Amazon has a ranking system for the products they offer. Rankings for books currently go from 1 to over 7 million, with 1, of course, being the best. We’re #1! When deciding whether or not to inventory a used book, CD, or DVD, we consider several factors, perhaps the most important being the Amazon Sales Rank. I have compiled our sales data for the last three years (2006-2008) to create the charts below. To help you gauge the scale of the dataset, I’ll just tell you we sold over 50,000 individual inventory items, almost entirely used.
You can see we sold 90% of items with an Amazon Sales Rank ranging from 1 to 500,000 at the time we listed the item. The percentage of items sold drops off steadily as as the Amazon Sales Rank increases (worsens). We sold only 5% of the items we listed having a rank from 6.5 million to 7 million.
The most popular books tend to have low prices in the used market. There is generally high demand and high supply. When lots of people buy the book new, there will inevitably be lots of used copies available, and used prices will drop rapidly as sellers jockey to make the sale.
The next chart is similar to the chart above, but additionally shows what percentage of total sales came from each Amazon Sales Rank grouping. Items with an Amazon Sales Rank of 500,000 or better accounted for 62% of items sold, and 53% of sales in dollars. Notice that for items ranked from 500,001 to 1 million, percentages for number sold and sales in dollars equal out at 17%, and in groupings beyond that the dollars out pace number sold.
We tend to allow relatively inexpensive items in to our inventory as long as the item has a good sales rank. We also allow items in to our inventory without such a good sales rank as long as the price is high enough to justify the effort it takes to list the item, and the shelf space to store it for possibly an extended period of time.
We have very intentionally chosen to concentrate our efforts on the 2 or so million highest ranked books. We routinely pass-up books that, based on data we have collected, have very little chance of selling. We prefer to channel items that are less likely to sell back to our clients or the Ann Arbor Reuse Center than to have them languish on our shelves unread. Which raises the question, how long does it take an item to sell?
We sell books, CDs, and DVDs on consignment, which requires tracking each item very carefully. We know exactly when an item is listed, and when it sells. Therefore, we can easily report on how many days it takes, and we can group the items by Amazon Sales Rank.
Like all of the charts in this post, the above is based on our used book sales from 2006-2008.
The surprising thing about the “days to sell” chart is that items in the middle ranks took the longest to sell. I am not yet sure what to make of this. It might help to break the data down by year (I took a quick look at just 2006-2007, and the pattern is the same). Don’t let that distract you though. Items with an Amazon Sales Rank better than 500,000 on average sell in about 30 days. That’s a pretty useful piece of information.
Results, of course, may vary. You also have to consider condition, customer feedback rating, and selling venues (for us that is Amazon
, Alibris, Abebooks, Books by Chance).
The data used in this post include sales of books, CDs, DVDS, audio books on cassette, and VHS. Amazon has separate rankings for books, music, and movies and the range of rank values varies for each from 1 to N. I lumped everything together. We sell primarily books, and the vast majority of our sales occur on Amazon.
The following table shows numeric data for the first two charts.
| Amazon Sales Rank |
% sold |
% of Total # Sold |
% of Total $ Sold |
| 1-500000 |
90 |
62.60 |
53.08 |
| 500001-1000000 |
82 |
16.63 |
16.65 |
| 1000001-1500000 |
75 |
7.72 |
9.29 |
| 1500001-2000000 |
65 |
4.34 |
5.89 |
| 2000001-2500000 |
54 |
2.33 |
3.21 |
| 2500001-3000000 |
47 |
1.29 |
1.97 |
| 3000001-3500000 |
43 |
0.89 |
1.56 |
| 3500001-4000000 |
36 |
0.46 |
0.85 |
| 4000001-4500000 |
34 |
0.27 |
0.65 |
| 4500001-5000000 |
31 |
0.14 |
0.41 |
| 5000001-5500000 |
20 |
0.06 |
0.16 |
| 5500001-6000000 |
18 |
0.03 |
0.06 |
| 6000001-6500000 |
16 |
0.01 |
0.06 |
| 6500001-7000000 |
5 |
0.00 |
0.00 |
One final point. I omitted items that were unranked at the time listed. We sold 34% of items that were unranked at the time of listing, and they sold on average in 242 days. That’s a pretty good percentage. Unranked items are wildcards, generally worth the risk if the price is decently high. The thing to watch out for, however, is other editions of the same title that have low prices and good sales ranks.
Making the Charts
The following MySQL command shows how I generated the data for the last chart “Average Number of Days for an Item to Sell Grouped by Amazon Sales Rank”. Of course, this is what worked for my inventory database, and the way I have it set up. There are just two tables at play. One that holds the inventory, and another with sales information. The best trick here is the one used to calculate the age. You’ll see the number 86,400. That’s the number of seconds in a day. Inventory items without an Amazon Sales Rank, or with a rank over 7 million were omitted. Also of interest may be the calculations used to cluster sales by rank in increments of 500,000.
select concat(round(floor(skuinventory.salesrank_orig/500000)*1000000/2)+1, '-', round(floor(skuinventory.salesrank_orig/500000)*1000000/2+500000)) as `Amazon Sales Rank`, floor(avg(if(sale_status='sold', @age:=(unix_timestamp(sold.payments_date) - unix_timestamp(skuinventory.t_create))/86400, NULL))) as `# Days` from skuinventory left join sold on skuinventory.sku=sold.sku where year(t_create)>=2006 and year(t_create)<=2008 and salesrank_orig<7000000 and salesrank_orig is not NULL group by floor(skuinventory.salesrank_orig/500000);
I like TN Ernie’s number