Wednesday, September 25, 2019

Hacks for Amazon Marketing

After my last post, I had an odd spike in ad expenses with no return, and I wanted to know why. So I did another report on my AMD ads experiment, with a breakdown by ad placement. It showed me that 35% of my recent ad budget was completely wasted on First page Top placement, with not a single sale. When looked for a way to adjust the number of ads by placement to eliminate this, I found none. AMS customer service said that they will take it under advisement as a new feature in the future.

Placement Impressions Clicks Cost Per Click Spent Sales  average cost return orders percent budget
Product pages 56792 78  $   0.61  $ 47.54  $29.90  $   4.75      0.63 10
Rest of search 21624 34  $   0.65  $ 22.10  $23.92  $   2.76      1.08 8
First page Top of Search 2113 31  $   1.23  $ 38.18  $     -  


Until then, how do I avoid flushing so much money?

1. Product Pages Only

Well, Rest of Search has no knobs, but I can set the base bid at 10 cents, and after the ad is created, go to the last line of the "campaign settings" tab and increase bids for Product Pages only by 800%. In this way, I can ensure that no First impressions will be made and I can bid on my old keywords for Product Pages only for under 80 cents. I can then adjust the bid per keyword for this submarket. 
The first day only cost me $1.66, but I haven't seen any purchases yet. After an initial burst of 2000 impressions, Amazon throttled the exposure to 3 impressions per hour, too low to get any clicks. I suspect after the first day, it favors campaigns that produce a profit.

2. Rest of Search

If I want Product Pages and *some* of the Rest submarket which as twice as lucrative, I could set my default bid to 65 cents (the average cost per click of that category) and only bid down. The max bid is half the average of the First page prices; however, on the first day of this test, the campaign somehow reset itself to "dynamic up and down" and generated three useless $1 "top" clicks to go with the 3 sales that the other categories earned. I set it back to "down only" and will retry. Of the 13 settings that you can change, Strategy is the only one that has its own off-screen Save button, so be careful.

Unfortunately, when you go "down only," even if you keep increasing the bid, it only gives access to Rest markets 10 percent of the time. It feels like I am being penalized for not giving Amazon free-rein with my budget. Indeed, in the five days after the change, I only got 2 clicks total. Up until the change, I would have expected 65 clicks for the same period. Since it takes an average of 6 clicks per sale last week, I sold nothing.

3. Skipping Days Manually

With years of Amazon reporting data to go on, I know that Tuesdays are my worst days (and Sundays are weak). Looking at Amazon ad data, none of the clicks generated revenue on that day of the week. So, I will use the pause button on the campaign to manually prevent the expenditure on that day. Reducing my ad budget 14 percent with the same monthly purchase rate increases my profit.

4. Manually Enabling on Key Days

From my sword-and-sorcery series, I know that Monday afternoons were my biggest time for purchases, making almost as much as the rest of the week combined. Therefore, I only turn my Doors campaign on for that day. Make sure that the ad duration lasts through the desired dates or turning it on the night before will only get you an expiration message when you try to check the stats the next day.

Tuesday, September 17, 2019

AMS Ads Lessons Learned


Indie authors would love to wave a magic wand and have ads transform their backlist into best sellers. It doesn’t work that way. For the first week, AMS is more like visiting the doctor or mechanic after going without for years. You pay a bunch of money to find out what needs fixing. Be willing to pay upfront for this valuable information and make changes accordingly. If you’re lucky, you can use the tools available to find a niche for your book that is profitable and satisfies your customers. I know this because I played amateur sleuth like one of my characters. I’ll give concrete examples from my experiments below.

1. Don’t try this with 99 cent books
Your cut of a 99 cent e-book is less than 30 cents. AMS bid averages start in the 35 cent range. The best categories are closer to 90 cents. This means that you’ll never make a profit. Even if you’re willing to eat the huge loss short-term, my one-day comparison experience is that people are less likely to click on a 99 cent novel than a 2.99 option. If you don’t value the product, why should anyone else?

2. Use Kindle Select Novels with a Paperback Option
Although it is difficult to track, novels that subscribe to KDP Select have much better results. Kindle Unlimited subscribers who don’t have to pay extra are much more willing to try your book. Once they read a few pages, they should be hooked. Even with stand-alone novels, I’ve generated almost as much follow-on revenue through KU as during the actual sale of units. My “Doors to Eternity” epic fantasy didn’t benefit from the exclusive Amazon contract the way all my sci-fi novels did, but I flipped it on after a few days to make my ad dollars go twice as far. Note that I sold a higher percentage of paperbacks than normal with these ads. (11% compared to the normal 0.2 percent) Unfortunately, the AMS reports won't tell you which keyword sold your paperback, and the total sales on the summary page will be mysterious more than those on the spreadsheets because of this.
Lastly, I noticed that read-through to the second book in any series made ads even more cost-effective. The higher your read-through rate, the better the effects. Because 67 percent of readers of “Jezebel’s Ladder” buy “Sirius Academy,” and there are five books in the series, one sale has the weight of three.
Below is the comparison of direct sales versus KU. Note the two-day lead time.

