WhatsApp Customer Support Playbook: SLAs, Hours & Handoffs
A field-tested operating manual for WhatsApp customer support: SLA tiers, business hours and away flows, routing rules, escalation paths, CSAT measurement and a sample five-agent operating model.
A customer's payment fails at 9:40 p.m., and instead of filing a ticket she messages the WhatsApp number printed on your website. Whether she becomes a retained customer or a public complaint depends less on your product than on the operating model behind that number. This playbook is that model: hours, SLAs, routing, escalation and measurement for a support channel run properly on the official WhatsApp Business API.
TL;DR
- Run WhatsApp support like a contact centre, not a group of people replying from personal chats: publish hours, define SLA tiers and give every conversation an owner.
- Design the away flow as carefully as the live one — after the 24-hour window closes, you can only reopen a thread with an approved template.
- Route by intent and skill, keep canned replies for structure rather than soul, and escalate on defined triggers instead of gut feel.
- Ask for CSAT once, right after resolution, and review it weekly alongside response and resolution metrics.
- Steal the five-agent operating model at the end and scale it from there.
Why WhatsApp support needs an operating model
Email forgives you: a four-hour reply is unremarkable there. WhatsApp does not. Customers borrowed the channel from friends and family, so silence reads as being ignored — and the platform itself enforces a rhythm through the 24-hour customer service window: free-form replies are allowed only within 24 hours of the customer's last message, after which you need an approved template. We break down the categories and fees in WhatsApp conversation categories and the 24-hour rule.
Speed is table stakes; consistency is the moat
Any team can answer fast on a quiet Tuesday. The test is the Monday sale spike, the agent on leave, the angry refund at 8 p.m. An operating model — tiers, owners, escalation triggers, handover rituals — is what keeps quality flat when volume spikes. It is also what makes the channel measurable, which is the difference between "support seems fine" and knowing that it is.
Set SLA tiers before you buy any tool
A single "reply fast" target guarantees misallocation: agents panic over routine questions while a payment failure waits politely behind them. Split conversations into priority tiers and give each a first-response and resolution target:
| Tier | Typical trigger | First response | Resolution |
|---|---|---|---|
| P1 Urgent | Payment failure, outage, angry customer | Under 5 min | Under 2 hours |
| P2 High | Blocked order, delivery stuck, billing dispute | Under 15 min | Same shift |
| P3 Normal | Product questions, how-to, pre-sales | Under 1 hour | Under 24 hours |
| P4 Low | Feedback, documentation, feature requests | Under 4 biz hours | Under 72 hours |
Treat these as starting points, not gospel — a clinic triaging appointments and a D2C brand mid-sale will calibrate differently. What matters is that the tiers exist, agents know them, and breaches become visible somewhere other than your memory. If you are scaling a shared queue, the setup in our WhatsApp team inbox guide is the natural foundation.
Publish hours, then design the away flow
What a good away message contains
Your away message is your night shift. It should do four jobs in a few lines: set the expectation, offer self-serve, leave an urgent path, and stay human. Greeting and away messages like this are native to WhatsApp's business tooling — the craft is in the copy:
Thanks for messaging {{business}}! We're away right now — we reply between 9 a.m. and 7 p.m., Monday to Saturday. Track your order anytime: {{link}}. For anything urgent, reply URGENT and we'll prioritise it first thing tomorrow.
Configure the same hours in your WhatsApp Business profile, on your website's contact page and in the greeting new chats receive. Mismatched expectations are the root of most "they never replied" complaints.
The after-hours plus 24-hour trap
Here is the operational detail most teams learn the hard way. A customer messages at 11 p.m., your team replies at 9:05 a.m. — fine. But if a thread sits unanswered past the 24-hour window, because of weekends, holidays or staffing gaps, free-form replies are locked and your only way back in is an approved utility template ("Your ticket has been resolved…"). Keep a small set of approved reopening templates ready so a missed thread never becomes a lost customer.
Assignment and routing rules
Routing decides who owns a conversation the moment it arrives. Without rules, ownership becomes "whoever noticed first", which is how threads get double-replied or quietly dropped. Three signals cover most teams:
| Signal | Example | Route to |
|---|---|---|
| Intent keywords | "refund", "invoice", "where is my order" | Billing or orders queue |
| Contact attributes | VIP tag, repeat buyer, wholesale | Senior agent |
| Source | Click-to-WhatsApp ad vs organic inbound | Pre-sales vs support queue |
Keyword-based auto-replies combined with assignment rules in a tool like Whaterakt can triage most of this before a human even opens the queue — see how to automate WhatsApp replies for doing it without over-botting. Language routing matters too in India: a Hindi-first customer should land with a Hindi-fluent agent, not discover the mismatch three messages in.
