MERSO IT

Automation · Customer messaging

Zalo OA and ZNS automation: order updates and remarketing

MERSO IT Insights · Updated July 2026 · 8 min read

Automated Zalo notification flow sending order updates from a shop system to a customer's phone

Why Zalo first: for a Vietnamese customer base, Zalo is where order updates actually get read. Email open rates are poor, SMS is expensive per character and impersonal, but a ZNS message with the tracking number lands in the same app the customer already uses all day. Done right, Zalo becomes both your delivery-notification channel and — carefully — your remarketing channel. Done wrong, you burn weeks on template rejections and get your OA restricted.

Two channels, two rulebooks

Everything on Zalo splits into two mechanisms with different rules:

  1. ZNS (Zalo Notification Service) — pre-approved template messages sent to a phone number. The customer does not need to follow your OA. Built for transactional notifications: order confirmed, handed to courier, delivered, COD collected. Every template must pass Zalo's review before you can send it.
  2. OA messages — conversational messages to users who interact with your Official Account. Consultation-style replies are generally tied to a window of roughly 48 hours after the user's last interaction; outside it, your options narrow sharply and shift toward paid or restricted message types.

The automation consequence: transactional flows ride on ZNS keyed by phone number, while care and remarketing flows ride on the OA and must be engineered around the interaction window.

The transactional backbone

Wire your order pipeline so each state change fires the matching template: order confirmation at checkout, tracking number when the courier webhook reports pickup, a delivery confirmation, and a COD receipt where relevant. If you already automated fulfilment as described in our ViettelPost integration guide, these are the same webhooks — you are adding one more subscriber, not a new pipeline. This is the messaging layer of the end-to-end stack.

Designing templates that pass review

Template review is the single biggest schedule risk in a Zalo project. The most common rejection we see: promotional wording inside a transactional template. "Your order has shipped — use code X for 10% off next time" reads like a delivery update to you and like an advertisement to the reviewer. Rules that keep approval smooth:

  1. One purpose per template. A delivery notice notifies about delivery. Nothing else.
  2. Every parameter earns its place — order code, tracking number, amount. Reviewers reject vague free-text parameters that could smuggle arbitrary content.
  3. Write the Vietnamese customer-facing text carefully and formally; sloppy tone invites scrutiny.
  4. Submit templates early and in a batch. Review takes days, and you will iterate.
Plan for rejectionTreat template approval as a project dependency with lead time, not a checkbox. Keep a template register: purpose, parameters, status, rejection reasons. Teams that track this ship in one review cycle; teams that don't discover a rejected template the night before launch.

Care flows inside the 48-hour window

When a customer messages your OA — a question, a complaint, a reaction to a campaign — the window opens. This is where your chatbot earns its keep: instant answers on shipping fees, sizes and order status, with human escalation for hard cases, exactly as we architected in the customer service bot article. The automation subtlety is window bookkeeping: record the timestamp of every inbound interaction per user, and have flows check remaining window time before sending. A follow-up scheduled naively for "24 hours later" will silently fail — or violate policy — if the window has closed.

Remarketing within policy

Segmentation is legitimate and useful: tag OA followers by product interest, purchase recency and interaction history, then send targeted content to users you are entitled to message. What gets shops in trouble is pretending remarketing is transactional. Keep the two lanes separate in code, not just in intention: the transactional sender should be technically unable to send marketing content, and the marketing sender should check consent, tags and window state before every send.

Consent and opt-out under Decree 13

Vietnam's Personal Data Protection Decree (13/2023/NĐ-CP) makes consent and purpose limitation concrete obligations. Collect marketing consent explicitly, store when and how it was given, and honour opt-outs immediately across all channels — a customer who opts out on Zalo should not resurface in your SMS list. Build opt-out as a first-class event in your customer store, not a manual spreadsheet task.

Phone numbers: the silent killer

ZNS targets phone numbers, and Vietnamese numbers arrive in your systems in every format: 0912…, +84912…, 84912…, with spaces and dots. Normalise to one canonical format at ingestion — every checkout, every import, every chat handoff — and validate before sending. A large share of "ZNS not delivered" tickets we investigate are normalisation bugs, not platform issues.

Cost control and quota management

Message typeCost profileAutomation rule
ZNS transactionalPaid per message, by template tierSend on state change only; dedupe per order and state
OA reply in windowLow / includedPrefer for care; bot answers first
Marketing sendsPaid, quota-boundBudget per campaign; segment tightly

Meter everything: a counter per template per day, alerts on unusual spikes (a retry loop can spend your monthly budget in an hour), and a monthly cost-per-order figure so messaging spend stays visible next to your courier and gateway fees in the same ledger you use for COD reconciliation.

Sequencing the build

Ship delivery notifications first — highest customer value, purely transactional, easiest approval. Then the care bot, then segmentation, then campaigns. Zalo also slots into your broader content distribution alongside Facebook and TikTok; see our auto-posting article for that layer.

Want Zalo notifications and remarketing on autopilot?

MERSO IT builds ZNS pipelines, care bots and compliant segmentation — and for suitable projects we demonstrate the core workflow before any payment.

Talk to Mersoid, our AI consultant