Home » Amazon, Ebooks, Self-Publishing » An Experiment in Amazon Sales Rank

An Experiment in Amazon Sales Rank

2 June 2016

From author Dan Koboldt:

Few things are more obsessed-over by modern authors than Amazon Best Sellers Rank. For most of us, it’s a form of numerical torture, updated hourly. Although the precise inner workings of the algorithm are known only to Amazon, there are a few general accepted principles:

  • The more a book sells, the higher (closer to 1) its rank goes.
  • Rankings change rapidly, often moving thousands of positions within hours.
  • Paid and free books are ranked separately.

Beyond these basic principles, there is little certainty. The basic formula for Amazon Bestseller Rank seems to be sales over time, meaning that a book must continue to sell to maintain a high position on the lists. In theory, it might therefore be possible to achieve a high ranking if a book sold a modest number of copies in a short amount of time. I decided to test that theory with an experiment.

. . . .

To assess the impact of rapid sales in a short time frame, I organized a coordinated buy of The Rogue Retrieval on May 12, 2016. Importantly, I didn’t make any purchases myself, for two reasons:

  1. I’d already bought the e-book for both of my Kindles, so Amazon wouldn’t let me.
  2. I didn’t want to appear to be gaming the system, or “buying the bestseller list” as is common practice in the business nonfiction world.

Instead, I turned to friends and fans, most of them fellow authors, who were willing to help. More than 50 people stepped up, and agreed to purchase the Kindle e-book at a specified time on that day. We targeted 8:00 a.m., which meant that some who live overseas bought at around midnight, NYC time (morning for them).

Here are some questions that I hoped to answer:

  • How long does it take for Sales Rank to reflect new sales?
  • What rank could be reached with 50+ sales in a single day?
  • What’s the effect on subcategory rankings?

. . . .

The first twenty-four hours into my experiment offered some insights:


The first boost (2 hours) came from sales the day before, when some eager volunteers pulled the trigger a bit early. The two main peaks occur at hour 7 (European volunteers) and hour 16 (US volunteers). Thus, the lag isn’t quite 12 hours as I previously believed, but closer to 8 hours.

. . . .

In total, about 57 people bought The Rogue Retrieval within a 24-hour period. Although we targeted 8 a.m., sales continued throughout the day. The rank peaked about noon the following day at 4,748 in the Paid Kindle Store:


It might surprise some that so many sales only propelled my book to #5,000, but this speaks to the size of the Amazon bookstore. If we assume it ranks 1 million books (an underestimate), this ranking is in the top 0.5%.

Notice another property of Amazon Sales Rank: increases happen quickly (as “jumps”), whereas the post-sale decrease in rank happens more slowly. Especially for the first couple of days. After 72 hours, however, the drop-off becomes more precipitous. This seems to support something I’ve read about Amazon’s internal ranking algorithm: sales from previous days carry some weight. The window seems to be 48-72 hours.

Link to the rest at Dan Koboldt and thanks to Ann for the tip.

Here’s a link to Dan Koboldt’s books. If you like an author’s post, you can show your appreciation by checking out their books.

Amazon, Ebooks, Self-Publishing

10 Comments to “An Experiment in Amazon Sales Rank”

  1. Good stuff. Now the book is back down to #192,000 or so. It’s got 36 reviews so I’d figure a book that liked would do better. Alas, it’s not. It didn’t have the staying power, doesn’t have the sustainability of earnings that we’d hope for.

    So how can you change that? The book is certainly long, at 421 pages, and it’s new, having come out in January.

    It’s not part of a series, however.

    Since it does have the reviews, and since I’m assuming most are legitimate (not family and friends) then I’m also assuming this book has potential.

    I’d start on a Book 2 and then right away a Book 3. Maybe box-set them after that. I certainly would not make the first book free or even $0.99, however. I would raise the price to $3.99, though.

    You’ll have to pull it out of its current categories as well. Currently sci-fi is very tough to rank in, at least what you’re doing with Time Travel (#1 rank of #4 and #100 rank of #33,000, so you’re not gettin’ in that one).

    Might be better trying Alternate History, Metaphysical & Visionary, Cyberpunk, Steampunk, maybe even Classic. Surely you can rank better in one of those (in the #70,000 to #95,000 range).

