It’s October 17th — Day 47 of the journey, and the numbers perfectly illustrate how unpredictable this whole process can be.
After one of my weakest days earlier in the week, we saw a strong rebound yesterday:
💰 Sales: $1,103
💸 Ad Spend: £358 (≈ $481)
📈 Profit: $622
That’s a solid day — anytime we break the $1,000 mark in sales, we’re right where we need to be.
Today’s pacing, though, tells the familiar story:
$438 by 4:45 PM, trending toward maybe $600–700 by end of day.
This is the rhythm of Facebook ads: bounce, dip, rebound, repeat. The key is to keep your head steady while your metrics dance around you.
Lesson of the Day: The Hidden Carousel Trap
Today’s big takeaway isn’t glamorous — but it’s one of those technical lessons that saves you money and sanity once you learn it.
It’s about Facebook Carousel Ads, and how one simple oversight nearly tanked my results.
The Setup: A Great Carousel That Started to Fatigue
My top-performing ad for weeks has been a carousel ad — a series of slides that tell a story, frame by frame, guiding prospects through the problem, solution, proof, and offer.
It’s worked beautifully because it acts as a mini-sales page inside Facebook. By the time someone clicks through, they’ve already followed the emotional arc of the offer.
But even the best creatives have a shelf life.
After running for a while, the frequency crept up and the cost per purchase began to rise.
So I decided to refresh it.
The Mistake: Letting Facebook Crop My Images
For the new carousel, I thought I’d get clever and make portrait (vertical) images.
The logic seemed sound — most people view on mobile, right?
Instead of manually designing for every placement (feed, stories, reels, etc.), I figured I’d just upload the tall images and let Facebook’s cropping tool handle the rest.
Bad idea.
When I reviewed the performance:
Carousel Version Unique CTR (All) Outbound CTR Original 3.2% 1.94% New (portrait-cropped) 1.93% 1.29%
That’s nearly a 40% drop in click-through rate — and it wasn’t because of the copy or offer.
When I checked the ad preview, I saw the issue:
Facebook had cropped the slides inconsistently. Some text was cut off. Headings overlapped. Images looked awkward in the feed.
The ad looked broken.
The Fix: Custom Images Per Placement
Here’s what I did to correct it:
Created 1080x1080 (square) versions of each slide.
These fit perfectly for Facebook and Instagram feeds.
Manually replaced images in each major placement:
Facebook Feed
Instagram Feed
Facebook Reels
Instagram Reels
Stories and Right Column
Previewed every placement to make sure nothing was cropped, off-center, or missing text.
Once that was done, I relaunched the ad.
And the difference was instant:
Carousel Version Unique CTR (All) Outbound CTR After Fix (square images) 3.1% 1.97%
The numbers jumped right back up — almost perfectly matching my original high-performing carousel.
The Lesson: Don’t Rely on Auto-Cropping
Facebook’s automatic resizing might be convenient, but it’s not designed for precision.
If your slides include text, headlines, or CTAs (and they should), always:
✅ Upload properly sized images for each placement.
✅ Preview all placements before publishing.
✅ Check after publishing — especially in mobile view.
It’s a little extra work, but it can easily mean the difference between $10 and $20 CPAs.
Other Updates
📊 The Caples-inspired story landing page split test is live.
Too early for data — I’ll have first insights in the next 2–3 days.
🔄 The upsell test officially closed yesterday (control held strong).
🎨 Next up: optimizing carousel copy order and testing whether to include price on the final slide.
Key Takeaways
✅ Never trust Facebook’s auto-crop for carousel placements.
✅ Upload custom images for each placement (feed, Reels, Stories, etc.).
✅ CTRs can tank just because your visuals render poorly.
✅ Fixed formatting took CTR from 1.9% → 3.1%.
✅ Keep iterating — every micro-optimization adds up.
jonathanhowkins.com
I want to help Course Creators succeed in predictably and profitably generating more leads and sales using Facebook Advertising.