Amazon Conversion Rate by Category: What the Numbers Actually Mean

How Amazon calculates conversion rate — and where it gets muddled

Amazon's own term is Unit Session Percentage: units ordered divided by sessions, shown in Business Reports under Detail Page Sales and Traffic. That's not orders divided by sessions — a single order with three units counts as three units against one session, which quietly inflates the number for anyone selling multipacks or add-on items.

Worked example: a countertop appliance listing gets 12,400 sessions in a month and sells 1,116 units. Unit Session Percentage = 1,116 / 12,400 = 9.0%. If those 1,116 units came from only 980 orders, the order-based conversion rate reads 7.9% on the exact same traffic — a full point lower. Which definition a benchmark article used is rarely stated, and that's the first reason category tables disagree with each other.

Sessions themselves get counted inconsistently too. A shopper who opens your listing on the Amazon app, closes it, and reopens it on desktop can register as two sessions for one person. None of the category benchmark tables circulating right now disclose whether their data adjusts for this.

Category benchmarks — and why they don't agree with each other

The ranges cited most often across seller sites, taken category by category:

Electronics and Home & Kitchen sit at the top of that spread because buyers arrive already comparing specs rather than browsing. Clothing, Shoes & Jewelry sits at the bottom because size and fit uncertainty pushes shoppers to open several listings before buying one. That part of the story is consistent across sources.

What's not consistent is where the whole scale sits. One widely read benchmark page puts the site-wide average at 8%–12%; another puts it at 2%–5%. Both can be technically correct — they're measuring different traffic mixes. A listing getting 90% branded search traffic converts far higher than one getting 40% of its sessions from prospecting placements, and neither article tells you which mix its benchmark assumes.

If your number is below the range — what to check before you touch the listing

A conversion rate below its category range isn't automatically a listing problem. Work through these before you rewrite a single bullet point:

• Stockouts and Buy Box suppression: Amazon still counts the session even when the shopper couldn't actually buy.
• A recent price move relative to your closest competitor, not your own history.
• A drop in review count or star rating after a return spike or a bad review run.
• A change in traffic mix — a new campaign type added sessions that weren't there in the prior period.
• Seasonality specific to the category, which moves the whole range up or down regardless of listing quality.

If none of those moved, the listing is probably fine and the traffic mix changed instead — which is the mistake covered next.

The mistake we've made ourselves: blaming DSP for a conversion dip that wasn't a listing problem

We've had this conversation on our side of the table more than once: conversion rate at the listing level drops the month a DSP prospecting campaign starts, someone assumes the targeting is off, and the reflex is to pause it. That reflex is usually wrong, and it's worth naming plainly because we've made it too, early in a program, before checking the audience mix.

Prospecting and online-video placements are built to reach people who have never seen the brand. Those sessions convert at a lower rate on the first visit — that's not a flaw, it's the definition of new demand. Across 30 of our advertisers in July 2026, the book as a whole ran at 6.04x return on ad spend measured across the whole account, not the best-performing line item, with 20.1% of the 57,137 attributed purchases coming from shoppers new to the brand. A chunk of that new-to-brand volume would score as "low-converting" if you judged the DSP placements by the same yardstick as a Sponsored Products campaign on branded search terms. It isn't the same job.

It's also worth separating cost from conversion, because they get quoted against each other by mistake. The $0.41 CPC that circulates in Amazon DSP commentary is an online-video number; the blended figure across that same 30-advertiser book, mixing display and video, was $1.42 CPC at a $4.00 CPM. If you're comparing a listing's conversion rate to what a campaign "should" cost per click, make sure both numbers describe the same placement type.

How to benchmark yourself instead of a category average

A published range tells you roughly where you should sit. It can't tell you why you moved. For that, use your own data:

• Pull your own trailing 90-day Unit Session Percentage before you look at any published range — it's the only number measured the same way every time, on your actual traffic.
• Segment sessions by traffic source where you can. Amazon Marketing Cloud shows DSP-driven sessions separately from organic and Sponsored Products sessions — the only way to tell whether a blended dip is a listing problem or a traffic-mix change.
• Compare the same weeks year over year, not this month to last month. Seasonality moves category conversion rates more than most single optimizations do.
• Treat the direction of your own trend as the signal, not the absolute number against a category table you didn't produce.

Where reMKTR fits

reMKTR runs Amazon DSP as a managed service and reconciles DSP against sponsored ads inside Amazon Marketing Cloud, specifically so a prospecting-driven dip in listing-level conversion rate doesn't get misread as wasted spend. We measure incrementality with holdouts and matched controls, because last-click attribution — and a session-level conversion rate — can't tell you whether a shopper would have bought anyway. That's true whether or not you ever run a campaign with us; it's just how to read the number correctly.