Here’s an interesting post about AMS.
More product searches start on Amazon than anywhere else, even Google. It’s the world’s biggest bookstore and by far the largest ebook retailer.
But Amazon Marketing Services (AMS) is still very much a work-in-progress, particularly the slightly pared-back version authors get to advertise books.
Self-publishers tend to focus on making books visible on Amazon. Aside from being a market leader, and having famed frictionless purchasing, there is another key reason why such a focus often gets the best return. Unlike other popular sites, anyone visiting Amazon is generally there for one reason: to buy stuff. You aren’t interrupting them while they share dank memes with friends, or search how fast a raven can fly during winter.
AMS is often referred to as “new” but it has been around for more than two years now. While AMS offers a variety of ads to third-party sellers which can increase app downloads, drive traffic to websites, or boost sales, we’ll specifically focus on the bits open to self-publishers: the Sponsored Product ads and Product Display ads for selling books.
AMS has seen an explosion in popularity this year, with a range of courses and webinars and books and workshops all promising to teach you how to be an AMS whiz. They are probably all over-egging it at least a little bit, because the platform is fundamentally under-developed, and hasn’t changed much from what was first launched in beta a couple of years ago (and I’m told it has implemented little of the feedback provided by beta testers).
Success on AMS is tricky to attain, frustratingly fleeting, and difficult to scale. There are some pretty basic flaws with the system that are holding us back from becoming better advertisers.
The marketing world is filled with talk of Amazon plans to take on the dominant duo of Facebook and Google in the advertising space. I have a bit of experience in this area, having previously worked for Google AdWords and managed multi-million dollar campaigns. I recently refreshed my AdWords certification to make sure those skills are up to date, I’ve been running Facebook campaigns for several years now, gaining a pretty good understanding of that endlessly complex platform, and I have gained some good experience recently with BookBub ads also.
One thing is clear: Amazon has quite a bit of work to do if they are going to convince large brands and big marketing agencies to spend significant portions of their budgets on AMS.
I think we can assume Amazon isn’t done iterating, though, and with that in mind I’ve pulled together some suggestions on how to make things better for everyone – authors getting a better return when advertising their books, Amazon themselves through increased advertiser spend, and readers too in the form of more relevant and useful ads.
Some of this feedback is quite negative – actually nearly all of it is very negative indeed – but the sole motive is to get features and tools which will allow me to be a smarter advertiser, one who can spend more at AMS. I hope it’s taken with that spirit in mind.
One of the first obstacles you might run into is ads getting rejected, often for unfathomable or inapplicable reasons, and any appeals regarding same can take seven days or more to process as AMS has no actual direct support at the moment, and everything must go via KDP – and customer service agents there don’t seem to know the policies very well.
This isn’t good enough. We’re spending enough money on the platform to get actual support. Remember, we’re not suppliers in this context, we are customers of the AMS platform, and Amazon’s famed customer focus is a little blurry here. We shouldn’t have to wait seven days for a response to an email. It usually means the money I planned to spend on AMS gets spent on Facebook or BookBub instead.
Some examples of how this plays out in practice:
I was split testing different creatives for Let’s Get Digital. One focused on how it was endorsed by many other writers, and I had some phrasing like “the guide that bestsellers recommend.” That was rejected as books can’t claim bestseller status without it being proven. But the ad wasn’t claiming bestseller status (even though it has actually sold enough), it was claiming the book was endorsed by bestsellers, which is demonstrably true, and visibly so on the product page. Anyway, I went around in this loop with Amazon, with seven days between responses, and wasn’t getting anywhere so I just changed the ad. The next ad text said it had “over 300 five-star reviews” and that was rejected for referring to reviews “which are dynamic and can change.” Maybe that was an edge case anyway, fair enough.
An author friend has written a romance novel which has the words “Rock Star” in the title. But her ad was rejected and her appeals ignored on the grounds that Amazon has rules about referring to the star-rating in ads. At this point you would be forgiven for thinking that Amazon is testing out some kind of customer service AI, with decidedly mixed results.
2. Making Changes
After you have set up your campaign and navigated any potential rejection issues, you may realize you made an error in your ad text, or you might have thought of something snappier. If you want to go back and edit your creative, sorry, that’s not possible. You have to start a brand new campaign, and lose any positive click/purchase history on your keywords. Everything should be editable once a campaign is live. This is pretty basic stuff, something a smaller platform like BookBub can easily handle, let alone Facebook.
