August 6, 2026
Amazon Orders vs. dataryot Orders
Having trouble matching dataryot's numbers to Seller Central? One report says one thing, dataryot says another — and nothing is broken. Amazon counts the same shopping trip three different ways.
One checkout, three numbers
Say a customer checks out once with two tubs of protein powder and one shaker bottle — that's the picture above. Amazon records it as 1 order, 2 order items, and 3 units:
Orders
One checkout by one customer. However many products or quantities are in the cart, it's still one order.
Order Items
One line per unique product in an order. Two different products in the same checkout = two order items.
Units Ordered
Every individual unit, counting quantity. Two tubs of the same protein = two units.
The rule of thumb
Units Ordered ≥ Order Items ≥ Orders
They only match when every order is a single unit of a single product. For most brands, orders run 5–10% below order items — the gap is just customers buying more than one thing at a time.
Which number is your report showing?
Here's the part that catches almost everyone: Amazon doesn't really report orders. The Sales & Traffic Business Report — the one most sellers screenshot — shows Total Order Items and Units Ordered. That's by design: Amazon's analytics are built around ASINs — sessions, page views, and conversion are all tracked per product — so the item grain is what lines up with the rest of its reporting. Read "Total Order Items" as "orders" and every other order count will look wrong by a few percent, in the same direction, every month.
Amazon keeps a glossary of Business Reports terms (Seller Central login required) that defines each one — worth bookmarking for the next time two reports disagree.
So what does dataryot count?
The opposite of Amazon's item-first view. dataryot's repeat analytics — Repeat Order Rate, Repeat Customer Rate, LTV — are built on actual orders, because purchasing behavior lives at the order level: it's about how often a customer comes back, not how many products were in the cart. Three things follow:
- Orders, not items. A checkout with three products counts once — not three times.
- Shipped orders tied to a customer. An order can only be labeled "new" or "repeat" once it ships and is attached to a buyer, so cancelled and not-yet-shipped orders don't appear.
- Same-day orders are bundled. Two orders from the same customer within 24 hours count as one purchase occasion — otherwise one shopping trip split across two checkouts would look like a repeat customer.
- FBA orders only (for now). dataryot's order metrics currently cover FBA orders — FBM orders aren't included yet.
- Days run on UTC. Amazon's US Business Reports bucket each day in Pacific Time; dataryot uses UTC day boundaries, so a handful of orders at each day's and month's edge can sit in different buckets between the two systems. That's normal, not a discrepancy.
So dataryot's repeat-analytics order total always sits a bit below Seller Central's "Total Order Items" — by design, and by a stable, explainable amount. The repeat rates themselves are unaffected: numerator and denominator use the same basis.
Quick reference
| You want… | Look at… |
|---|---|
| A number that ties to Seller Central | Sales & Traffic (order items / units, Pacific Time) |
| How many checkouts happened | Orders — distinct order IDs |
| New vs. repeat customer behavior | dataryot repeat analytics (order-based) |
One real month, three ways
Here's a single brand's June — Amazon's Sales & Traffic report on the left and middle, dataryot on the right. Same month, three definitions, three numbers that are all correct:
Sales & Traffic — Units Ordered
6,276
Every individual unit, quantities counted. The biggest of the three numbers.
Sales & Traffic — Total Order Items
5,492
One line per product per order — the number most Amazon reports show.
dataryot — Orders
4,918
5,492 order items collapse into 5,101 actual orders; same-day repeats bundle down to 4,918 purchase occasions.
Read it backwards and the rule of thumb plays out exactly: 6,276 units sat on 5,492 order lines, placed through 5,101 checkouts, by what dataryot counts as 4,918 distinct purchase occasions. Units Ordered ≥ Order Items ≥ Orders — every time.
Parker Watts
Founder @ dataryot
See your real order data.
Repeat rates, lifetime value, and profitability — on definitions that hold up.