A Tale of Two Facebook Ad Strategies

I mentioned in a previous post that I commissioned an artist to hand draw a new cover for Ahelno. I want to move away from A.I. images, not only to support other creatives, but also because I don’t want readers to think my writing is also A.I.-generated, which it is not. 

I love the new cover (which I’ll reveal in a later post), but I admit I was also curious. Social media is often an echo chamber, and it’s full of people who loudly denounce generative A.I. and say they will never buy a book with an A.I. cover, and I believe them. But are these loud voices representative of the book-buying population as a whole? The science nerd in me couldn’t resist. I had the perfect way to test the hypothesis that readers are more likely to choose a book with a hand-drawn cover over an A.I. cover.

I set up an A/B test campaign in Facebook (Meta). One ad had the new cover, the other had the old cover. The headline and other text used was the same on both ads, and I let it run. After three weeks, here are the results:

	Clicks	Cost per Click	Cost	Impressions	Books sold*	Estimated Royalties
A.I.	321	$0.19	$60.72	6602	9	$16.25
Hand-drawn	393	$0.15	$57.75	5085	12	$20.07

*Books sold are measured across all formats. KENP pagereads were divided by pagecount per book

So the hand-drawn cover wins, but it’s not a slam dunk. There is a slight preference, but I’ve only sold three more books with the hand-drawn cover than the A.I. cover.

Now, I lost money on those ads, (about $82) which was fine. I viewed this as market research and not an actual opportunity to try to promote my books. I’ll let it run a bit longer to see if the numbers move over time, but I’ll cut my losses at $100. 

I’m launching A.I. Human in September, and I plan to make a heavier market push then than I made for Ahelno. So while I was learning ads, I decided to test the two most well-known marketers’ advice on Facebook ads.

A caveat: Every blog you read will tell you not to invest in ads when you have only published one book. I am not here to disabuse you of that notion. I decided to test ads as a learning opportunity, and I fully expected to lose money. Again--market research.

The two methods I tested were David Gaughran’s Facebook Ads for Authors on Youtube and Matthew Holmes’ Facebook Ads for Authors: A Complete Step-by-Step Guide on Kindlepreneur.

Both are useful guides in learning how to set up Facebook Ads. One yielded better results, but one held the key piece of advice that revealed the true winning strategy.

I won’t go through a side-by-side comparison, but here’s the key takeaway on the guides:

Holmes recommends setting up an automatic campaign and letting Facebook’s machine learning do the work to find your audience and drive traffic. Gaughran recommends ignoring all of the Facebook automations and setting up a manual campaign, where you target your audience more narrowly.

I set up two campaigns: one following Holmes’ method, and one following Gaughran’s. I uploaded a set of five ads (some using the hand-drawn art, and some using the A.I. cover and other A.I. images). I gave both a campaign budget of $10 per day.

For the Machine Learning campaign (Holmes’ method), I allowed Facebook to choose where to place ads, and to dynamically rotate through a few different headlines and copy to determine which worked best with audiences.

For the Manual campaign (Gaughran’s method), I set all the parameters in stone. I also set the engagement metric to be link clicks instead of the Facebook default of landing page views per his advice.

First, the engagement results.

	Clicks	Metric	CPC	Cost	Impressions
Machine Learning	1636	Landing Page View	$0.10	$169.21	14,004
Manual Campaign	914	Link Clicks	$0.17	$151.91	14,963

At first glance, the Machine Learning ad is far and away the clear winner. Theoretically, Landing Page Views are a more valuable metric. This shows people who went to your page and waited for it to load. Link clicks just mean people clicked on your ad. They might have clicked on the ad to expand the text, or clicked it by mistake without actually going to your landing page. Therefore, the fact that 44% more people actually went to the landing page on the machine learning ad is truly impressive, and the cost per click (CPC) is much lower. So Facebook’s Machine Learning campaign was both more efficient and more effective. Or was it?

Holmes’ method seemed to produce better results, and his method also included something Gaughran’s video did not discuss: Amazon Attribution links. People may click on your ads, but do they actually buy the book? Without knowing your conversion rate, you don’t truly know how effective your campaign is, and whether it's profitable. 

Gaughran’s video recommends pasting your book’s Amazon page directly into the ad, whereas Holmes includes a video guide on how to set up an Attribution link through Amazon for each ad. That way you can track which ad resulted in a sale. People who click on the link will still be taken to your book’s Amazon page, but if they purchase the book or download it on KDP, that data is recorded. And that data told a completely different story.

	E-books & Paperbacks	KENP read	Total books*	Cost	Royalties
Machine Learning	4	18,668	49	$169.21	$98.29
Manual Campaign	16	33,403	98	$151.91	$191.43

*Total books includes all formats added together, with KENP pagereads divided by pagecount per book

Thanks to the Amazon attribution links, I can see that despite the more impressive engagement numbers in the prior table, the Manual campaign sold twice as many books as the Machine Learning campaign, and even turned a small profit.

If I hadn’t added in the Amazon Attribution links, I would have assumed the Machine Learning campaign was the more effective one, and would probably have used that method when it came to my next book. 

Another benefit of the Attribution links is that I can see which individual ads are resulting in sales and pagereads. In both campaigns, Facebook determined that one ad was the winner and basically showed that one and not many of the others. But I could see from the numbers that there was another ad that was equally as effective, so after a while, I turned off one ad in both campaigns to see how the other one would perform, and about half of my sales came from that one. I would not have known that if I hadn’t used the Attribution links. This also prevents ad fatigue, so you can draw a campaign out longer.

Now, as previously discussed, I lost money overall on the ads, which I expected. But, for $113 (my total loss), I got my book out to 147 more people, and this resulted in 15 more star ratings per platform, a few reviews, and ten preorders for my next book! I call that a win.

Before ads (4 weeks of sales)

Books sold: 128

Amazon:  4.2 stars, 16 ratings, 11 reviews

Goodreads: 4.04 stars, 23 ratings, 14 reviews

After ads (includes prior sales + 3 weeks of FB ads)

Books sold: 388

128 prior month, 147 due to FB ads, 113 new organic sales

Amazon:  4.1 stars, 32 ratings, 12 reviews

Goodreads: 4.08 stars, 38 ratings, 16 reviews

So when it comes to the next ad campaign when launching my new book in September, I will follow David Gaughran’s method, but with the addition of the Amazon Attribution links as recommended by Holmes.**

If you have any valuable insights to share from your own journey with Facebook ads, drop a comment below!

**This is, obviously, one small case study. Your mileage may vary.