    Most of the time we just have things in the wrong genre so we’re not getting the best audience.

    Perhaps this book has potential.

    Perhaps yours does as well.

    • Since it does have the reviews, and since I’m assuming most are legitimate (not family and friends) then I’m also assuming this book has potential.

      The method used to produce this sales ranking was to recruit a whole bunch of friends, including other authors, to buy the book on a particular day and write reviews.

      This is not the proper audience of readers. The attempt was to see if visibility could increase sales.

      That job is up to the book’s content. If it is a good book to SF readers, it will be popular. If not, it will have a hard time, regardless of its initial or one-day sales rank artificially produced.

      Readers are very smart. They know what they want, and are unhappy if fooled into getting something not up to snuff.

    • N.E. Montgomery

      Dan’s book is trad-pubbed, so he’s going to have a harder time leveraging some of the strategies open to the indies, like box-sets and price changes, etc.

      But it was an interesting experiment well-planned (he’s a geneticist, so I’m not surprised by that).

  2. 1 day (or so) isn’t enough- 57 sales aren’t enough really to do anything.

  3. These results are consistent with my experiences: an eight hour lag from sales to rankings change (very noticeable with a Bookbub ad), and 50 sales roughly equals a 5,000 rank. (Typcially, 10 sales is about a 10,000 rank, 100 is a 1,000 rank, 1,000 is a 100 rank…) Thanks for sharing!

  4. Unknown excuse for a writer, 8 sales no comments over the first 3 days Rank: #51,831.

    So they might have gotten more of a pick-me-up for steady sales over several days instead of a sudden ‘dump’.

    (Any bets Amazon does the delay to help prevent ‘gaming’ their system?)

  5. I watch ranking pretty closely during the first 30 days of a new release.

    Based on what I’ve recorded (yep, I have a spreadsheet), it takes roughly 18-24 hours from the first sale before the first ranking appears.

    Depending on how many copies sell per day, the ranking can update anywhere from hourly to every 8 hours. Often, a hourly change is only 1-6 spots.

    My last series release’s best day was on Nov. 28, 2015: over 250 sales, 185 which were on Amazon.com. Its final rank (via Author Central) for that day was #759.

    On Nov. 29th, it sold 151 copies on Amazon.com, and the final rank for the day (Author Central again) was #709. That was its highest ranking, period.

    #4k-#5k rankings, that particular title sold 1-52 copies daily (just looking, I’d say maybe an average of 35-38 copies per day, over a 5 day period, Days 38-42).

    I have noted that over the past 2 years, it takes more daily sales to reach the same rankings previous releases hit (most noticeably on Amazon.com and Amazon.co.uk).

    And since I’m terrible about constantly checking sales reports throughout the day, I’ve always seen that sales do not instantly affect ranking. They’ve always taken 4-8 hours before a comparable ranking change appears.

    FYI, I don’t ask people to buy at certain times on certain days. I just announce the new release on social media (and my mailing list). Day 1 of the release I used as my example, it sold 51 copies on Amazon.com, and its record AC rank for the day was #5285.

    Amazon and D2D don’t send out new release emails the day of release. Those haven’t gone out for 3-7 days after publishing.

    • “Recorded” not record. Sorry.

    • Reality Observer

      “I have noted that over the past 2 years, it takes more daily sales to reach the same rankings previous releases hit (most noticeably on Amazon.com and Amazon.co.uk).”

      Now that says volumes about the claim of “decreasing e-book sales.”

      I think the market is also now firmly into the “quality differentiation” phase, where consumers are becoming more discriminating.

      No, not saying there that your books aren’t quality, Scath – just that there are more quality books out there than before, so your potential buyers have more choices. The wild and woolly days are over, where someone writes anything at all, slaps a horrible cover on, and sells gangbusters to a public desperate for anything to put on their reader.

      • There absolutely are more quality books out than before. I browse Amazon a lot, and am seeing more of them than stinkers now. The stinkers still exist, they’re just being buried.

        And no worries. I’m well aware some look down their noses at my UF works, because they’re not literary masterpieces. (Not saying you do!)

        They can look down their noses all they want, because that having to sell more copies to hit the same rankings thing also means each new release is doing better (roughly 25% better) than the last one. I’m amazingly good with that. 🙂

        So is my bank account.

Sorry, the comment form is closed at this time.