3. Manually Target
I’ve never seen an automatic campaign get significant clicks. If it does get clicks, they often don’t lead to sales for the first several days, and you won’t know why or be able to control this. The point of this exercise is for you to learn. That’s the real benefit for what you’re spending initially. If you knew what sold or why people liked/disliked your product you wouldn’t be in this situation. I use an initial budget of $10 a day. Plan on burning that much per book to find out what you don’t know. Try to start out with at least 75 keywords/phrases and add to the list each day as you find out what works. For the first book I wrote, Scarab, I could only think of 29 phrases. Only three got above 50 impressions, and only “Ready Player One” was something other than a sci-fi category name. As much as I thought fourteen-year-old video-game players will like the book, it isn’t really marketable. If I couldn’t think of enough good reasons somebody would want it, no one else will. With all my other experiments, other good keywords kept popping up every day. I was able to use the data to change the price and blurb on my novels to reflect what I learned. For example, 100% of Andre Norton fans who clicked on my Doors ad purchased the book, so I added her to the blurb and my keywords. I also added her to the list of phrases on my YA novel, an even better fit.

4. Be Patient
Data in Amazon takes one to three days to be recorded. Nothing will show up on your sale board for 24 hours. Even if you get sales on your board, you won’t always know which title sold for a while. KU purchases take even longer to register. Furthermore, book sales happen in cycles. For example, my epic fantasy sales dip on Sunday and Tuesday, but spike heavily on Monday at 1, after my target audience of IT guys comes back from lunch and checks their e-mail. You need several days of data before discarding anything. Be aware that the reporting interface is buggy. When you change to request Targeting keywords, it resets the span of the report to LAST month, giving you an empty useless spreadsheet. I recommend the "Month to Date" version.

5. Preconceptions are the First Casualty
However, on day three, you can begin changing the search keywords that you entered in your initial KDP setup for your e-book. The words you thought described your book are probably irrelevant garbage. Now you can prove what is popular and what positively associates with your product. For example: my magic school novel “Tells” had Harry Potter in the description, but everyone uses that label. Only .2 percent of impressions go anywhere. I had no idea that Richelle Mead or Harley Merlin series fans would convert so much better (up to 2.5 percent). The keywords Magic and Intuition were a waste of space, and I replaced them with the two successful ones. For my “K2 Virus,” no one cared about Virus, Korean, or Martial Arts, but Pandemic, Assassin, and Tom Clancy were huge winners. Evidently, buyers wanted a spy thriller more than a research paper. Why customers are interested is something we haven’t been able to see before. Take advantage of what you’re learning to make your book better long-term, not just for the span of a sale. 

6. If Amazon Won’t Spend Your Money, It Can’t Be Spent
After four days, “Quantum Zero Sentinel,” one of my newer novels hadn’t spent even $3. I had very few clicks. Why not? It had a 4.8 rating and had a decent initial sales bump with my regular fans. However, there were several problems. First, the targeted Batman/Orphan Black fans weren’t interested. The more popular authors who resonated were Dan Simmons, Neal Stephenson, and William Gibson. Altered Carbon even had a good showing. Wow! Huge compliments, but I need to radically change my blurb. The category that most closely aligned with the novel wasn’t hard sci-fi, but Spy Thriller, with a whopping 2.5% click rate. My biggest problem? The cover doesn’t say any of this to prospective readers. I’ll have to spend another $200 plus to fix the problem. Fortunately, my Facebook marketing research of alternate images says that more than enough people will buy the novel to pay for this change. In the meantime, thousands of people in my niche were exposed to my brand for next to nothing. Next time they see one of my books, they’ll be more likely to buy.

7. Not All Clicks are Good
False clicks happen, and they hurt, especially early on when you’re adjusting your campaign. Sometimes it’s an accident or a fluke. I had this happen most often with popular titles like “Game of Thrones,” “Warhammer,” or “Star Trek.” You need to reduce the chances of a mistake using negative keywords like RPG, DVD, Movie, or TV so that you don’t pay for someone’s fat fingers. After filtering, the hit rate for several of these is much lower (1 per 3000 views). I eventually got rid of these phrases, too, because even if the Star Trek customers do buy my space opera, it won’t be the experience they’re looking for and my product reviews will be lower. This is an exercise in finding the right audience for your writing. Since K2 Virus has a transgender character, I made that one of the original keywords. However, the types of books that come up when you search that word on Amazon are very specific sorts of erotica—nothing like my novel. I turned this keyword off after several false hits because I don’t want to mislead people. As this decision very subjective, I make it a rule to switch off only one of these false leads a day (or a couple closely-linked ones like Klingon and Star Fleet). Then watch how the revised list performs on your reports tomorrow. Eventually, you’ll have a smooth-running machine.

