See how fast you answer, and how satisfied customers are
SLA policies put a deadline on every conversation. CSAT surveys ask the customer how it went. Both are reported in Analytics, on the same filters as the rest of your numbers.
500 conversations, no credit card.
What's in CSAT & SLA
- SLA policiesOne metric per policy, with three rising thresholds on it.
- CSAT surveysOne question rated 1 to 5, shown as emoji, stars or numbers.
- Opening hoursThe countdown pauses while the team is closed, unless you say otherwise.
- Warnings and breachesWho is told, which webhook fires, and what a rule does about it.
- SLA reportingCompliance figures, broken down by policy, team, agent, channel and market.
- Follow-up QuestionsA second question that depends on the score you got.
- DeliveryEmail or SMS, and what happens on channels that cannot carry a survey.
- Prevention WindowThe least number of days between two surveys for one customer.
- LanguagesA default language plus the ones you support, each with its own wording.
- Satisfaction reportingSatisfaction, average rating, response rate and the 1–5 split.
- What customers wroteThe comments themselves, and an AI summary of the themes in them.
- Export and APICSV, the REST API, MCP tools and a BigQuery view.
One metric, three thresholds, one countdown
A policy measures one thing — First Reply, All Messages, which is every next reply, or Resolution Time — and puts three rising thresholds on it: At risk, then Urgent, then Breach, set in days, hours, minutes or seconds. A live countdown badge then sits on the conversation and in the list, so nobody has to work out what is closest.
The countdown pauses outside the team's and the channel's opening hours, and it knows about timezones and holidays. A policy that has to run around the clock is set to "Ignore office hours" instead. Policies apply per team and channel; the conversation's priority picks which of the three thresholds it is measured against. There is no manual pause on a single conversation.
What happens when a deadline gets close
Each level the conversation crosses notifies the person it sits with and fires a webhook — sla.warning on the way up, sla.breached at the end. The same events start automation rules, so escalation is something you write rather than something the product decides for you: reassign to a team lead, raise the priority, post to a channel, or all three. Automation is where those rules are built.
- The owner is notified at each level
- sla.warning and sla.breached fire as webhooks
- Automation rules can act on both — escalation is a rule you write
Are we fast enough?
That is the name of the page in Analytics, and it answers itself with three figures in the Promises band: first-response SLA compliance, resolution SLA compliance, and the two combined. Conversations that no policy covers are left out of the rate and counted on their own, so the number is not quietly propped up by work nobody promised anything about.
- SLA policy is a filter and a breakdown, next to team, agent, channel, market, priority and tag
- A "No SLA policy" bucket holds everything no policy covers
- Clocks run in business hours unless the policy ignores office hours
- SLA fields reach BigQuery through the conversations view
One question, rated 1 to 5
The survey asks for a rating from 1 to 5, shown as Emoji, Stars or Numeric. You write the question yourself and it can name the agent who answered. A comment field is optional, and Follow-up Questions can be attached to particular scores — one more question, text or rating, for the people who were not happy.
The survey page is hosted, carrying your logo and colours, and the thank-you message at the end is yours to write. The Languages tab holds a default language plus the ones you support, each with its own wording.
- Rated 1 to 5 — no NPS scale, no thumbs
- Invitations use a basic template or one from Email Studio, with your own From name and address
- A verified sending domain is needed before surveys can be switched on
Who gets a survey, on what, and when
Surveys go out by email or SMS. Pick whether one follows the conversation's own channel or always uses the same one, and set a Fallback Channel for conversations whose channel cannot carry a survey at all — a phone call, a web chat.
Sending is driven by automation rules. Quick Setup turns on the common case in one switch: send a survey when a conversation is archived. For anything narrower, the Send CSAT Survey action goes in a rule of your own, so team, channel or tag decides who is asked.
- Prevention Window — the least days between two surveys for one customer, 90 by default
- Held until it is between 09:00 and 20:00 where the recipient is
- Never sent to people who unsubscribed, contacts with no valid address, colleagues on your own domain, test conversations or unanswered calls
- Survey links expire after 30 days by default
Did we solve it?
