Ourus · one decision, traced
This is one order.
It left a warehouse on Monday and arrived at this address on Tuesday. Nothing unusual about it.
Thursday
It came back.
A collection, a handler, an inspection, a restock. The sale reverses and the margin on it is gone.
$17
to process one returned parcel
2×
the cost of the original delivery
Same street, same week
So did four
others.
Returns are not an edge case. They are the normal condition of the business.
~19%
of everything bought online is sent back
$850B
US retail returns in a single year
One city · one month
Every return is
a route back.
Each arc is a parcel travelling in reverse to a processing centre.
0
returns from this city this month
$0
spent moving them backwards
What a fraud model sees
One number: how often you send things back.
Cross the line, get friction or a ban. This is how nearly every returns tool on the market works.
85%
of retailers deployed AI to catch return abuse
45%
say it actually works
Now add the number it never sees
Let the city rise
by profit.
Every building grows to the annual profit that household generates. Same city, same returns, one more dimension.
The picture inverts
The tallest tower is the one it just banned.
She buys in three sizes, keeps the expensive one, returns the rest. Highest return rate on the block. Highest profit in the city.
And the one it never saw
Real abuse is short, not tall.
A 9% return rate looks clean on any dashboard. What comes back is an empty box.
What Ourus does instead
Send agents
through the city.
Every household becomes an agent with its own tolerance. The policy is tested against all of them before one customer feels it.
The result
An $850K swing on one policy.
No new customers acquired. No new products. The same month, decided differently.
Measured across 12,400 simulated households
Fraud loss recovered—
Profitable customers lost—
Net margin, yearly—
Policy v1 · proposed on the data alone
Block anyone who returns more than 30% of their orders.
Illustrative model on published returns economics. Not client data.
Households simulated
0
Returns 40%
$4,200 profit a year
Returns 9%
Empty boxes · −$980 a year
FLAGGED FOR BAN
Ourus
Everyone is building AI
to flag the return.
The money is in what your
best customer does next.
We simulate that before the policy ships.
Optimized by math. Validated by simulation.
Selecting first design partners now.
ourus.ai
Ourus