Update: That FB Ad Description Trick? The Data Changed My Mind.

Apr 8

A while back I wrote a post about a simple Facebook Ad tweak – filling in the Description field with an AI-generated, trope-stuffed description of the book – that appeared to cut my Cost Per Unit by 29% for one campaign. Two out of four campaigns improved, and I said at the time that the data was “promising but inconclusive” and that my next step was to test more descriptions at scale.

Well, I finally did. And I owe you an update – because the data tells a different story now.

The Original Finding

For those who missed the original post: Facebook gives you a Description field when you set up an ad – that little text block underneath your primary copy that most advertisers just skip right over. The idea was that maybe a strategic, keyword-loaded description could work like an Instagram hashtag and help Facebook’s algorithm target the right users.

The initial results were mixed but intriguing. One book’s UK campaign saw a 29% improvement in CPU with a description. But the other tests were split – two improved, two didn’t. I said I needed more data, and I meant it.

The Expanded Test

So I ran the test again – this time across three different romance novels in KU by different authors in different subgenres. Each book had a set of ads running with a description and an identical set running without one. I let them run for weeks, accumulating thousands of dollars in spend and tens of thousands of clicks before pulling the data.

Here’s what happened:

Book 1: The description version won – barely. $4.29 CPU vs $4.65 for no description. That’s an 8% improvement, which sounds decent until you realize it’s a fraction of the 29% I reported originally.

Book 2: No description won. $3.56 CPU vs $4.32. The version without a description was 18% more efficient.

Book 3: No description won again – and it wasn’t close. $3.70 CPU vs $6.32. The no-description version was 41% cheaper per unit.

  
    

  
      Book
      Description?
      FB Spend
      Clicks
      CPC
      Units
      CV%
      CPU
    

  
      Book 1
      No
      $2,276
      15,114
      $0.15
      489
      3.2%
      $4.65
    

      Book 1
      Yes
      $5,062
      31,672
      $0.16
      1,180
      3.7%
      $4.29
    

      Book 2
      No
      $1,000
      5,191
      $0.19
      281
      5.4%
      $3.56
    

      Book 2
      Yes
      $2,724
      17,572
      $0.16
      631
      3.6%
      $4.32
    

      Book 3
      No
      $3,492
      30,018
      $0.12
      944
      3.1%
      $3.70
    

      Book 3
      Yes
      $269
      2,128
      $0.13
      43
      2.0%
      $6.32
    

      Averages
      
      
      
      
      
      
    

      No Description
      $2,256
      16,774
      $0.15
      571
      3.9%
      $3.97
    

      With Description
      $2,685
      17,124
      $0.15
      618
      3.1%
      $4.98
    



  

The Averages Tell the Story

When I averaged the CPUs across all three books, the no-description ads came in at $3.97 per unit compared to $4.98 for the ads with descriptions. That’s a 20% advantage for leaving the description field blank – the exact opposite of what I originally reported.

The conversion rates told a similar story. Ads without a description converted at 3.9% on average, while ads with descriptions converted at just 3.1%.

One thing I did notice: Facebook’s algorithm gave significantly more spend to the ads with descriptions in two of the three tests. MW1’s description ads absorbed more than double the budget, and Book 2’s description ads got nearly triple. This makes sense – more text gives Facebook’s AI more to work with when deciding who to serve the ad to. But more spend doesn’t mean better performance, and in this case, the algorithm’s preference didn’t translate to a lower cost per acquisition.

So What Changed?

Honestly – probably nothing changed. The original test was likely just a small sample that happened to favor the description. It happens. That’s why we test at scale.

If I had to guess why the description might actually hurt in some cases, I’d point to congruence. When someone clicks your ad and lands on your Amazon page, you want as few surprises as possible. The description field adds another piece of text that the reader processes before they click – and if it doesn’t perfectly match what they find on the product page, it might create just enough friction to hurt conversions. But that’s speculation, not data.

The Takeaway

If you implemented the description trick after my original post – don’t panic. The data suggests it’s not going to destroy your campaigns. But it’s also not the easy win I thought it was. Across three books, leaving the description blank performed better on average.

My recommendation? Leave it blank. There are much higher-impact things you can spend your time testing – like covers, ad copy, audiences, and creative formats. The description field just isn’t one of them.

And hey, this is exactly why I keep testing and sharing updates. If I only ever posted my wins and never corrected the record when the data shifted, that wouldn’t be very helpful. The whole point of this is to get closer to the truth – even when the truth is “yeah, that thing I told you about? Never mind.”

Note: This data is from KU romance novels running Traffic campaigns on Facebook to US-based, women-only audiences aged 21-65+. As always, your results may vary depending on your genre, audience, and creative.

Ben DeMeter