Card integration is the easy part. Cash on delivery and BNPL are where UAE ecommerce builds get expensive, and both change your order model rather than just your checkout.
The UAE payment mix is not the Western one
A checkout designed around card payments will underperform in the UAE, and the reason is structural rather than technical. Cash on delivery remains a meaningful share of retail orders across the region, and buy-now-pay-later has been adopted quickly at the mid-basket range.
This matters because COD and BNPL are not payment methods you toggle on. Both change what an order means in your data model, and retrofitting them into a system that assumed card-on-checkout is where UAE ecommerce projects overrun.
Cash on delivery changes your order model
With card payment, authorisation and order creation happen together. With COD, an order exists for days with no money attached, and the money arrives as physical cash held by a courier who remits on their own schedule.
Your system needs to model that gap explicitly. That means an order state for 'awaiting cash collection' distinct from 'paid', reconciliation against courier remittance files, a defined process for partial deliveries where a customer accepts some items and refuses others, and handling for failed delivery where stock returns and no money was ever collected.
COD also has a fraud profile card payments do not: order-and-refuse costs the customer nothing and costs you shipping both ways. Most UAE operators eventually implement per-customer COD limits or a verification step above a basket threshold. Build the hook for that early even if you do not enable it at launch.
Comparing the main gateways
Network International is the largest regional acquirer with strong bank relationships across the Gulf. Best suited to established merchants with volume; onboarding is more involved and typically requires a UAE trade licence and corporate bank account.
Telr is aimed at SMEs, with faster onboarding and a simpler integration surface. Supports AED and multi-currency. A reasonable default for smaller catalogues and early-stage merchants.
PayTabs sits between the two, with good regional coverage beyond the UAE. Worth choosing if you plan to expand into Saudi Arabia or Egypt, since it saves a second integration later.
Stripe now operates in the UAE and has the best developer experience of any option here. The trade-off is weaker support for local rails — you will still need a separate solution for COD, and BNPL coverage differs.
BNPL: Tabby and Tamara
Both are widely used in UAE retail and both work similarly — the provider pays you in full, takes a merchant fee, and carries the customer credit risk.
The integration detail that matters is eligibility. Approval depends on basket value, customer history and the provider's own risk model, so you cannot assume availability. The correct pattern is to query eligibility server-side before rendering the option, so customers do not select BNPL and then get declined at the final step. A declined BNPL attempt at the last screen is a lost order, not a downgraded one.
Also plan the refund path. BNPL refunds flow back through the provider rather than directly to the customer's card, and the timing differs from card refunds. Support teams need to know that.
VAT and invoicing
UAE VAT is 5% and applies to most goods and services, with specific categories zero-rated or exempt. Your catalogue needs a VAT treatment per product rather than a single global rate, and your invoices must show the VAT amount separately with your TRN.
For COD orders the tax point is a common source of confusion. Get accounting sign-off on whether you recognise at dispatch or at cash collection before you build the reporting, because changing it afterwards means reprocessing historical orders.
A sensible build order
Model orders to support unpaid states before integrating anything. If your schema assumes payment at creation, every subsequent step fights it.
Then card via your chosen gateway, then COD with reconciliation, then BNPL last — it is the most self-contained and the easiest to add once the order model is right.
Sandbox thoroughly. All the providers above offer test environments, and the failure modes worth rehearsing are the unglamorous ones: partial delivery, BNPL decline at final step, courier remittance mismatch. Those are what actually generate support load in month two.