WhatsApp Lead Routing and Assignment Rules

One WhatsApp number, three or ten people answering it. It works fine at twenty conversations a day and falls apart somewhere around sixty. The symptoms are always the same: two agents reply to the same customer within a minute of each other, an enquiry sits untouched for four hours because each person assumed someone else had it, and nobody can say who owns anything.

Routing rules fix this. They are not complicated, and the common mistake is making them far more elaborate than the team can sustain.

The two failures worth designing against

Everything else is secondary to these:

• Double reply. Two agents answer the same customer, often contradicting each other on price or availability. The customer loses confidence immediately, and internally it causes arguments about whose lead it was.
• Silent drop. Nobody replies at all, because the conversation belonged to everyone and therefore to no one. This is the more expensive failure, because you never find out it happened.

Any routing scheme that reliably prevents these two is good enough to start with, even if it is crude.

Start with round-robin and a timer

Before any clever rules, put this in place:

1. Every new conversation is assigned to one agent, automatically, in rotation among those marked available. One owner, always.
2. If the assigned agent does not reply within N minutes, it reassigns. Fifteen minutes is a sensible starting value for sales, shorter if you are competing on speed.
3. Reassignment is visible. The next agent can see it was passed on, so nobody assumes it is fresh.

That is the whole first version. No categories, no tags, no manager approving anything. It eliminates both failure modes and takes an afternoon to configure. Most teams that think they need sophisticated routing actually just needed this.

Why the timer matters more than it looks: the agent who is genuinely busy is also the agent least likely to say so. A timer handles that without anyone having to admit anything.

Rules worth adding, in order of value

Once round-robin is running, add these only when you can point at a real problem they solve.

1. Language. In India this is usually the highest-value rule. A customer writing in Marathi, Tamil or Bengali converts noticeably better with an agent replying in the same language, and worse with one who switches them to English. If your team has that mix, route on it.

2. Returning customer to the same agent. If a contact has spoken to someone before, send them back to that person. The customer does not re-explain, and the agent already knows the history. This is cheap to implement and customers feel it immediately.

3. Product or service line. Only where your agents genuinely know different things. If one person handles AMC and another handles new installations, route on the first message content or a menu choice. If everyone can answer everything, this rule only adds delay.

4. Value or source. Leads from a paid campaign, or above a certain deal size, to your stronger closers. Worth it when the spread between your best and average agent is large.

5. City or territory. Last, and only when territory is real — a technician who covers that area, or a branch that serves it. Routing by city when it changes nothing is a common way to make a simple system complicated.

Availability has to be honest

Routing only works if the system knows who is actually free. Three things make that true:

• Agents mark themselves available or away, and it is normal to be away. If marking away is treated as slacking, nobody does it and the queue routes to people on lunch.
• Caps per agent. Beyond a certain number of open conversations, stop assigning. An agent holding thirty open chats is replying slowly to all thirty.
• Shift awareness. The rota should drive availability, not memory.

After hours is a routing decision, not an accident

A large share of Indian consumer enquiries arrive after 8pm. What happens to them is a design choice:

• Queue them, do not drop them. They should be assigned at the start of the next session, not left in an unassigned pool that somebody has to remember to check.
• Auto-reply with a specific time. "We will reply by 10am tomorrow" holds people. "We are closed" does not.
• Morning order matters. Oldest first, not newest first. The 9pm enquiry has already waited twelve hours.

Handover rules, which matter as much as routing

Assignment decides who starts. Handover decides what happens when that person cannot finish. Without rules here, conversations get silently abandoned at shift end.

• Nothing changes hands without a note. One line on what the customer wants and what was promised. Handovers without context make the customer repeat themselves, which is the single most irritating thing in a support conversation.
• Shift end is a sweep, not a hope. Open conversations are reassigned to the incoming shift explicitly.
• Escalation is a route, not a guess. Complaints, refunds and anything legal go to a named person, immediately, not around the floor.

If a bot handles first contact, the handover from bot to human needs the same discipline — the trigger phrases and the timing are covered in chatbot versus live agent.

Measure three things, not twenty

Routing can be measured to death. Three numbers tell you whether it is working:

1. First response time, by agent. If one person is consistently slow, that is a coaching conversation or a capacity problem, not a routing problem.
2. Reassignment rate. How often the timer has to step in. Rising means someone is overloaded or availability is dishonest.
3. Unanswered after one hour. This should be close to zero. Anything else means leads are still leaking.

What good looks like for the first of these is in WhatsApp response time benchmarks for India, and the wider set is in the metrics that actually matter.

The mistake almost everyone makes

Designing a routing tree with eight branches before the team has ever worked from a shared inbox. It looks thorough and it fails for a predictable reason: every branch is a rule someone has to maintain, and when reality does not match the tree, conversations fall into gaps nobody checks.

Start with round-robin and a fifteen-minute timer. Run it for a month. The rules you actually need will announce themselves — usually language first, returning-customer second. Add those two and most teams never need a third.

Routing when there are only two of you

Most advice assumes a floor of agents. A large share of Indian businesses doing this well have exactly two people answering — often an owner and one assistant — and the rules are different at that size.

At two people, round-robin is usually wrong. The owner and the assistant are not interchangeable: one can quote and commit, the other cannot. What works instead:

• Everything lands with the assistant first. They answer the common questions, collect the details, and handle anything procedural.
• Escalate on content, not on rotation. Pricing beyond a set range, custom scope, complaints, and anyone who asks for the owner by name.
• The owner's queue is visible to both. The failure mode at this size is the owner being the bottleneck without anyone being able to see it.

The timer still matters, but it should escalate upward rather than sideways. If the assistant has not answered in fifteen minutes, the owner sees it — not because the assistant is unreliable, but because at two people there is no third option.

What to do with repeat and spam contacts

Two categories quietly consume agent time and neither needs a human:

• People who message every campaign and never buy. Worth tagging after the third cycle. Not to ignore them, but so they do not sit at the top of the queue ahead of a live enquiry.
• Wrong numbers and job seekers. Every business number in India gets these. A quick tag and a saved reply handles them in one tap rather than a conversation.

Neither should be deleted. Both should be out of the queue the agent works from, because the cost of routing is attention, and attention is the scarce thing.

Frequently Asked Questions

What is the simplest routing rule that works?

Round-robin among whoever is marked available, plus a timer that reassigns if the owner has not replied within about fifteen minutes. It needs no categories, no tagging and no coordinator, and it removes the two worst failures — double replies and conversations nobody owns.

Should leads be routed by language or by city?

Language is usually the higher-value rule in India, because a customer writing in Marathi or Tamil converts noticeably better with an agent who replies in the same language. City only matters when territory is real — a technician who covers that area or a branch that serves it.

What should happen to messages that arrive after hours?

They should be queued and assigned at the start of the next working session, not left in an unassigned pile. An automatic reply stating a specific time — "we will reply by 10am" — holds the customer, and the queue makes sure the morning shift actually sees it, oldest first.

How do I stop one agent from hoarding leads?

Cap open conversations per agent and watch the reassignment rate. An agent holding thirty open chats is replying slowly to all of them, which looks like ownership and behaves like a bottleneck. Caps fix it without anyone having to be accused of anything.

How many routing rules is too many?

If a new team member cannot explain where a lead goes in two sentences, you have too many. Every branch is a rule someone must maintain, and conversations fall into the gaps between branches nobody checks. Most teams need round-robin, a timer, language, and returning-customer — and nothing else.