REFERENCE · RETURNING CUSTOMERS
A guide for the team

One customer, one thread — no matter the channel.

When a customer comes back about something while their last contact is still in flight, the right answer is almost never to start a new conversation from scratch. Atender recognises them — across every channel — and offers to continue where they left off. This page covers the three ideas that shape the feature: who counts as returning, how Atender confirms it, and what happens once it’s confirmed.

i
A note on terminology.
Detection is the moment Atender realises an incoming contact already has an open conversation. Triggers decide which existing conversations qualify as a candidate for matching. The customer-facing question — “is this the same issue, or something new?” — is the prompt. The timeout action is what runs if the customer doesn’t answer the prompt in time. Each lives in its own settings panel, configured one channel at a time.
The match

Who counts as returning

A returning customer is one Atender already knows on this account, with at least one conversation that hasn’t been put away yet. The match is based on identity — not on which inbox the message came through.

1Identity, channel-agnostic — how Atender recognises the contact in the first place
☎
Phone
Caller ID matches the contact
✉
Email
From-address matches the contact
✉
SMS
Sender number matches the contact
☺
Web chat
Session, cookie, or signed-in identity
The match crosses channels on purpose. A customer who emailed yesterday and calls today gets connected to that same conversation. The detection key is the contact, not the inbox — which is exactly what a busy customer service team needs when one issue arrives by three routes.
Identity is necessary; it isn’t sufficient. The conversation also has to qualify ↓
2Triggers — which existing conversations are eligible to match
Trigger default: on

Unanswered conversations

The customer has an active conversation where the ball is still in your team’s court — the last message came from the customer, not from an agent. The most common case: a follow-up while the original is still waiting on a reply.

Trigger default: on

Snoozed conversations

An agent has parked the conversation for later. Useful when the customer can’t wait and reaches back out before the snooze expires — you’d rather pick the same thread back up than start fresh.

Trigger default: on

Recently resolved conversations

The customer’s previous conversation was marked done, but only just. Within the lookback window, “done” still counts as “recently in play” — perfect for quick “wait, one more thing” follow-ups.

Archived is always out. An archived conversation never matches, regardless of triggers. The four lifecycle states behave the same way they do everywhere else in Atender: active, snoozed, and done are eligible for matching (gated by triggers); archived is the line that says “this is over.”
And one global cut-off, set per channel ↓
3Lookback window — how far back “recently” reaches
Per channel

Hours, configurable

Defaults to 24 hours; tune it to match your average resolution time. If your team typically closes things in minutes, a long window will stitch unrelated contacts together. If you regularly take days, a short window will miss the customers most likely to call back about the same thing.

The confirmation

How Atender confirms it

Identity plus eligibility tells Atender there’s a candidate. Before doing anything irreversible, it asks the customer — in the way that fits the channel they came in through.

1The prompt — one question, in the grammar of the channel
Across channels

A one-beat confirmation, shaped to the medium

The exact form of the prompt changes from one channel to the next — voice, chat, and messaging each have their own conventions for how a person answers a quick yes-or-no with the least friction. The intent is the same everywhere: the customer answers in a single beat, and Atender knows which thread the next message belongs in.

Email is the deliberate exception. Atender doesn’t ask on email. Reply-based threading already implies continuity, so a matched email is auto-merged into the existing conversation by default — one less round-trip in a medium that’s already slow.
The prompt copy is yours to write. Each channel ships with a sensible default and you can override it per channel. If you want the prompt to read like a turn-away (“you already have an open conversation; we’re still on it”), the wording is a setting, not a code change.
If the customer doesn’t answer, Atender still has to do something ↓
2Response timeout — how long to wait for a reply before falling back
Per channel

Minutes, configurable

Defaults to 60 minutes. After the timeout, the prompt is no longer accepted and the channel’s configured timeout action runs — the next concept covers what those actions look like.

The resolution

What happens once it’s confirmed

There are two ways the answer arrives: the customer responds to the prompt, or the prompt times out. Either way, the result is one of three actions — with two channel-specific shortcuts on top, for the places they earn their keep.

1The three timeout actions — what runs if the customer doesn’t answer the prompt
Action Do nothing prompt was the answer

Leave it where the prompt left it

Nothing further happens automatically. If the prompt copy itself directs the customer back to their open conversation, this action is, in effect, a turn-away — the prompt told them what to do and you trust them to do it.

Action Add to existing presume continuation

Append silently to the matched conversation

Treat the new contact as a follow-up on the existing thread. The agent picking it up sees the entire history, including everything the previous handler tried — no “can you explain again?” on the customer’s side.

Action Create new presume new issue

Start a fresh conversation as if not matched

Useful when your team’s workflow treats every contact as its own ticket, or when the cost of accidentally bundling two unrelated issues is higher than the cost of a duplicate.

And two shortcuts that only make sense on specific channels ↓
2Channel-specific accelerators — phone and email get extra options that earn their keep
Phone skip the menu

Skip IVR for returning callers

Bypass the phone tree entirely and connect the caller to an agent. Once a caller is recognised as returning, navigating the menu again is just friction — you already know roughly what they want.

Phone jump the queue

Priority queue placement

Available once Skip IVR is on. Returning callers go to the front of the agent queue, ahead of new callers. Sensible default for second-attempt callers; double-check it against your queue policy before flipping it on.

Email no prompt round-trip

Auto-merge without asking

The default behaviour for matched emails. If reply-based threading found the same thread, that’s usually evidence enough; asking the customer to confirm in a fresh email would be slower than the resolution itself.

All channels override the copy

Custom prompt per channel

Replace the default prompt text on any channel — tone, language, instructions for what to do if neither option fits. Leave it blank to use the shipped default.

This feature was built for busy teams. Customer service at scale means people reach out through whichever channel is fastest for them — phone today, chat tomorrow, email when they remember. Returning customer detection is the seam that turns that multi-channel reality into a single-thread experience: one customer, one issue, one conversation, no matter how many doors they walk through. Inboxes stay calm, the next agent picks up exactly where the last one left off, and customers never have to explain themselves twice.
At a glance

How the three ideas fit together

Identity opens the door, triggers decide who’s eligible, the prompt confirms, and the action resolves. Tune each per channel; let archived conversations stay closed.

A returning customer
Match who returned
Identity (channel-agnostic)
Phone Email SMS Web chat
Triggers
Unanswered Snoozed Recently resolved
Plus a per-channel lookback window. Archived never matches.
Ask confirm before acting
Confirmation style
Channel-native prompt Email: silent merge
Response window
Timeout, in minutes
Prompt copy is editable per channel.
Resolve the outcome
Timeout actions
Do nothing Add to existing Create new
Channel accelerators
Skip IVR Priority queue Email auto-merge
“Do nothing” can act as a turn-away — the prompt was the answer.