Estimate Amazon Sales From BSR: 113 Days of Our Own Rank

✨ Written with AI assistance from our own order and ranking data, and reviewed by Avanta Global EOOD before publication. Learn more

Short answer: Best Sellers Rank does track sales, but loosely and late. On our own listing, rank followed the orders of the previous two to three days, not today's. Days with the same sales level landed at ranks 2.4 to 4 times apart, and the mapping drifted by up to 1.9× from one month to another. A rank-to-units formula fitted on June–July underestimated September sales by about a third. BSR tells you roughly how much something sells. It says nothing about whether it makes money.

This is one listing in one sub-category on Amazon.com.au. The method carries over to any listing. Our numbers do not.

What BSR is

Best Sellers Rank is Amazon's ordering of products by sales within a category. Amazon says it is calculated "using sales volume data", and "recent sales count more than older sales" (Amazon's BSR guide). A product has one rank per category it sits in. Rank 1 is the best seller, and the number is relative: your rank can move because other products sold more or less, even when your own sales did not change.

Amazon publishes no formula and no rank-to-units conversion. Every "BSR to sales" figure is someone's model.

What we measured

We sell one product on Amazon Australia: a bubble mailer bundle, fulfilled by Amazon. Since 10 June 2026 our system has recorded its rank every day from Amazon's catalogue API: the top-level rank in Stationery & Office Products, and the sub-category rank in Envelopes & Mailing Supplies. Until 24 July that was one reading a day. From 26 July it is four readings a day, six hours apart, and the day's value is their median.

We matched it against our own orders by purchase time (Sydney), excluding cancellations.

Days of rank history	113 (10 Jun – 30 Sep 2026)
Orders · units, 10 Jun – 29 Sep	413 · 577
Sub-category rank, best – worst	#24 – #895
Top-level rank, best – worst	#190 – #7,166
Top-level rank ÷ sub-category rank	about 8 (range 5.6–11.5)

The two ranks moved together almost perfectly, so below we use the sub-category rank.

Rank follows yesterday's orders, not today's

We compared each rank reading with the units sold in separate windows before it (Spearman correlation; −1 would mean more sales always meant a better rank):

Units sold in this window before the reading	Correlation with rank
0–12 hours	−0.17
12–24 hours	−0.31
24–36 hours	−0.51
36–48 hours	−0.44
48–72 hours	−0.43
4–7 days	−0.35

Sales from the last twelve hours barely showed. The strongest link was to sales a day to a day and a half earlier, and sales from a week back still counted. Amazon's KDP help says something similar for books: rankings are "updated at least once per day, but changes may take up to 2 days to appear" (KDP Sales Ranking help).

In practice, the orders of the previous three days explained rank best. That number of orders matched rank more closely than the number of units did (−0.88 against −0.84). On 31 August one buyer ordered 25 units in a single order, and the day finished at 35 units, our biggest day. Rank moved from #49 to #41 over the next day, then fell back to #78. A bulk order counts for a lot of units and not much rank.

What rank corresponded to 1, 2, 5 or 10 sales a day

We grouped each day by average units per day over the previous three days, and read that day's rank:

Units per day (previous 3 days)	Days	Median sub-category rank	Middle 80% of days
about 0 (under 0.5)	3	#592	#319 – #895
about 1	11	#341	#147 – #600
2–3	27	#153	#88 – #268
about 5	44	#105	#61 – #148
7 or more	27	#62	#32 – #77

The first row has only three days, so its range is the lowest and highest rank.

A curve fitted through all of it puts 1 unit a day at about #343, 2 at #197, 5 at #95 and 10 at #55. The scatter around that curve is large: the typical day sat a factor of 1.45 above or below it. Read the other way, estimating units from rank missed by 22% on a typical day and by more than 60% on one day in ten, and that is on the very days the curve was fitted to.