And it’s not just ad text. On Amazon, you can’t change your campaign name once it is live, meaning you can be left with the garbled nonsense that inexplicably forms the default text. You can change your bids, but you only individually. When you have 1,000 keywords in a campaign, this can take quite some time indeed.
As a result, people often just copy the campaign and start again. But then the new campaign will have wildly different results, on all metrics. Impressions, clicks, and sales can all be dramatically different for reasons that are completely indiscernible, which I’ll talk about more below.
3. Reporting Delays
As anyone who has ran an AMS campaign will know, it can take three days or more before sales will appear in your ads dashboard. Problem is that clicks (and related costs) come in sooner than that, and impressions come in even quicker. You might see impressions on the first day and clicks on the next, but often have to wait until the third day to see if that exposure and spend is turning into sales. This makes it really tricky to measure success.
Maybe this isn’t an simple problem to solve (but I’d note BookBub does it far better). It would be better if all numbers had the same delay. I don’t mind as much if it takes AMS three days to report impressions. If clicks, sales, and impressions are coming in at the same time, then I can measure things properly. Right now, it’s impossible.
4. No Customization
AMS requires a lot of trial and error, and that when you get something to work, you are often loathe to touch it in case it stops working. This means you can have several pages of campaigns, with most of them defunct and no way to archive them, or organize your account in any useful way. Even crudely ordering the dash by Campaign Status will put the Terminated accounts on the top, meaning the active ones can spill onto the next page.
What we really need is more tools to see the data we need in an instant, so we can make smarter decisions (i.e. spend more money on ads that are more relevant to customers). I should be able to filter Active campaigns only. I should then be able to slice and dice those campaigns and order by ACoS or Impressions or Spend or whatever I want. Again, this is pretty basic stuff that every competing platform can handle. AdWords, for example, had this functionality ten or fifteen years ago.
5. No Filters
On that note, I should also be able to filter my dashboard by date so I can see what served today, or what my spend was this week, etc. At the moment, everyone is flying blind. We only get cumulative lifetime numbers, which is crazy! To find out what you spend in the last week, or yesterday, or in July, you have to manually track those numbers as you go along. There is no way of doing it after the fact.
Again, this is standard at Google and Facebook and BookBub and anywhere else that has a paid advertising platform. Why Amazon doesn’t provide this basic data is beyond me. Again, it’s preventing me from making smart decisions, and the dumb money is flushing out the savvy advertiser – threatening the long term-health of the platform, for everyone, including Amazon. I could track it manually, but I just don’t have the time. Which means I either don’t do it and make bad decision on AMS… or spend the money elsewhere.
6. Sales & ACoS
This is a big one. Sales and ACoS are the two metrics we use to judge the success of our campaigns. But they aren’t reliable. ACoS is just a function of Sales, so I’ll focus on the latter.
We are given just the raw sale price of any books sold as a result of an AMS ad. So if my book retails for $2.99, that’s the number that will appear in the Sales column, regardless of whether I received 70% royalties or 35% royalties for that particular sale, or whether delivery fees were deducted, etc. Needless to say it would be much more valuable to get a number related to what I’m actually receiving for that sale.
Much more serious is the problem related to paperback sales. I could have an e-book retailing for $4.99 and a paperback retailing for $12.99. All of these sales are just thrown together. So I could see Spend on one campaign of $16 and Sales of $25.98, and conclude that the campaign is profitable. However, if that figure relates to two paperback sales rather than a number of e-book sales, I could be making a huge loss. And because we only get cumulative lifetime numbers (which is completely nuts), it becomes impossible to tell over time whether the sales number contains paperback sales which are hugely skewing the numbers.
The net result is that we end up wasting lots of money on unprofitable campaigns. This is unsustainable for Amazon, and means readers are getting served more untargeted ads than they should.
7. Information Vacuum
We are constantly guessing with AMS. Did this previously successful and ROI-positive campaign stop working because it had a low CTR overall? Will it restart if I prune the worst keywords? We never really know for sure.
Facebook ads can stop mysteriously too, but it’s usually a case of fiddling with a few things to get them running again. With AMS, this process is maddening. You can go through the laborious process of upping your bids on hundreds or thousands of keywords, and you will have to wait three days to see if that has done the trick – and have to manually compare the numbers to see if any effect is present, because all we get is cumulative lifetime data. (Have I said how crazy that is?)
