Open Ads Manager this year, and there’s a fair chance your ROAS is lower than it was twelve months ago. Same budget, same product, same audience settings. If you’ve typed “why is my Meta ROAS dropping” into Google lately, you’re not alone. Meta changed how it picks ads, and a few old habits don’t fit the new system anymore.
Here’s the short version. Meta’s AI now decides who sees your ad, and it mostly makes that call by reading the ad itself. So the biggest levers to improve ROAS on Meta ads are your creative, your offer and the quality of your tracking data. Fiddling with audiences matters a lot less than it used to.
Below, we cover what changed, how to handle targeting, how many ads to test, and what to check when your numbers slip.
Why does Meta Andromeda’s creative diversity matter more than your audience settings?
Meta picks ads in two steps. First, a retrieval system cuts millions of possible ads down to a few thousand for one person. Then a ranking step chooses what to show. Andromeda is the first step. Meta started rolling it out in late 2024 and has described it as roughly 10,000 times more complex than the older retrieval setup.
For you, it means the system looks at what your ad says and shows, then matches it to people who might care. If all your ads look alike, it treats them as one ad. Take five videos with the same person and the same script, only the opening line changed. That’s one idea in five outfits.
Real Meta Andromeda creative diversity looks more like this:
- A founder talking straight to the camera
- A customer review or short testimonial
- A product demo with no voiceover
- A before and after
- A comparison with the usual alternative
Each one speaks to a different type of buyer, so each gives Meta a different way to find people. A different problem, a different person on screen, a different format, and a different first three seconds. Changing a button colour or a headline doesn’t count.
Broad targeting vs interest targeting on Meta: what to pick in 2026?
Meta has been merging and retiring detailed interest options since 2025, and more went in early 2026. On new campaigns, Meta Advantage+ targeting is the default. Meta treats the interests you type in as suggestions, not hard limits. So that tidy interest stack you built a few years ago may be partly gone and partly ignored.
Does broad always win? Not quite. Broad needs data to work with. Here’s a rough guide many advertisers follow. It’s a rule of thumb, not a Meta rule.
| Your account today: | Where would we start? |
| 50+ purchases or leads a week | Broad or Advantage+ audience |
| 15 to 50 a week | Advantageous audience, with your best interests as suggestions |
| Under 15 a week, or a new pixel | Interests or lookalikes, then widen |
| Very niche B2B offer | Interests plus custom audiences |
| Cart abandoners, past buyers | Custom audiences (retargeting) |
On broad targeting vs interest targeting on Meta, lead gen advertisers should watch lead quality, not just cost per lead. In an older public test, Advantage+ gave cheaper leads, but the cost per qualified lead came out higher than with interest targeting. Cheap leads that never pick up the phone won’t improve ROAS on Meta ads. If you run lead forms, add a qualifying question or two, and send real outcomes like booked calls or sales back to Meta.
Also, don’t judge a switch after three days. Run broad next to your current setup for at least two weeks, as a challenger.
How many creatives to test on Meta ads (a number you can actually use)?
Meta hasn’t published one number for Andromeda. It used to suggest six or fewer ads per ad set, and plenty of scaled accounts now run 10 to 20 or more. But one well-known Meta consultant tested 25 and 50 ads in a set and saw no clear gain. More isn’t automatically better.
So how many creatives to test on Meta ads comes down to budget and signal. Meta says an ad set needs around 50 results a week to leave the learning phase, and every ad inside it shares that pool. Say a sale costs you ₹800. Fifty sales is ₹40,000 a week for one ad set. Split that across four ad sets, and each gets about 12 results. That’s why we lean toward fewer ad sets with more distinct ads in each.
Here are starting points. They’re ours, not Meta’s.
| Weekly budget: | Ads per ad set: | New concepts each week: |
| Small (one ad set barely hits 50 results) | 4 to 6 | 1 to 2 |
| Mid | 8 to 12 | 2 to 4 |
| Larger | 12 to 20 | 4 or more |
Plan for most of them to lose. Teams often report that only around one in ten tests turns into a real winner. Give each test about a week, judge on cost per result and not click rate alone, and leave live ads untouched, since edits can reset learning.
