Hard-gated confirmation: the COD operator playbook

Most COD performance failures in LATAM are not last-mile failures. They are confirmation failures, masquerading as last-mile failures.

If a merchant ships an order before a human has reached the customer, the carrier becomes the QA layer. The carrier is not designed to be the QA layer. So orders that should never have shipped are routed, dispatched, attempted, retried, and finally returned — and the entire cost of that loop, including the carrier handling fee, lands on the merchant. The order shows up in the dashboard as RTO, but the failure was upstream of the carrier.

Hard-gated confirmation moves the QA layer to where it belongs: before dispatch.

If a Fufills order is not confirmed by a human inside our published SLA, it does not ship. No exceptions, no carrier-decides, no merchant-overrides.

This is the operating principle most COD operators in LATAM never adopt, because it shrinks the funnel. It is also the only reason our RTO target is 10–15% on accounts using the full execution stack, against an industry baseline that runs 25–40% on ungated COD.

A confirmed order is one where a human in our risk-control call center has, on a recorded call:

1. Verified the buyer is the buyer. Not a friend, not a roommate. The person who placed the order.
2. Verified the address is deliverable. Not a hospital. Not a wrong-zip apartment. Address verified against carrier coverage in that lane.
3. Verified intent to pay on delivery. Tone, language, hesitation patterns. Disposition coded.
4. Verified product, qty, price, currency. No surprises at the door.
5. Locked the delivery window. "Tomorrow morning" or "Friday afternoon" — not vague.

Anything that fails any of the five steps does not move to dispatch. It either re-enters the retry queue, gets re-routed for re-confirmation, or is killed and refunded.

A first-attempt failure does not mean the order is dead. It means the order is paused. The SOP says:

• Attempt 1 — within 15 minutes of order placement, while intent is high.
• Attempt 2 — 2 hours later if Attempt 1 failed (no answer, voicemail, language mismatch).
• Attempt 3 — next business day, at a different hour of the day to catch the buyer in a different context.
• Attempt 4 (final) — 24 hours after Attempt 3. If unconfirmed, the order is closed.

Four attempts, capped. Past four, the recovery rate drops below the cost of the next attempt. That is when the operator becomes more expensive than the abandoned cart.

Every attempt closes with a coded outcome. We use roughly 12 disposition codes across the LATAM book, but the four that matter most are:

• CONFIRMED — proceeds to dispatch.
• RESCHEDULE — confirmed, but with a delivery window the buyer named (and we honor).
• CANCEL_BUYER — buyer declined. Refunded, removed from carrier flow.
• UNREACHABLE — capped attempts hit, no contact. Closed, no dispatch.

Disposition codes are how a COD operation distinguishes between failed delivery (carrier problem) and failed sale (confirmation problem). Without them, every loss looks identical in the dashboard and the operator chases the wrong fix.

The argument against hard-gated confirmation is that it shrinks the funnel — confirmed orders are fewer than gross orders, and gross orders are what the merchant celebrates on day one.

The argument is correct and irrelevant. Confirmed orders are also the only orders that produce revenue. The math:

The ungated funnel shipped 1,000 packages to net 650 deliveries. The hard-gated funnel shipped 720 packages to net 612 deliveries. Hard-gated delivered 94% as many real orders with 28% less last-mile cost and far fewer reverse-logistics events.

This is the central trade in COD operations: do you optimize for the dashboard number on day one, or for the revenue number on day thirty.

In production, hard-gated confirmation is one of five linked SOPs that make up the execution stack: Confirm → Dispatch → Deliver → Collect → Transfer. Each step has its own SOP, its own disposition codes, its own SLA window. Confirmation is the first hard gate. Collection is the second.

This is why we describe Fufills not as a 3PL but as a COD enablement platform: the value is in the gates and the reconciliation, not in the trucks.

Three questions an operator should ask their current COD provider this week:

1. What is your confirmation rate, by lane, last 30 days? (If they cannot answer, they do not measure it. If they cannot measure it, they do not enforce it.)
2. What is your attempt cap and your retry cadence? (If the answer is "we call until we reach them," the cap is too high. Past four attempts, the marginal cost exceeds the marginal value of a confirmed order.)
3. Show me your last 100 UNREACHABLE dispositions — what was the median time-to-cap? (A healthy book closes most unreachable orders inside 48 hours. A bloated book chases for a week.)

If those three questions go unanswered, the operation is running ungated. Which means it is running at industry-average RTO. Which means the merchant is paying for confirmation as a feature without receiving it as a result.

That is the gap Fufills was built to close.