The CSAT score is 4.6 out of 5 with an 8% response rate. The product manager treats it as the voice of the customer; the operations director rightly does not. Survey response rates in ServiceNow are mostly a design problem, not a delivery problem, and the default templates are tuned for ease of authoring, not completion.
One question wins
A single-question survey (“Was this resolved to your satisfaction? Yes / No / Partly”) completes at 30-50%. The same survey with five questions completes at 12%. The same survey with ten questions completes at 6%. If you need ten questions of data, send the one-question survey to everyone and a follow-up long-form survey to the subset who answered partly or no.
Trigger timing matters more than wording
A survey sent at incident closure has higher response than one sent the next morning. A survey sent at incident closure but only after the user has confirmed resolution has higher response still. Avoid surveying users whose ticket was closed by SLA timeout — they are not happy and the survey will not change that, but it will erode trust in the survey itself.
// Trigger condition
current.state.changes() &&
current.state == 6 && // Resolved
current.close_code != 'Closed by SLA' &&
current.opened_by.preference_survey != 'opt_out'Frequency capping is non-negotiable
A power user closing 30 tickets a week receives 30 survey emails by default and stops opening any of them. Cap surveys per user per period — one per user per week is a reasonable starting policy. Maintain the cap state in sys_user_preference so it survives across survey definitions.
Question phrasing that does not lead
“How satisfied were you?” is a leading question — most people round up out of politeness. “How likely are you to recommend our service to a colleague?” produces honest answers. “What is one thing we could have done better?” produces qualitative gold. Avoid five-point Likert scales for high-volume surveys; binary or ternary scales reduce decision time and lift response rate.
Anonymous vs attributed
Attributed surveys allow you to follow up on negative responses, which is operationally valuable. Anonymous surveys produce more honest responses, especially on management-related questions. Pick deliberately. If attributed, say so in the survey intro — hidden attribution destroys trust if discovered.
Assessments are not surveys
Assessments score respondents against a scale; surveys collect feedback. Use assessments for skill evaluation, vendor scoring, and internal control attestation — anywhere the output is a quantified rating used for decisions. Conflating the two leads to assessment templates with feedback questions that do not roll up into a score.
Common failure modes
Survey results aggregated only at the global level — break down by support group, fulfiller, and category before any conclusion. Survey response treated as a service KPI without a response-rate floor — a 5.0 score from 4 responses tells you nothing. Survey questions changed mid-period without versioning — versioning the assessment definition (asmt_metric_type) preserves comparability.
Implementation sequence
Run a one-question pilot for 30 days against one team. Measure response rate, satisfaction distribution, and qualitative themes from the optional comment field. Only after that pilot expand to additional questions or additional teams. The instinct to launch a 12-question survey to the whole organization on day one is the most reliable way to get unusable data.
What to do this week: pull the response rate on every active survey definition; anything below 15% is a candidate for redesign or retirement.
