What Actually Wins the Buy Box? A Deep Dive Into the Data (for Booksellers)

I've been selling books on Amazon for over 15 years. In that time, I've heard every theory about how to win the Used Buy Box - drop your price by a penny, get more feedback, switch to FBA, maintain a 99% rating. The forums are full of confident explanations, many of them contradicting each other.

Amazon tinkered with their Buy Box algorithm a few times in 2025, then locked in their new rules right around November 1st. It feels like the changes are here to stay, so we might as well dig in and try our best to reverse engineer the 'ol algorithm.

Let's turn to the data!

(special thanks to Claude Code, Keepa, and caffeine for powering this research)

If you're not familiar with the Buy Box, it's the Featured Offer on Amazon where you can buy anything with the least amount of friction. One click adds it to your cart - either a new copy through the New Buy Box, or a used copy through the box underneath. It looks like this:

It's estimated that around 75% of used books are purchased through this magical box. So, it's pretty important to "win" the Buy Box. This post dives into how it used to work and how it works in this new year.

The Sample Data

Using the Keepa API, I pulled 100-200 random ASINs from known booksellers with various inventory sizes, trying to represent a broad sample. I started with 2,500 ASINs and stripped out the non-book data, leaving us with 2,343 book ASINs - capturing 17,872 individual seller offers. For each listing, I recorded who was winning the Buy Box, who was competing for it, and every metric I could measure: condition, price, fulfillment method, feedback counts, seller ratings, rotation patterns.

Then I started isolating variables. What actually moves the needle? What's just noise?

Some of what I found confirmed the conventional wisdom. Some of it didn't. And some of it surprised me enough that I went back and re-ran the numbers to make sure I hadn't made a mistake.

Key Finding #1: The algorithm no longer prefers FBA offers above the rest of the crowd

I pulled historical Buy Box data from January 2025 to compare against January 2026. Same ASINs, same methodology, one year apart.

The differences aren't subtle at all:

FBA went from dominating the Buy Box to... not. In January 2025, FBA sellers won 85% of Buy Boxes in my sample. One year later? Just 13%.

The way the game used to be played was to be the cheapest used FBA offer and you'd nearly always win the Buy Box. The algorithm seemed to favor "vetted" sellers, so it was tougher to win the Buy Box until you had racked up roughly 10 feedback ratings. This wasn't a hard and fast rule, but it seemed to raise the bar for just launched sellers until they proved themselves.

Something has fundamentally and dramatically changed.

Bottom line: If you switched to FBA specifically to win the Buy Box in books, it might be worth revisiting that assumption.

Key Finding #2: Price is only half of the equation

Roughly half the time in 2026, the Buy Box is occupied now by the cheapest offer on the listing - regardless of fulfillment method. If price isn't the only metric that can earn the Buy Box, what else is in play for the other half of the Buy Box wins? I'm glad you asked - more on that below.

Key Finding #3: Condition is king

Check out these differences between which offers won the Buy Box in 2025 and 2026:

In January 2025, the algorithm didn't seem to care much about condition. Good books won more than two-thirds of the Buy Boxes. Acceptable won nearly 17% - a real share. Very Good was actually underperforming at 12.6%.

Fast forward to the present: Very Good now accounts for nearly 6 out of 10 wins. Acceptable collapsed to less than 1%. And Collectible? Zero wins across the entire dataset. eBay is your best bet for Collectible books, and listing them as Collectible on Amazon is useful to get around the "Priced too High" alerts, but it won't win you the Buy Box.

The algorithm appears to be sorting offers into condition tiers first, then evaluating within those tiers. If that's true, an Acceptable offer isn't really competing against a Very Good offer or even a Good offer - they're in different lanes.

Finding #4: The pricing sweet spot

This brings us to the question every seller asks: how much can I charge and still win the coveted Buy Box?