And even if you go through all that rigmarole, you might just find that you misdiagnosed the reason, and that your ads haven’t resumed their previous level of serving. A standard process of elimation that can take a few hours on Facebook or BookBub can take days or weeks on AMS. Often you just give up and start a new campaign. Which then has completely different results for equally indiscernible reasons.
I know that Amazon is famously tight-lipped. I know how tech companies operate. They want all your data, but don’t want to share any of theirs. Even their data about you!
But, really, some more information about how AMS works will make us better advertisers. When I worked at Google they were just beginning the process of being more open about how everything worked, and now they have a full outreach effort teaching advertisers how the system works and what best practices are. I hope Amazon goes on a similar journey – it really does benefit everyone, without needing to compromise any proprietary information.
8. More Relevancy
Did I mention I want to be a better advertiser? This is important, not just so I can get a good return and add to my already impressive collection of Fabergé eggs, but also so that we are all serving more relevant ads to customers, which will increase user trust in the ads, and increase CTRs, and make the ads more viable, and make everyone more money, including Amazon.
I’m sure relevancy is a factor in the ad auction somewhere, but because AMS is a total black box, I don’t know what exact role it plays – which makes it harder to optimize my campaigns. Is it better to have multiple 1,000 keyword campaigns, for the same book, targeting every tangentially comparable author and title, and to prune as I go? Or is it better to regularly start new campaigns with all my keyword winners from each campaign into one keyword supergroup?
I have my suspicions that the relevancy or quality score the system assigns is at a keyword level rather than at a campaign (or account) level, but can anyone really say that for sure? There could also be campaign-relevancy weighting applied at a lower level. Or I could be wrong. AMS isn’t easy to figure out.
I can say this though: whatever relevancy is built into the system isn’t weighted heavily enough. The penalty for being a bad advertiser is too low. (By bad advertiser, I mean someone serving untargeted ads – taking a scattergun approach and just targeting anything and everything.) Which means the dumb money is flushing out the good advertisers. Which means the ads will get less relevant and more expensive over time, which means users will click on them less, and so on. I’d argue this process is well underway already.
Increasing the relevancy factor in deciding which ad is served will have the opposite effect. It’s the long-term view, the one I hope Amazon will take, and the one which will cut across their core proposition less – i.e. the aim to show customers the products they are most likely to purchase.
You know what else could help? Some more options other than just “broad” keywords. I think “exact match” options are available in the wider AMS system, it would be good to see that here too, given that particular innovation hit AdWords in the early 2000s.
9. For The Love of All Things Holy: Borrow Data
This is tricky for Amazon for two reasons: first, it’s something that would have to be engineered specifically for us. Second, KDP doesn’t currently give authors borrow data. We get reads, but can only estimate borrows. And not very accurately either.
That said, we really need borrow data. The key metrics of Sales and ACoS are already a little misleading for reasons mentioned above, but being in KU ads another twist. Borrows are a function of visibility, so AMS ads can and do have a positive effect on reads in our KDP dashboard. But we don’t get any metrics in AMS related to them. Ebook sales, paperback sales, and that’s it. No borrow data – and reads often make the difference between a profitable campaign and a money loser. At the moment, we can only guess, which means we are probably often killing good campaigns and letting bad ones roll.
AMS has huge potential. Huge. But it hasn’t developed much since launch. Reporting is like an early beta that should have been updated before going live. It’s amazing that the only data we get is cumulative lifetime numbers. It makes it difficult to optimize, and hugely time consuming too.
The tools we do have just aren’t fit for purpose. Most of the major issues above surround problems related to reporting, presentation of data, and general usability – basic issues which should have been resolved already at this point.
AMS is also too much of a black box. Amazon doesn’t really share how the system works in any kind of detail – it’s all pretty vague and opaque – and we can’t figure it out because we aren’t given enough data. I’ve been running AMS ads all year on and off, and I’m not entirely confident I can say that I’ve gotten that much better at using the platform – which stands in marked contrast to BookBub, and even Facebook, despite it’s incredible complexity and constant changes.
If we had more data and better tools, everyone would benefit – Amazon and readers too.
* * *
I’m very open to the possibility that I’m a dumbass, so if I’m talking nonsense above or you have solved any of these issues, or have any general AMS tips, I’m happy to hear them in the comments!