Building a survey
Surveys are built in the console (console.html, supervisors only). Keep them short: three to five questions, under a minute, is where completion rates stay high.
Question types
| Type |
Scale |
Reported as |
Counts for the agent by default |
| Rating |
1 to 5 |
CSAT (% of 4s and 5s, and the average) |
Yes |
| NPS |
0 to 10 |
NPS (% promoters minus % detractors) |
No |
| Effort (CES) |
1 to 7 |
average effort |
No |
| Yes / no |
1 or 0 |
share of yes |
No |
| Multiple choice |
one of 2 to 9 options |
counts per option |
– |
| Free text |
typed comment |
listed with the response |
– |
Every scored answer is also brought to a 1-5 scale so surveys with different questions report together, and so Arena receives one scale.
Counts for the agent. Tick this on questions about the agent's handling (satisfaction with the service); leave it off for questions about the company (would you recommend us, was it easy, was it resolved). Only agent-attributable scores go into the agent report and to Arena, which keeps agents from being marked down for pricing or policy.
Driver. A short label for the report (Wait time, Resolution). Questions without one are reported under their text.
Required. Scored questions are required by default; free text is optional unless you tick it. In chat, a customer skips an optional question by replying with a dash.
Asking why only when it matters
Any question can carry a rule: ask only when an earlier question is at most N. The starter survey asks "What could we have done better?" only when satisfaction was 3 or lower. This keeps happy customers' surveys short and gives you a comment on every poor one.
Who gets which survey
Each survey has a target:
- Queues: the survey applies to contacts handled in any of the listed queues.
- Contact attribute: the survey applies when the contact carries the attribute, with a specific value or any value. Set the attribute in your flow (for example
survey = sales).
- Neither: the survey is the default for any contact.
When several surveys match, the attribute match wins, then the queue match, then the default. One survey per contact, ever.
Deciding in the flow. Often the inbound flow already knows whether a contact should be surveyed, before any agent is involved. Set the contact attribute surveycx to skip and no survey runs on any channel for that contact, whatever the targeting; set it to a survey's id to force that survey. The attribute can be set by a Set contact attributes block, from CRM data in a Lambda, or by the agent desktop through the Connect API. The same applies to contacts reported through the API: "attributes": { "surveycx": "skip" }.
Channels limit where the survey runs: voice (IVR), chat (in the chat), SMS and email (invitations with a web link), web. A survey with no SMS or email channel never sends an invitation.
Sample % surveys only that share of eligible contacts, chosen deterministically per contact.
Low score
Each survey sets the low-score rule: any scored answer (or one question) at or below a threshold on the 1-5 scale flags the response and opens a follow-up Task. See Closed loop.
Pausing and preview
A paused survey stops being offered immediately; answers already collected stay. Preview opens the web survey exactly as a customer sees it, without saving anything.
Languages
Set the survey's language tag (fr-CA, es-US) so the web page is marked correctly for screen readers. Write the questions in that language; one survey per language, targeted by queue or by a language contact attribute, is the simplest way to run several.
Current as of version 0.1.0. See the release notes for what changed since.