What $27M in Meta Ad Spend Taught Us 👀

Welcome to your latest edition of Welcome Tomorrow Insights, your go-to source for the latest in growth marketing strategies and impactful industry updates.

Aug 05, 2025

TL;DR:

Let the algorithm do its job. Meta works best when you give it space, don’t over-segment audiences, restrict placements, or pause campaigns too quickly. Trust the system to learn and optimise.
Creative is the real growth lever. High-performing campaigns are powered by native, scroll-stopping creatives not high production value. Test often, but don’t rotate too fast.
Don’t fall for misleading metrics. ROAS and CTR can be deceptive. Focus on what really matters: incrementality, CAC, and full-funnel impact.
Performance requires simplification. Simplify your campaign structure, messaging, and strategy. Clarity, consistency, and clean data signals win every time.

After managing over $27 million in Meta ad spend across African markets, we’ve started noticing something: many brands ,even the experienced ones, end up relying on similar strategies, without always knowing what’s making them work.

And when things stop performing, it’s hard to know whether to wait it out or switch things up. We've been there too.

In fast-moving markets, this kind of uncertainty is common. Success often feels hard to replicate, and performance drops can seem random even when you're doing “everything right.”

That’s why we started documenting patterns we kept seeing across accounts.

What follows are 14 lessons we have learned.Some of them go against popular advice, but they’ve helped us move from guesswork to real, repeatable performance.

Let’s get into it 😊

1. Killing a campaign too quickly ruins performance

It’s one of the most common mistakes.

The campaign is humming along. CPA jumps for two days. Panic sets in and boom, the budget gets cut in half or paused completely.

But Meta campaigns aren’t static rows in a spreadsheet. They’re dynamic learning systems. Every campaign builds targeting signals, behavioural models, and delivery patterns over time. Interrupt that, and you reset everything.

Unless the trend is structurally negative, short-term volatility is not a red flag. Zoom out. Look at 14-day data. Trust the flywheel.

2. Your prettiest ad is probably underperforming

Brands often spend big on production, high-end lighting, drone shots, polished edits. The result? An ad that users scroll right past.

Why? Because on Meta, anything that looks overly produced can trigger “ad blindness.” People are trained to ignore the hard sell.

What performs better? Creatives that clearly speak to a pain point and offer a solution. Content that feels relevant, human, and valuable in-feed. It’s not about being raw, that works better on TikTok. On Meta, it’s about resonance.

This isn’t a call to be sloppy. It’s about designing for the environment. Meta is a content-first platform, your ads need to connect before they convert.

3. Broad targeting wins more often than narrow segmentation

The instinct is to get hyper-precise. Build detailed personas. Layer interests, demographics, behaviours.

But Meta’s algorithm doesn’t need that kind of hand-holding anymore. With billions of data points, it can find the buyers better than you can define them.

When you restrict audience size, you limit delivery, reduce learning, and create artificial ceilings.

What works better? Give the algorithm space. Start broad. Let your creative qualify your audience, not your targeting filters.

4. When something works, duplicate it, don’t rebuild it

Performance marketers love to tweak. You’ve got a winning ad? Let’s make it “even better,” right?

Except most of the time, the “improved” version underperforms.

The problem is: when you change the format, script, or angle too much, Meta sees it as a new creative. You’re back to square one. New learning phase. No momentum.

Instead, duplicate the top performer and change one thing. The first 3 seconds. The CTA. The hook. That’s how you create iteration without resetting success.

5. Retargeting looks great on paper. But a lot of it is wasted spend.

This one stings.

Retargeting often delivers sky-high ROAS. But in many cases, those users would’ve bought anyway, with or without the ad.

Meta’s attribution model is happy to claim the credit. But your actual incremental lift? Minimal.

Retargeting only works when it’s focused on high-intent actions:
→ Added to cart
→ Visited a specific PDP
→ Watched 75%+ of a product demo

And the lookback window should be tight, 3 to 7 days max. Beyond that, you’re just paying to take credit for organic conversions.

6. ROAS is useful but dangerously incomplete

Most advertisers optimise for ROAS. But here’s the issue:

ROAS = revenue tracked by Meta ÷ ad spend.

That means:

It doesn’t tell you if Meta actually caused the sale.
It doesn’t tell you if the customer is new.
It doesn’t factor in whether those sales would’ve happened anyway.

The more established your brand, the more inflated your ROAS gets, because you’re paying to reacquire loyalists you already had.

The real metric to obsess over? Incrementality.
→ Are your campaigns driving net new growth?
→ Is your overall revenue rising in step with ad spend?

Use ROAS as a directional metric, not as a KPI in isolation.

7. If your ad doesn’t stand for something, it fades into the feed

Neutral messaging feels safe. Broadly appealing. Professional.

But on Meta, neutral = invisible.