Canned replies versus the personal touch
Canned replies are compression, not personality. They shine for the skeleton of a conversation — acknowledgement, standard instructions, closing — and collapse the moment a customer is upset or negotiating.
- Canned is fine for: greetings, order-status formats, return addresses, business hours, "looking into this, give me five minutes".
- Never canned for: complaints, refund negotiations, anything with emotion in it, anything the customer wrote more than two sentences about.
The blend that works in practice: open with the customer's name and their specific issue in your own words, use canned blocks for the procedural middle, and close with a personal line plus a clear next step. Agents should be able to type the free parts comfortably — if they can't, your training rather than your template library is the bottleneck.
Escalation paths that actually work
Every support channel needs a spine. Tier 1 agents resolve the routine; a Tier 2 senior or floor lead takes what exceeds scope or authority; a Tier 3 — manager or founder — handles the rare exceptions such as legal threats, high-value disputes and safety issues. The design principle is triggers, not judgement calls:
- SLA breach imminent or already happened.
- Escalation keywords: "legal", "police", "consumer court", "social media".
- Third contact about the same issue — chronic threads burn CSAT and morale.
- CSAT rating of 1 or 2 on a just-closed conversation.
- High-value contact attribute combined with any complaint.
When escalating, the handoff must carry context: a two-line internal note on what was tried and what the customer was promised. A bare "please handle" is how the customer ends up repeating their whole story — the single most corrosive experience in support.
Measuring CSAT on WhatsApp
CSAT belongs at the moment of resolution, not on a monthly blast. The cleanest pattern: when an agent marks a conversation resolved, a single-question rating goes out while you are still inside the 24-hour window, so it stays free-form:
Hi {{name}}, glad I could help today. On a scale of 1–5, how was this chat? One tap is all it takes — it genuinely helps us improve.
Keep it to one question, never ask twice in a thread, and let agents see their own scores — CSAT used for coaching rather than policing is how it stays honest. What counts as good? Well-run channels commonly sit in the mid-80s percent happy or better; treat that as directional and benchmark against your own last month. For the full measurement stack around CSAT — response, resolution and revenue metrics — read the 12 WhatsApp analytics metrics that matter.
A sample five-agent operating model
Here is a concrete roster for a typical SMB handling a few hundred conversations a day, structured so coverage, escalation and handover all have named owners:
| Seat | Coverage | Owns |
|---|---|---|
| Floor lead (senior) | 9 a.m.–6 p.m. | Escalations, QA, CSAT review, templates |
| Agent 1 | 9 a.m.–6 p.m. | Orders and delivery queue |
| Agent 2 | 11 a.m.–8 p.m. | Billing and refunds queue |
| Agent 3 (part-time) | 2 p.m.–8 p.m. | Overflow, documentation, P4 tickets |
| Agent 4 (rotating) | Weekend shifts | All queues, triage only |
Three rituals make the roster work: a 15-minute overlap between shifts for handover, an end-of-shift rule that every open thread gets an internal note and a promised next step, and a Monday 20-minute review of last week's SLA breaches and CSAT lows. The model scales down (two people wearing two hats each) and up (add queue specialists before you add generalists).
FAQ
How fast should a business reply on WhatsApp?
For urgent tiers, minutes; for routine, within the hour during business hours is a defensible target. The real answer is whatever your published SLA says — customers forgive a stated four-hour window far more easily than an unstated forty-minute silence.
What happens if the 24-hour window expires?
You lose the ability to send free-form messages on that thread. Your options are an approved utility template for factual, transactional follow-ups, or waiting for the customer to message you again, which reopens the window. This is exactly why away flows and reopening templates matter.
How many chats can one agent handle?
It varies enormously with complexity — pre-sales catalog questions and refund disputes are different jobs. Teams commonly land somewhere between 15 and 30 active conversations per agent per shift; track your own trend rather than importing someone else's number.
Should sales and support share one WhatsApp number?
One number with routed queues, yes; one number with no separation, no. A shared inbox with intent-based routing gives customers a single familiar contact point while keeping pre-sales and post-sales work separately measurable.
Do we need a chatbot for WhatsApp support?
A thin layer helps: greeting, business hours, order-status lookups, flagging urgent messages. Full bot-led support tends to frustrate, because customers on WhatsApp expect a human quickly. Automate the repetitive 20 percent and staff the rest.
How do we handle support at night or on holidays?
Publish the hours, run the away flow, and decide explicitly whether urgent items get a callback rota or a firm next-morning promise. The mistake is implying round-the-clock availability you cannot staff — that gap generates your angriest threads.
An operating model is boring by design — tiers, owners, rituals — and that is exactly why it survives busy weeks. Start with the SLA table and the away message this week, add routing and CSAT next, and you have a support channel that scales without heroics. Whaterakt gives you the shared inbox, assignment rules, roles and analytics to run this whole playbook on the official API — see pricing and pick the plan that matches your queue size.