Fraud Screening and Order Holds: What Happens When a DTC Order Gets Flagged
Every DTC brand eventually holds an order it is not sure about. Here is how fraud screening, manual review and release rules work between your store and your 3PL, and what a held order costs while it waits.
Somewhere between checkout and the packing bench, a small number of direct-to-consumer orders stop moving. A risk score comes back high, a billing address does not match, five orders arrive from the same card in four minutes. The order goes on hold.
What happens next is rarely designed. It gets improvised the first time it happens, and then that improvisation becomes the process. Since a held order is one that a customer has already paid for and is already waiting on, it is worth deciding the rules deliberately.
What actually triggers a flag
Most stores and payment processors score orders before they ever reach a warehouse. The common triggers are unremarkable on their own:
- Billing and shipping addresses in different regions
- A shipping address that has been used with several unrelated cards
- Multiple failed authorization attempts before one succeeds
- Order value far above your normal average
- Freight forwarder or parcel-locker destinations
- A rush of identical orders in a short window
None of these prove anything. Plenty of legitimate customers ship gifts across the country, buy far more than average, or retype a card number after a typo. That is the whole difficulty: the signals that correlate with fraud also describe ordinary good customers, which is why an automatic decline threshold set too tightly turns into lost revenue you never see reported.
Who owns the decision
The practical question is where the review happens, and there are only really three arrangements.
| Model | Who reviews | Works best when |
|---|---|---|
| Store-side only | Brand's team, before the order releases | Order volume is moderate and someone checks often |
| Processor rules | Automated scoring, auto-cancel above a threshold | Volume is high and margins tolerate false declines |
| Hold at the warehouse | 3PL pauses the order, brand decides | Orders reach the floor fast and holds must catch up |
The arrangement matters less than making sure only one of them is authoritative. When a store rule and a warehouse rule both act on the same order, the usual result is an order released by one system and still sitting on a shelf in the other.
The part most brands skip: where held orders physically go
A hold is a status in software, but the goods are real. If an order was already picked when the flag came through, something has to happen to the tote or the finished parcel.
Three rules are worth writing down before you need them:
- A defined hold location. Not "the end of the bench." A labeled shelf or cart, so held parcels are never mistaken for outbound ones.
- A time limit. How long a held order may sit before someone must decide — commonly one or two business days.
- A disposition. If the hold becomes a cancellation, is the parcel opened and restocked, or kept intact in case the customer clears review?
Without the third rule, held parcels accumulate. The inventory shows the units as shipped or allocated, the shelf shows something different, and cycle counts start turning up discrepancies nobody can explain.
What a hold costs while it waits
Holds are not free, and the costs are not only the fraud you avoid:
- The pick is done twice if the order is released after restocking
- The units are unavailable to other customers while allocated
- Carrier cutoffs get missed, so a cleared order ships a day or more later
- Customer service handles a contact that would not otherwise exist
That last one is easy to underestimate. A held order almost always generates a message, because from the customer's side nothing has happened since checkout. A short proactive note — that the order is in review and when they will hear back — is usually cheaper than the support thread that starts otherwise.
Holds versus deliverability problems
One distinction worth keeping clean: a fraud hold and an undeliverable address are different problems with different fixes. A flagged order needs a human decision about risk. A bad address needs correction before a label prints, which is a validation step rather than a review step — we covered that path separately in catching bad DTC addresses before the label prints.
Similarly, an order that shipped and then went missing is a carrier matter, not a screening one, and it follows the process described in what to do when an order is lost, damaged, or stolen. Folding all three into one "problem orders" queue is how genuinely urgent cases end up waiting behind routine ones.
A workable setup
For most growing DTC brands, a reasonable configuration looks like this: screening runs at the store or processor, only clearly fraudulent orders auto-cancel, everything ambiguous goes to a review queue a named person checks at least twice a day, and the warehouse receives only orders that have already cleared. Holds that reach the floor get a labeled location and a deadline.
That is a modest amount of structure, and it is mostly about deciding once rather than deciding each time. If you are working through how your order flow should hand off to a fulfillment partner, our DTC fulfillment services page covers where those handoffs normally sit.
