Start with these five checks.
- Separate customer-confirmed resolution from system closure.
- Link contacts by customer, intent and episode window.
- Compare clean resolutions with repeat-contact paths.
- Inspect transfers, reopenings and policy exceptions.
- Quantify avoidable contacts and customer impact by cause.
Ask for only what can change the answer.
- InteractionsTickets, calls, chats and transcripts
- Episode linksCustomer, intent and repeat window
- WorkflowRouting, transfers and close events
- KnowledgeArticles, policy and version used
- Product contextIncident, release and known issue
- OutcomeResolution, effort, refund and churn
Test one reversible move.
Choose one common intent and four weeks of completed episodes. Replay a proposed close, routing or knowledge rule and review every episode whose outcome changes. Run a small live cohort only if repeat contact falls without false closure, unnecessary transfer or policy violations.
A decision your team can use.
- 01A contact-episode dataset
- 02A ranked failure-path map
- 03Customer and workflow root causes
- 04A guarded historical replay
Common questions.
What counts as a repeat contact?
A new contact about the same underlying need within a stated time window; define both intent and window explicitly.
Can ticket data alone answer this?
Often not. Calls, customer confirmation and workflow events can reveal failures hidden by ticket status.
Should we add more agents?
Only after arrival patterns, coverage and avoidable repeat demand are separated.