The algorithm responds to engagement. Comments. Shares. Reactions. And you only get that when you say something specific, even if it excludes part of the audience.

This isn’t about being controversial for attention. It’s about clarity. Speak directly to a pain point. Take a position. People pay attention to conviction.

8. Don’t micromanage placements. Let Meta optimise them.

Every advertiser has placement biases. You assume Reels are where conversions happen. Or you always disable Marketplace because it “feels low-quality.”

But Meta’s system knows more than we ever could, when your audience is most likely to engage, and where.

We’ve seen top-performing placements come from unexpected channels even ones clients originally excluded.

Let Meta test all placements. If it doesn’t work, the algorithm will de-prioritise it. But if you block it upfront, you’ll never know.

9. High CTR doesn’t always mean strong performance

CTR feels like a win. It’s visible, immediate, and easy to chase.

But high CTR can be deceiving especially if CPM is high or your traffic doesn’t convert.

👉🏽 On web, a high CTR might lead to a bounce or low ROAS if the landing page doesn’t convert.
👉🏽 In app campaigns, it could mean a lot of low-quality installs, especially if your install-to-event rate is poor or users drop off before key in-app actions.

The real questions are:
→ What’s your CPC?
→ What’s your conversion rate (on-site or in-app)?
→ What’s the blended CAC?

CTR can be a useful directional metric when comparing creatives under controlled conditions. However, in isolation, it offers limited insight and its implications vary significantly depending on whether the objective is web traffic or app engagement.

10. Refreshing creatives too often can tank momentum

You’ve heard it before: “Rotate new ads often to avoid fatigue.”

But constant refreshes mean constant learning phases. If you swap creatives too quickly, the algorithm never gets a chance to optimise delivery. You keep re-setting your own progress.

Instead, build a strong base of proven performers and test incrementally alongside.

Winning ads deserve to live. Don’t pull them just to feel busy.

11. Over-segmentation hurts account performance

This is an old-school media buying habit that won’t die:
→ A campaign for every objective
→ An ad set for every interest
→ A budget for every micro-audience

But Meta learns best when it has consolidated signal density.

Simplify your structure:
→ Combine ad sets by funnel stage
→ Use CBO (Campaign Budget Optimisation)
→ Reduce budget fragmentation

Fewer campaigns. More data per campaign. Better results.

12. Your ad should say one thing. Just one.

The urge to pack every benefit, guarantee, testimonial, and feature into one ad is strong.

But that’s not how people read.

On Meta, attention is brutal. If the user doesn’t get your value prop in the first few seconds, you’re gone.

Instead, lead with one clear promise. What’s the single thing that makes someone stop and say: “I need this”?

Everything else can live on your website.

13. Not every “unprofitable” campaign is a failure

We’ve seen this over and over: a top-of-funnel campaign looks like it’s underperforming. No conversions. Low ROAS.

But it’s feeding your retargeting pool. Warming your audience. Seeding new messaging.

When evaluated alone, it looks weak. In the context of your funnel, it’s essential.

So before you shut it down, ask:
→ Is it generating qualified traffic?
→ Is it fuelling lower-funnel success?
→ Is it teaching us something new?

Not every campaign has to convert directly to be worth running.

14. Don’t use Target CPA too early

Target CPA sounds efficient. You cap your cost, and Meta delivers within your range. Easy.

But here’s the catch: it only works well if you already have conversion volume.

Without it, you strangle learning before it starts.

Instead, use “Maximise Conversions” in the early stage. Let the algorithm explore. Once you’ve built conversion density, you can shift to Target CPA for tighter control.

Final takeaway

Here’s the thread that ties all 14 truths together: great Meta performance comes from knowing when to intervene and when to get out of the way.

Too many brands sabotage their own results with:

Over-targeting
Over-editing
Overreacting

But if you feed the algorithm high-signal creative, clean structure, and time to learn, it works.

At Welcome Tomorrow, this is exactly what we help brands do. Whether you're trying to unlock profitable scale or course-correct a stagnant ad account, we’d love to help.

👋🏾 Want to share this newsletter with a friend? Send them here to subscribe.

Wambui’s Hot Take 🔥

It’s easy to assume the algorithm will favour the ad with the lowest CPA or highest ROAS. But that’s not always what happens. Meta is optimising for predictability, not just performance.
An ad that performs decently and behaves consistently is often prioritised over one that spikes results but delivers inconsistently. Why? Because the algorithm can allocate budget more confidently when it can model an outcome.
So if your flashiest ad doesn’t scale, it might not be because it’s bad, it might just be too erratic for Meta to trust.

Wambui Mutero, Paid Media Manager at Welcome Tomorrow

Join Our Community 👥

Come hang with us. Check out the blog, follow along on socials, and stay in the loop as we share the wins, lessons, and experiments shaping growth marketing today. 🤝🏾
PS: The really good stuff? You’ll find that on our TikTok. 👀😂