells: auto-applied recommendations, "increase your budget" nudges, and treating every Advantage+ toggle as mandatory. Retire interest stacking, hyper-segmented ad sets, and manual bid fiddling as a daily hobby.
Meta's best practices have a conflict of interest
Let's start with an uncomfortable fact: Meta's best practices are written by a company that makes money when you spend more. That doesn't make all of them wrong. It does mean you should read them the way you'd read "gaming tips" published by a casino. Some are real strategy. Some are just house edge with better branding.
We've run enough accounts to sort them. Here's our list, in three buckets.
Bucket 1: Actually follow these
Consolidate your structure. One CBO campaign with a few ad sets beats fourteen campaigns with forty ad sets. Every ad set you add splits signal and starves learning. This has been true for years and it's more true now that the algorithm does the targeting work. If your account looks like a filing cabinet, that's not sophistication, it's fragmentation.
Run CAPI, and run it well. The pixel alone is a suggestion box. Conversions API is the actual plumbing. Deduplicate properly, send the full event payload, and check Events Manager like it's your job, because it is. Bad signal in means bad decisions out, and no amount of creative genius fixes broken measurement.
Feed it creative volume and variety. This is the closest thing to a cheat code left in Meta ads. The accounts that win test constantly: new hooks, new angles, new formats. Not because every test wins, but because the winners pay for all the losers ten times over. Creative is the variable you control most, so it's where your effort should go.
Stop touching things. Every significant edit resets learning. Budget changes over 20% in a day, objective swaps, attribution window changes: each one tells the algorithm to start over. Make a change, then leave it alone for 3-4 days minimum. The number of accounts we've seen "underperforming" purely because someone fiddles daily is depressing.
Measure what matters. Platform ROAS is a directional indicator, not a P&L. Judge the account on blended MER, contribution margin, and actual bank deposits. An account at 2.1x platform ROAS that's growing profitably beats an account at 4x that's shrinking. Every time.
Bucket 2: Meta upsells wearing a best-practice costume
Auto-applied recommendations. Meta will happily apply its own suggestions to your account automatically. Some are fine. Some will raise your budgets, expand your targeting into placements you deliberately excluded, and turn on creative "enhancements" that rewrite your ads. Review every recommendation manually. The button that says "apply all" is not your friend.
"Increase your budget to exit learning." Sometimes true. Often just Meta asking for more money. Before you scale spend to satisfy the learning gods, ask whether the ad set deserves more budget at all. Pouring budget into a mediocre ad set doesn't fix it, it just teaches the algorithm to find mediocre customers faster.
Every Advantage+ toggle, always on. Advantage+ audience, Advantage+ placements, Advantage+ creative enhancements: each is situationally useful and none is universally correct. Advantage+ creative enhancements, in particular, will happily crop your video, rewrite your copy, and add music you didn't choose. Test them like anything else. "Best practice" is not a synonym for "turn everything on and hope."
The 50-conversion rule as religion. Yes, ad sets need roughly 50 conversions a week to exit learning properly. No, that doesn't mean you should force it by smashing unrelated products into one ad set or setting absurd budgets on day one. It's a guideline for structure, not a quota to hit by any means necessary.
Bucket 3: Retire these
Interest stacking. Combining twenty interests into one ad set was a 2019 workaround for small audiences. In 2026, with broad targeting and Advantage+ doing the matching, it's astrology. Either go broad or test interests cleanly, one per ad set, with real budgets. The stack tells you nothing about which interest worked.
Hyper-segmented ad sets. Separate ad sets for every age bracket, gender, and placement made sense when you controlled delivery. Now you're just fragmenting signal and paying for the privilege. Segment when you have a creative reason (different creative for different audiences), not a spreadsheet reason.
Manual bidding as a daily hobby. Bid caps and cost controls have their place (we wrote a whole article on them), but daily bid fiddling is not media buying, it's anxiety with a dashboard. Set strategy deliberately, then let it run.
Duplicating ad sets to "reset learning." Dead for years. Still recommended in Facebook groups by people who last ran an account profitably during the Obama administration.
Judging creative on day-one ROAS. New creative needs spend to find its audience. Killing ads after $12 of spend because ROAS isn't 4x yet is how you murder winners in the crib. Judge early creative on thumbstop and hold rate, not revenue.
The short version
If you only take five things from this article: consolidate structure, fix your measurement, test creative relentlessly, stop touching the account daily, and measure profit instead of platform ROAS. Everything else is commentary.
The meta-lesson (sorry): whenever Meta or a guru hands you a "best practice," ask who profits if you follow it. If the answer is Meta's quarterly earnings, verify before you obey.
FAQ
What is the single most important Meta ads best practice in 2026? Creative volume and variety. The algorithm handles targeting and delivery; the input it can't generate for itself is fresh creative. Accounts that test constantly outperform accounts that optimize settings.
Should I follow all of Meta's recommended best practices? No. Treat them as suggestions from an interested party. Consolidation, CAPI, and creative testing are genuinely good advice. Auto-applied recommendations and blanket "increase budget" nudges deserve skepticism.
How many ad sets should a Meta campaign have? As few as your testing plan requires. Most accounts do best with consolidated CBO structures: one prospecting campaign with a handful of ad sets, one retargeting campaign. Dozens of ad sets fragment signal.
Are Advantage+ features mandatory now? No. They're tools, not commandments. Advantage+ audience and placements work well for many accounts, but creative enhancements should be tested, not blindly enabled. Evaluate each on performance.
How often should I check my Meta ad account? Daily monitoring for breakage (spend pacing, tracking, disapprovals), but structural changes only when you have a diagnosis. Checking daily and changing daily are very different activities.
Written by the MetaMaxd team. We run Meta ads for brands that are tired of agency theater: no vanity metrics, no "hacks," just accounts engineered to make money. If your account is full of best practices and empty of profit, that's what our audits are for.