Here is an actual sample day with my decisions color-coded.

keyword views clicks cost per click spent sales
assassin 961 3 $ 0.99 $ 2.98  $  2.99
espionage 633 1 $ 0.64 $ 0.64
medical thrillers 622 7 $ 1.14 $ 7.98  $  5.98
spy novels 616 1 $ 0.65 $ 0.65
pandemic 589 12 $ 0.87 $ 10.38 0
tom clancy (BROAD) 576 1 $ 1.33 $ 1.33
tom clancy 450 1 $ 1.54 $ 1.54
plague 292 6 $ 0.76 $ 4.55  $  2.99
north korea 279 1 $ 0.63 $ 0.63
richard preston 190 2 $ 0.81 $ 1.62
spy 181 1 $ 0.59 $ 0.59
hard science fiction 83 1 $ 0.84 $ 0.84
military thrillers 47 1 $ 0.59 $ 0.59
outbreak 43 1 $ 0.48 $ 0.48
bioterrorism thriller 39 2 $ 0.46 $ 0.91  $  2.99

8. All Hat, No Cattle
The last case is the hardest. What happens if you get lots of clicks, but nobody buys? Ouch. You just shelled out good money for nothing. I had this happen with “Void Contract.” One of two things might be to blame. I’ll have to apply the scientific method to determine which. First, the image of an African man with a gun on the cover can be off-putting to American consumers. I found with Facebook experiments that British and Indian customers have no such bias. Unfortunately, I can’t test this theory because AMS doesn’t run anywhere but the US, and I don’t want to buy another new cover today. Second, and more likely, is that the detailed pitch is too dull for the audience. I wrote the book as an homage to the 70s author Alan Dean Foster and peace between all species, but today’s military sci-fi fans (Zahn, Dietz, Green, Cook) are the ones buying it. No room for touchie-feelie with these guys. The pitch has to grab them by the nose and shout, “Someone who deserves it is gonna get hurt, and you can watch.” Testing will take a week because first I have to wait a day for the e-book to unlock after I changed the keywords. Then, the new pitch can take another day to percolate through to the Amazon page. Finally, I’ll need at least three days of revised data collection to judge the performance of the new one. Because of the long wait, I need to word-smith each revision carefully. I’m considering the following replacement:
  • Max doesn’t want to be an assassin, but his Xhosan-African genes and hunter training make him invisible to empaths, the perfect weapon for hunting alien war criminals across the Gigaparsec of known space. After they’re all gone, he isn’t sure how he’ll fit into peacetime society. That’s not a problem now that Saurian mobsters have kidnapped his only surviving friend. Since evolved races can’t legally kill sentients, Max is going to have to get creative.
  • Fans of Dietz, Zahn, Vance, Heinlein, Walter Jon Williams, and Babylon 5 should enjoy this action-filled quest to other planets.

I’ll update the post when these results are available.

9. The Long View
For “midlist” authors like myself with a couple dozen novels, AMS works best as a marathon, not a sprint. I don’t want one-day burst of 500 sales, followed by obscurity. I want a steady three sales a day, building my “people who bought this” affiliations and increasing my positive reviews. I love it when a day after someone new likes one of my novels, an entire cluster of them sell. These are intangibles you can’t buy. The biggest success from the first wave of my AMS experiments was “K2 Virus,” which has a handful of keywords that pop 25 times a day, but when they do, I have a 4 percent conversion to sales. At the other extreme, I have words that get 200 views a day with only a half-percent click rate. You have to work these ads like a farm: planting, weeding, harvesting, and starting over the next season when conditions change yet again. This isn’t a one-time deal; rather, AMS ads are a tool in your arsenal to help you understand who your buyers are and what they want—marketing. This also informs what the topic of my next novel should be. If I can tell from research that it won’t have a market, I shouldn’t spend three months and a thousand dollars creating it.

10. Adapt
After over a week, I figured out that two of the keywords I used were getting sales but not earning as much as they cost. Therefore, I adjusted the amount I will bid so it stays inside the profit margin. This can be adapted continuously from week to week.

Targeting Cost Per Click (CPC) Spend ACoS profit point
assassin $ 1.40 $ 5.58 186.6% 0.75
medical thrillers $ 0.98 $ 10.82 180.9% 0.54