Amazon Ads Case Study

Amazon Sponsored Product Ads: Automatic Targeting vs. Manual Keyword Targeting [Case Study]

People have asked me the same question for the last two years.

Hey Man Have You Done Any Amazon Campaigns?

Up until a few months ago, my answer had always been the same.

“No, but I want to.”

That seems to be the answer for most of us — us guys who spend the majority of our day in Google Ads. We know Amazon advertising is the next step. We just don’t know how or where to start.

That’s why when my old high school classmate, Dan, messaged me saying he was doing stuff on Amazon, I was stoked.

We got together at a local bar and grill and caught up on all that had happened since we last ran into each other at a dance party in college (Dan’s life is more exciting than mine). Then we talked about business.

It was a short conversation.

Have you ever had one of those conversations when every single thing the person said made you think, “You really need to talk to [Mark, Britt, Tyler, etc]”?

Yeah, that’s what happened.

“You really need to talk to Jenny,” I said.

I call Jenny “The Amazon Woman” — not because she’s freakishly tall (she’s 5’7″-ish), but because she’s an Amazon SEO and CRO wizard. I don’t know the details of how she does it, but she makes products rank and sell on Amazon, and she’s really good at it.

I introduced them to each other then got out of the way so Jenny could work her Amazon-magic.

A Few Months Later

Fast forward a few months — Dan’s making a lot more money.

He’s crushing it with SEO and has a budget in mind specifically for sponsored product ads. He tells me that he’s been running some of his own campaigns with automatic targeting. I tell him I don’t know what that means. He describes them to me. They sound like Smart campaigns (Google’s automated campaigns formerly known as AdWords Express, tactfully rebranded as smart campaigns, because you need to be stupid to use them). I say if they’re anything like Smart campaigns, he’s better off trying his luck in Vegas.

But he says it’s working, so I look.

He’s right.

Amazon Ads Automatic Targeting Campaign Performance Results
Performance for Dan’s primary auto-targeting campaign. Sorry this looks bad on mobile. Click here to see it full-width.

I don’t know much about ecommerce. I just know you want a low ACOS and high ROAS. High and low are relative to your profit margins, but I know Dan’s, and this is profitable for him.

After looking at his auto-targeting campaigns, I asked Dan what he considered to be a good ACOS — in other words, if I built him a manually-targeted campaign, at what point could he wake up in the morning, log in to Seller Central, look at his ACOS and feel content that his business is in a good place.

He said anything under 20%.

I told him I’d do 10%.

The Process

Building an Amazon campaign with manual targeting based on data from an auto-targeting campaign is the same process as creating a better Google Ads account based on an existing one: You identify what works and cut whatever doesn’t.

It feels like cheating. Once I found out where to download search term reports, I knew exactly what to do:

1. Identify winning keywords.

To be specific, I mean identify recurring words that, when the data is filtered to only include those words, have a low ACOS and high ROAS.

For example, if you’re selling water bottles and these search terms are driving sales:

Best water bottles for hiking

water bottle for hiking trip

gift water bottle for hiking friend

It would make sense to target hiking water bottle as a keyword.

2. Identify negative keywords.

This is the same process as identifying winning keywords, but with inverse logic — which words have a high ACOS and low ROAS? Better yet, which words are getting clicks but aren’t generating sales at all?

Using the previous example, if you keep seeing hiking show up in search terms and your water bottles were meant for the gym or office, it might make sense to add hiking as a negative keyword.

3. Group keywords into similar themes.

I was able to determine four ad groups after sorting the winning keywords into similar themes. Here are three of them.

– Brand. I don’t know water bottle brands, but if Fiji or Evian sold reusable water bottles, the keywords in this ad group might be fiji water bottle or evian water bottle.

Some people think paying for clicks on brand searches is a waste of money. I think it’s an easy win and a cheap insurance policy in case competitors start bidding on your brand.

– Categories. I can’t go into much detail on this without revealing Dan’s niche. Another name for this ad group could be “types” — for water bottles, this ad group might contain keywords focusing on different water bottle colors, or sizes, or the material they’re made of (plastic, metal, glass, etc).

– Competitors. In this ad group I added keywords targeting searches for Dan’s competitors. See why it’s a good idea to bid on your own brand?

4. Determine bids based on historical conversion rates and target ROAS.

I told Dan that I would get him a 10% ACOS, so I determined bids based on conversion rates from the auto-targeting campaign with that goal in mind.


The first few days were ugly, like many campaign launches are. The campaign wouldn’t spend because I set the bids too low. After gradually increasing bids over those first few days, we finally spent money — and for a while the ACOS was just under 10%.

I could have kept it there, but Dan was missing out on sales. He said as long as ACOS was under 20% he’d be happy, so I bumped them up even higher. The ACOS increased to about 15%. Total sales increased too.

Amazon Ads Manual Targeting Campaign Performance Results
Performance for my manual targeting campaign built using Dan’s auto-targeting campaign data. Click here to see it full-width.

Here’s a side-by-side comparison.

Amazon Ads Automatic vs Manual Targeting Performance Comparison
Skeptics will say I skewed the numbers by allowing brand and non-brand data to coexist in the same campaign, and if we were spending more than a few hundred dollars, I would have broken them out separately. But delivery for brand queries stayed flat (I didn’t overdeliver to inflate the numbers). This data is clean. Here it is full-width.

So how much more efficient are manually-targeted campaigns than auto-targeted campaigns? Well, I doubled CTR and cut ACOS in half. Dan used to spend $30 to make $100 — now he spends $15 to make $100. That sounds like a lot more efficient to me.

Which Is Better?

That’s the question, right? Which targeting method is better: automatic targeting or manual targeting?

The results here indicate that manually-targeted campaigns drive better results, but it’s only because of data from the auto-targeted campaign that I was able to build such a winning campaign. If I had to build something entirely from scratch with no historical data, It wouldn’t have gone this well.

I think a better question is if you’re a total newbie building your first Amazon campaign, which targeting option should you choose?

My advice is this: Use automatic targeting first. Let Amazon find out what works, then use that data to do it better. Prioritize the derivative manual-targeting campaign first, but keep an auto-targeting campaign going to find new keyword opportunities.

What I Learned

This case study took a few months to write. Throughout that time I asked several people what they knew about Amazon ads. Most of them didn’t know anything. The few who did seemed convinced they had found their own Holy Grail and were skeptical I could do better.

The thing is, this wasn’t hard. Anyone with intermediate Google Ads experience could do this and charge a management fee based on the money they save clients, like a good accountant or financial advisor with taxes. This service wouldn’t be practical at Dan’s size. He now saves a little over $200 a month in ad spend — a moderate expense for his business, but not enough for someone to turn into a hustle. I did it for free, for the opportunity to learn and write this, and because he’s my friend.

You’d need whales — big clients with big money. I’m sure there are agencies out there already doing this and I’m just not familiar with them. Otherwise, to work with the Dans out there, you need to know what Jenny knows too.

But if you know all that, why not do it all yourself?