I’ll share a few testing tricks for Amazon ads from my bag of tricks. I used them with some degree of success to advertise my books, 100+ of my customers’ books, and sold tens of thousands of copies with profit.
Did I get your attention? Smokin’!
So, let’s start with a controversial statement: testing your Amazon ads makes little sense. That
This wonkiness is the real reason you don’t want to test your ads. The data AMS serves authors are unreliable at best and misleading at worst.
The first thing you need to do before even attempting to test anything is to get educated about AMS metrics. I wrote a whole monster post about this:
The second thing you need is time. I’ve never seriously tested my campaigns. I’m fine with my meager 100% ROI, so I don’t have a clue how much time you really need to tweak them to 500-1,000% ROI range. My best guess is a minimum of one hour a day, every day.
The third thing you need are keywords. A lot of them. Thousands. Or tens of thousands. Or hundreds of thousands.
There are other types of ads on Amazon than keywords ads. They are basically useless. They may work, but you cannot test them. Not in a sense I have in mind – to get meaningful data that you can use to repeat or enhance the experience in the future.
How to get keywords? Publisher Rocket is the best tool for this purpose in the whole universe.
It’s against some advice from gurus I’ve seen about testing. “If you bid low, you will get no traction and you will have nothing to test.”
It may or may not be true. Yes, you will most probably get traction if you bid high. By traction I mean impressions – the fabric of Amazon ads.
However, high impressions with high bids mean nothing. Every moron will get impressions if he highly overpays for them.
Besides, most authors and 99.99% of first-time authors are dirt-poor.
High bids are the quickest way to eliminate yourself from advertising.
You will spend a fortune even before you start testing, so you will get very little in return and quit.
And this “don’t bid low because you will get no result” is a big myth. I’ve been advertising well over 100 books and I always bid low. Whereas my results are not very impressive (100% ROI) they are very stable.
You may get close to zero results only when you test just a handful of keywords.
The starting bid? 10 cents. Yes, you can still get impressions for 10 cents on Amazon. If you feel adventurous or your book’s royalties per copy is above $4, bid 12-13 cents at the beginning. That’s all.
You see, when you bid that low, you average Cost Per Click will
It gives you a space for errors. Even if your Kindle costs only $2.99, you will get about $2.04 per sale. So, you can afford 20-29 clicks per sale. If your conversion rate is worse than that, you still won’t lose much money, maybe half a dollar on a sale. But if you would’ve bidden 30 or 50 cents per click you could’ve lost a fortune on a single sale.
Big Data, Not Flukes
The way Amazon ads and their reporting work makes analyzing the data on the campaigns level impractical. What you test, at least in the initial stages, is the overall performance of your book:
Does it get impressions? Does it get clicks? How many impressions convert into a click? How many clicks convert into a sale?
You set the baseline to measure your future campaigns against it.
The baseline shouldn’t be based on a single campaign or
several of them. I usually run
106 128 campaigns for one book. I can
easily find for almost each book campaigns that are bummers or great
performers. But it doesn’t mean the book is a drainer or great performer.
With one or a few campaigns you can quickly fall under the illusion: “Wow! This book sold 10 copies for 38 clicks! It’s a sure bestseller!”
Nope, it’s not. Someone simply bought 10 copies of a paperback. You got lucky. It means a single buyer out of 38 clicks. This is a terrible conversion ratio.
Another common illusion:
“Oh, there is no ACoS, this campaign is fine, unlike those losers with 3-digit ACoS.”
Yes, 3-digit ACoS is usually bad. But no ACoS ain’t success. It depends on the data.
No ACoS means only one thing: there were no sales for this particular campaign.
If you got only 5 clicks, it means nothing. You didn’t get enough data to draw any conclusions.
If you got 500 clicks and no sale, it’s a disaster. All those 3-digit ACoS campaigns are better than this drainer.
Don’t look at the particular campaigns’ metrics. Sum up their metrics and analyze the big picture. My rule of thumb: anything below 100,000 impressions is not enough data.
With 100k impressions you should have 40-50 clicks, more if your title, cover and blurb are great. You should get a few sales out of so many clicks. You have something to work with.
100k impressions are fine, but 1,000,000 impressions are better.
The more data you have, the better decisions you can make. The more the data you gather, the more you know about your book and how people interact with it. Getting a lot of impressions, or great conversion rates or plenty of sales with 1 or 3 campaigns may be misleading.
You might’ve chosen the target audience very well with keywords and that’s why you got those good results. But when you try to scale up with more keywords, the success story may abruptly finish.
I prefer to get decent general results and only then to fine tune the ads. However, 99% of the time I just stop on getting decent general results because fine-tuning takes so much time!
Having a million impressions, you should get at least 400-500 clicks. That means 400-500 people landed on your book page. How many of them bought?
If those conversion rates are good, meaning you pay less for clicks than you earn from sales, you can play with the advertising process without pressure. Add more keywords, tweak the ad’s blurb, tweak the headline or Editorial Reviews.
This break-even point gives you a lot of leeway with your testing.