The same sales, a different rank each month

The median rank on days averaging about 5 units a day, month by month:

Month (2026)	Days	Median rank at about 5 units/day
June	4	#132
July	11	#109
August	17	#70
September	12	#128

The same sales bought a much better rank in August than in September. We do not know why. Rank is relative, so competing listings probably sold differently, but we cannot see their sales. That is the point.

We tested it: a formula fitted on June and July predicted 4.1 units a day for September; we sold 6.3. A BSR calculator built on one period is already out of date in the next.

Why BSR-to-sales calculators disagree

Every calculator fits a curve like ours to some listings, categories, marketplace and period. Each choice moves the answer:

• Category size. Rank #100 in a small sub-category and rank #100 in a large top-level category are different sales levels. Ours differ by a factor of about eight on the same day.
• Marketplace. Each Amazon store ranks its own sales, and Amazon notes that the same product can rank differently in each (Amazon's BSR guide). A curve from Amazon.com says nothing about Amazon.com.au, or the reverse.
• Time. As above, the curve moved by up to 1.9× between months within one sub-category.
• Orders, not units. If rank follows orders, a product that sells in multiples (ours averaged 1.4 units per order) sells more units than rank suggests.

We have not tested any vendor's calculator, so we quote none.

Can BSR find profitable products?

It finds products that sell. It cannot say whether they make money: fees, product cost and ad spend are not in it. Our three price periods (full numbers):

Period (2026)	Median sub-category rank	Profit after all ad spend, per day
A: 12–29 Aug	about #62	A$21.6
B: 30 Aug – 10 Sep (deal)	about #71	A$59.8
C: 11–28 Sep	#136	A$32.3

Our best rank came in our least profitable period. In period C our rank was more than twice as bad as in A, and we made half as much again per day. In mid-July our referral fee more than doubled (the fee change). Rank has no way to show that, because fees are not part of it.

What we got wrong

For the first 45 days we recorded one rank reading a day, and our daily report judged the listing's direction from one reading to the next. In that period, one day's reading differed from the day before by a factor of 1.5 or more on 55% of days. After we switched to the median of four readings, that fell to 23%. And the rank moved the opposite way to recent orders on 14 of 41 days, against 12 of 55 later. Part of that change comes from higher sales since August, and we cannot separate the two. But one reading a day was not a trend, and we read it as one.

How to check your own

1. Record your rank every day, several times a day if you can, and keep the median. Note both the top-level and the sub-category rank.
2. Export your orders from Seller Central and count orders (not only units) per day, in the marketplace's own time zone. Leave out cancellations.
3. Line up rank against the orders of the previous 1–3 days, not the same day. Check which lag fits your listing best.
4. Group days by sales level and look at the spread of rank, not only the median. That spread is your error bar when you read a competitor's rank.
5. Redo it every month. If the mapping moves, any calculator will move with it.
6. Check profit separately, from your settlement report (how to read one). The settlement profit reader runs it in your browser.

FAQ

How accurate is estimating Amazon sales from BSR? On our listing, a formula fitted to our own history missed by 22% on a typical day and by over 60% on one day in ten. Fitted on June–July, it underestimated September by about a third.

How long does BSR take to react to a sale? On our data the strongest link was with sales 24–36 hours earlier. Sales from the last 12 hours barely showed.

Is a better BSR always more profitable? No. Our best-ranked period made the least profit per day of the three we measured.

About this article

Rank figures come from Amazon's catalogue API for one Avanta Global EOOD listing on Amazon.com.au, recorded daily from 10 June to 30 September 2026 (one reading a day until 24 July, four from 26 July). Orders come from our own orders report, 10 June – 29 September 2026 (413 orders, 577 units), by purchase time in Sydney time. Profit per period is from our settlement and advertising reports, as published in our ACoS article. Amazon's statements are quoted from its BSR guide and KDP help page, read on 30 September 2026. Measured on 30 September 2026. Currency is Australian dollars.