If condition creates separate lanes, and the algorithm favors higher conditions, then a Very Good book should be able to charge something over a Good book and still win. But how much?

If two books have the same price, the algorithm favors the book that's in a better condition. Nearly every time.

Within a 10% price premium, the higher condition wins almost 80% of the time. That's a strong advantage.

At 11-25%, it drops to a coin flip.

Beyond 25%, the lower-priced offer starts winning consistently.

This suggests there's a tolerance band - a range where condition carries you, and beyond which price takes over. The exact boundaries probably vary by category, competition level, and factors I haven't isolated yet. But the pattern is clear enough to be useful.

What I'm Still Figuring Out

This research raises as many questions as it answers. A few things I'm still chewing on:

Does this apply outside of books? My sample is entirely books. Media categories might behave differently than used books. At first glance, CDs and DVDs appear to be using the old Buy Box logic that still heavily favors the cheapest FBA offer. More research to come in future posts.

What triggers suppression? About 13% of listings in my sample had suppressed Buy Boxes - meaning Amazon decided no offer deserved that Add to Cart button. I have some hypotheses about the reasons for this (high price relative to list price, low FBA presence), but I haven't nailed it down yet.

How does rotation actually work? The data suggests the Buy Box re-evaluates every 4-7 hours among similarly-scored offers, with a 15-40 minute cooldown after repricing events. But I'm still mapping the mechanics.

Is feedback really irrelevant? The data shows sellers with zero feedback in the past year winning at meaningful rates. That surprised me. I want to dig deeper before making strong claims, but it appears that the playing field is leveled between established sellers and new sellers, and between FBA sellers and MF sellers.

Some Thoughts (Not Rules)

I'm hesitant to turn preliminary research into commandments. The algorithm could change tomorrow. My sample might have blind spots. And your business is likely different from mine.

But here's what I'm thinking about for my own sourcing and pricing:

• Condition certainly deserves more attention in 2026. If the algorithm really does favor Very Good over Good by a meaningful margin, then the extra few seconds spent evaluating condition at the sourcing stage could pay off at the Buy Box stage. I'm not sure yet how much that's worth in practice, but it's on my radar.
• Be wary about listing Acceptable books. They're almost entirely eliminated from winning the Buy Box. You'll still sell them, but not through the coveted Buy Box. Plan accordingly.
• The race to the bottom might be optional. If 54% of winners aren't the cheapest, then undercutting by a penny might be solving the wrong problem. Especially if you're already in the top condition tier on a listing. It may be worth looking through your offers for opportunities to raise your prices and still be the cheapest within your condition. Roughly a third of the time, the Buy Box was afforded to a seller who had the cheapest price AND a better condition. This is unnecessary as they almost certainly could raise their prices and still win the Buy Box. Pay special attention to your repricing logic to avoid leaving money on the table.
• FBA economics in books might need a second look. If FBA isn't providing the Buy Box lift it used to, the calculus changes. That doesn't mean FBA is bad - there are other reasons to use it - but "FBA wins the Buy Box" may not be one of them anymore. At least not in books. There still is a "Prime Bump" in play, but it's fairly small. More of that in future posts.

Where This Goes Next

This is the first in a series. I want to go deeper on each factor - condition, price, fulfillment, feedback, suppression, rotation mechanics. I want to test these patterns over time and see if they hold. And I want to answer questions I haven't thought to ask yet.

If you're the kind of seller who wants to understand why things work, not just follow rules that might be outdated, I think you'll find this useful.

Stay In the Loop

I'll be publishing more research as I dig deeper. If you want to get future posts first - before they hit the blog - join the email list. No spam, just data.

Work Together

I'm offering coaching and consulting for sellers who want to go deeper on strategy, sourcing, or pricing. I'm also taking on custom research projects - if you've got a specific question about the market, your competition, or your own data, I'd love to help you find the answer.

Email me to start a conversation and learn how to go upstream in your own business: caleb@upstreampursuits.com