When we plan creative testing at 360 Digital Idea, we build each batch around different angles and formats, test them against each other, then move the winners into scaling. Spending a little on testing first is cheaper than putting the full budget behind a guess.
Why is my Meta ROAS dropping? Five checks, in order:
Before you change anything, work out what kind of drop this is. The same ROAS number can come from very different problems.
| What do you see? | Likely cause: | First move: |
| Meta ROAS fell; store revenue didn’t. | Reporting change | Compare with your own sales data. |
| Conversions look low across the board. | Tracking gap (pixel, Conversions API, delays) | Check the Events Manager for missing or duplicate events. |
| CTR falling, frequency rising, same ads for weeks | Creative fatigue | Launch new concepts, not tweaks. |
| ROAS slipped after a budget increase. | Scaling | Look at the return on the extra spend. |
| CPM and CTR steady, fewer sales | Site or offer | Check page speed, price, stock, checkout |
On the first one: in March 2026, Meta changed attribution so click-through credit goes only to link clicks. Likes, saves, and shares moved into a separate “engage-through” bucket. Reported conversions fell for plenty of accounts even though real sales didn’t. Check your own sales numbers before you cut any budget.
Creative fatigue is where we’d look next. By the time ROAS drops, the ad has usually been tired for days, so watch frequency and click rate every week, not every month.
Scaling is the one people forget. You buy the cheapest sales first, so the next ₹10,000 almost always returns less than the first ₹10,000. That dip is normal. It’s only a problem if it keeps sliding.
At 360 Digital Idea, we check tracking and the landing page before we touch the ads, because a slow page or a broken event can look exactly like an ad problem. If you want a second pair of eyes, our team can go through your account as part of our Meta Ads service.
A simple 30-day plan to improve ROAS on Meta ads:
- Week 1: Fix measurement. Check your Pixel and Conversions API, and set a baseline from your own sales data.
- Week 2: Simplify. Fewer campaigns, fewer ad sets. Start one Meta Advantage+ targeting campaign next to your current one.
- Week 3: Launch 6 to 10 ads that are genuinely different from each other.
- Week 4: Read the results. Cut the losers, keep the winners running, and brief the next batch.
Then repeat every month. The accounts that hold their numbers tend to be the ones that keep feeding Meta new ideas.
Final thoughts
Meta’s AI has taken over a lot of the work that used to be yours. What it can’t do is come up with good ideas for you. Give it different, honest, creative, clean data and enough budget in fewer places, and you’ve got a fair chance to improve ROAS on Meta ads this year. Nobody can promise a specific number, because results depend on your product, price and market. If you’d like help with your own account, our team at 360 Digital Idea can review it with you.
FAQs
Q1. What is Meta Advantage+ targeting?
It’s Meta’s AI-led way of finding your audience. You can add suggestions like interests or age, but Meta can go beyond them if it expects better results. It’s now the default for new campaigns.
Q2. Is broad targeting better than interest targeting?
On broad targeting vs interest targeting on Meta, broad usually wins once your account gets steady conversions every week. New accounts, tiny budgets and very niche offers can do better starting with interests, then widening later.
Q3. How many creatives to test on Meta ads at one time?
Most accounts do fine with 4 to 6 distinct ads per ad set on a small budget, and 8 to 20 on a larger one. Different ideas matter more than a big count. Add new concepts every week.
Q4. Why is my Meta ROAS dropping when sales look the same?
Often it’s reporting. Meta narrowed click-through attribution to link clicks in March 2026, so fewer conversions get credited to ads. Check your own sales data first, then look at tracking, creative fatigue and your landing page.
Q5. What is a good ROAS on Meta ads?
It depends on your margin. Divide 1 by your gross margin to get your break-even ROAS. At a 40% margin, that’s 2.5. Anything above it is profit before other costs, and the right target varies by category.
Q6. Does Meta’s Andromeda creative diversity mean every ad needs to be a video?
No. Mix formats like images, short video and carousels, but the bigger gain comes from different ideas, angles and people on screen, not different file types.
Q7. How long should a creative test run?
About a week is a common minimum before you judge results. Wait for enough conversions, compare cost per result, and don’t edit live ads during the test.