The satisfaction page carries Satisfaction — the share of answered surveys that count as satisfied — beside average rating, response rate, open rate, send rate, coverage and satisfaction by topic. You decide where satisfied starts, at 4 or only at 5, and whether neutral answers are counted in the total at all. Satisfaction over time shows the 1–5 split, and each rating opens the surveys behind it.
- Filters: team, agent, channel, market, priority, tag, SLA policy, organization and custom fields
- Channel is the conversation's channel, so a survey emailed after a call counts under Voice
- A per-agent satisfaction column sits alongside
- AI-handled conversations have no team, so they appear as Unassigned
The comments, and what they add up to
Every comment is listed with its score and a link back to the conversation it came from. Generate AI summary reads them together and comes back with the overall sentiment, up to five things customers love and five that need attention — each with the number of mentions behind it — real quotes from the comments, and up to three recommended actions. It reads up to 100 comments, and it says no when there are too few to be worth summarising. The product labels the result as AI commentary, because that is what it is.
Take the numbers somewhere else
Satisfaction data does not have to stay in Atender. Pull it out as a file, read it from your own code, or let it land in the warehouse with everything else.
CSV export
The last 7, 30, 90 or 365 days, or all of it.
REST API and MCP tools
Read surveys and their settings, and record a rating, from your own code or from a coding agent over MCP & API.
BigQuery
Surveys arrive as their own view; SLA fields come through the conversations view.
SLA policies and CSAT surveys are set up here and show up in the rest of the platform — as rules that fire, numbers in reports, and a status on the wallboard.
Automation
Rules act on SLA warnings and breaches, and the Send CSAT Survey action decides who is asked.
Analytics
Both reports live here — Are we fast enough? for SLA, Did we solve it? for satisfaction.
Monitor
SLA status sits beside queue depth and agent availability on the live view and the wallboard.
MCP & API
Read surveys and settings, record a rating, and receive sla.warning and sla.breached as webhooks.
The deadline, and the answer.
Email and SMS. You choose whether a survey matches the conversation's own channel, always goes by email, or always goes by SMS. Conversations on a channel that cannot carry one — a phone call, a web chat — use the Fallback Channel you set. There is no survey inside the chat widget, in WhatsApp or Messenger, or in the IVR.
One rating from 1 to 5, shown as Emoji, Stars or Numeric — you pick which. The question is yours to write and can name the agent, there is an optional comment field, and Follow-up Questions can be attached to particular scores. There is no NPS scale of 0 to 10 and no thumbs up or down.
The Prevention Window sets the least number of days between two surveys for one customer, and it starts at 90. Beyond that, surveys are not sent to people who have unsubscribed, contacts with no valid email address or phone number, colleagues on your own domain, test conversations, or calls nobody answered. A survey is held until it is between 09:00 and 20:00 where the recipient is, and the link expires after 30 days by default.
Yes, on the same rules as any other conversation. They have no team behind them, so in the reports they appear under Unassigned. There is no separate AI-versus-human satisfaction metric.
It is the share of answered surveys that count as satisfied. You decide where satisfied starts — at 4, or only at 5 — and whether neutral answers are counted in the total at all. Average rating, response rate, open rate, send rate and coverage are reported beside it.
In Analytics, on the Did we solve it? page. The rating is not shown on the conversation in the agent workspace. Comments are listed there with a link back to the conversation they came from.
Team and channel, and nothing else — not tags, not market, not language. The conversation's priority does not pick the policy; it picks which of the policy's three thresholds applies.
Yes. The countdown pauses outside the team's and the channel's opening hours, and it knows about timezones and holidays. A policy that has to run around the clock is set to "Ignore office hours" instead. There is no manual pause on a single conversation.
No. Both are off until you set them up. SLA policies start working as soon as you write one. CSAT surveys need a verified sending domain before they can be switched on.
See Atender in action
Try the full platform for free. Or book a demo and we will walk you through it.
500 conversations, no credit card.