Why Your Meta Ads Are Stuck in Learning Phase — And Why Fixing It Is Not What You Think

saurav soni
11 hours ago
4 min read

Had a conversation recently with a brand running Meta ads that had been stuck in learning phase for three weeks. They'd done the research. They knew about the 50 events requirement. They were optimising for Initiate Checkout. The events weren't coming in fast enough and the campaign just sat there, burning budget, going nowhere.

Here's the thing — they were doing it wrong. Not because they hadn't learned enough. Because the advice they'd learned was being applied without understanding what's actually happening underneath it.

This is one of those areas where performance marketing gets genuinely misunderstood at scale — people hear the rule, apply the rule literally, and miss the point entirely. Let me break down what's actually going on.

The 50 events rule — what it actually means

The 50 events threshold is at the ad set level — not the ad level. This is the misunderstanding I see constantly. Meta aggregates all conversion signals across every ad inside a single ad set. So if you have three ads in one ad set and they collectively drive 50 Initiate Checkouts in a week, you're out of learning. The individual ads don't need 50 events each.

This is actually why consolidating ads within an ad set helps — more creatives means more delivery surface, which means faster signal accumulation. People running one ad per ad set are making the learning phase problem worse for themselves without realising it.

Why Initiate Checkout optimisation is probably the wrong choice

Initiate Checkout campaigns are not working for most brands right now — and the reason is straightforward. You're optimising for an event that happens mid-funnel, and Meta's algorithm is finding people who initiate checkouts, not people who complete purchases. On a good landing page that might be fine. On most landing pages, you're getting a load of abandoned carts and no ROAS.

The question nobody asks themselves is — how are you tracking ROAS on Initiate Checkout optimisation? If you're optimising for add to cart or IC events but measuring success on purchases, your conversion window, your pixel attribution, and your reported numbers are almost certainly misaligned. You're making budget decisions on data that doesn't reflect what's actually happening.

Same problem with traffic campaigns. I see brands optimising for traffic, getting good click numbers, and then wondering why sales aren't coming in. Traffic optimisation finds people who click. That's all. Whether they buy is a completely separate question that the algorithm isn't trying to answer.

The landing page problem nobody talks about

Traffic from Meta hitting a product page versus a collection page versus a homepage behaves very differently. For IC optimisation to work efficiently, the ad should land directly on the product page — with price visible, trust signals present, and the path to checkout as short as possible. Sending paid Meta traffic to a homepage or a collection page and wondering why IC events aren't accumulating fast enough is a landing page problem, not an algorithm problem.

The campaign doesn't need to be restructured. The destination does.

When you can't hit 50 events — what to actually do

If you don't have the budget or the volume to hit 50 purchase events per ad set per week, that's okay — some campaigns still perform well without hitting that threshold. But if you need to make smart bidding work properly and you're stuck below the data threshold, the best move right now is Advantage+ Shopping Campaigns.

ASC is built exactly for this situation — limited budget, limited purchase volume, can't hit 50 events per ad set in the standard structure. Meta's ASC aggregates signals at the campaign level across a broader audience, which means it can optimise meaningfully at lower volumes than a standard purchase campaign requires. It's not the answer to every problem but for smaller budgets trying to get out of learning phase, it's the most practical option available right now.

The bigger problem — performance marketing is maths, not vibes

Meta's algorithm is genuinely good. When you run a purchase campaign early on a fresh pixel, it often finds buyers — somehow, it works, and the early numbers look great. A lot of marketers see that early performance and think they've done a great job. Then they try to scale and everything falls apart, and they're back watching reels trying to figure out what changed.

What changed is that the algorithm ran out of easy wins in a small audience and now needs proper signals to find the next layer of buyers. And if you don't understand why it worked in the first place — what objective you were running, what the attribution window was, what the pixel was actually tracking — you can't diagnose why it stopped.

You can't be delusional in performance marketing. It's numbers. When something works, you need to understand why it works. When something stops working, you need to solve it like a maths problem from multiple angles — campaign structure, optimisation event, landing page, audience signal quality, budget versus volume requirements. Not gut feel. Not more creative testing. The actual maths.

This is the kind of diagnosis I do in every new account I take on — working out what the algorithm is actually being asked to do, whether it has enough signal to do it, and what needs to change structurally before throwing more budget at the problem. If you want that done for your account:

→ Book a Free Strategy Call

Subscribe to my blog