One Change at the Time
In theory, you can change one thing, for example the ad’s blurb and observe the results of your action. In practice, this method is far from foolproof. Yes, you’ve changed one thing, but Amazon changes the game’s rules all the time and they never (or VERY rarely) inform you what they changed.
So, you can change the blurb, and at the same time Amazon’s algorithm will decide that your book is no longer marketable. It will get only the fraction of impressions it used to get in the past. You think it’s a fault of a new blurb, but in reality it’s just the whim of the Juggernaut.
By the way, I’ve seen Amazon change its mind about a book multiple times. In most cases it means book’s decapitation – impressions plummet and plummet trying to reach zero – but I’ve also seen cases when a book that wasn’t getting much juice suddenly got an influx of impressions across all the campaigns.
So, you have enough unknown factors on your plate. You don’t need to add complexity by changing more than one factor at the time. Tweak one thing, let your ads run for at least a week and only then try to draw conclusions.
Why one week? It’s back to #2, to gather enough data. If your book gets a million impressions a day, well, one day of test may be enough.
Never Adjust a Bid Down
As far as I can tell, it’s a signal to Amazon to turn off this specific keyword altogether. Whenever I modified the bid down, the keyword stopped getting impressions.
On the other hand, adjusting the bid up is like adding fuel to a fire. It works.
Bid up the Winning Keywords
Thus, my winning testing strategy is to modify the bids of the keywords that are hits. I use two ways to that goal:
a) Increase bids on particular keywords
The problem with this method is that it’s very rare to find
a single winning keyword. You need a significant amount of data to decide the keyword
is worth doubling down.
What’s the significant amount? About 100 clicks. If a keyword got 100 clicks and made enough sales to break even, I’m willing to increase the bid to see if it is an accident or a rule.
The great thing about this method is that I don’t have to create a new ad which requires work and time.
b) Create superads
It’s so much faster to do. You download the data from old ads, and narrow down the keyword set to keywords getting the most impressions and/ or clicks. Of course, if you have the keyword with 100+ clicks and one or none purchases, you don’t include it in the super set.
Having the set, I create a new ad with higher bid. Amazon always has higher bids in higher esteem, so even if I use the same keywords in old ads Amazon will choose the superad to display.
Compare Apples to Apples
This is, by far, the trickiest trick among my tricks. If Amazon’s system wasn’t so wonky, it would’ve been easier to pull out. As things are right now, all you can do is to restrict the guessing to a minimum.
First, to compare anything, you need data. Never ever try to judge your ads performance by your gut feelings: “Oh, it seems to work fine!”
Download the data. Aggregate them. Put them into a sheet. Adjust your ads by adding one new factor (higher bids, different keywords, change of blurb, etc.). Let the ads run for a few weeks or however long it will take to gather enough information.
Download the data. Aggregate them. Put them into a sheet.
Compare the data. Draw conclusions.
This is how the process works. Violate just one step in it, and you can put all your data and analysis into a trash bin. You can as well take a crystal ball and look for cues if your ads will be profitable, or not.
Don’t compare Sponsored Product ads to your Lockscreen ads.
Don’t compare your romance book’s performance to my personal development guide’s performance.
Don’t compare 10-cent bid ads to 15-cent bid ads; unless, of course, this is the change you are testing right now.
Don’t compare the UK market to the US one. And, for God’s sake
Everything matters. You need to be a maniac of diligence to test Amazon book ads properly.
Testing means you always compare your ads’ performance after a change with the baseline. You need to establish a baseline in step #0. The first set of ads is your baseline.
Yes, you will have to make some wild assumptions sometimes; guesstimate; set hypothesis based on your gut feeling or on what other authors say. With AMs it’s unavoidable. But whenever you can, base your analysis on facts and data.
The Quick Summary
Bid low. 10 to 20 cents is enough in the beginning. The less, the better. It’s better to have no traction than to overbid and take yourself out of the advertising game at the very beginning.
Use big data for your analysis. Anything below 100,000 impressions is iffy. Anything below 40 clicks is iffy. Sometimes it’s all you’ve got, but don’t bet your budget on analysis based on such a meager data set. More is definitely better in this case.
Change only 1 factor at the time. The AMS system is mysterious enough even without complicating things by adding multiple layers of complexity.
Don’t bother adjusting the bids down. Leave them or turn them off.
Scale bids up. When you do that, do it gradually. Increasing the bid by 2 cents will not change much, but won’t also skew the whole game. If you increase a bid by 20 cents, you may max it out and compete for the top spots. The game at the top spot is very different than in the rear ranks.
Bid up single keywords – that way you don’t have to create a new ad from scratch. Remember to keep the data snapshots at the keyword level if you do this.
Be obsessive about data. In the end, this is what testing is – comparing one set of data to another and making informed decisions.
Take my advice with a pinch of salt. I’m not really an ads tester. If my ads don’t work, I prefer to give up instead of looking for ways to improve them. It’s faster. My performing ads make me enough money for living and I prefer to spend my time on other things.
Nonetheless, I can assure you that the above tricks will save you some time and a lot of money. If you are a newbie in the book advertising game on Amazon, use my tricks before you turn into something more advanced.