Health Measures
Measuring support deflection without hiding unresolved questions
Separate confirmed self-service resolutions from assumed deflections, support handoffs and unresolved questions when assessing community support value.
Measure the support needs a community resolves, then estimate which of those needs would otherwise have reached a supported contact route. A help-centre visit without a ticket establishes neither resolution nor an avoided contact. Report confirmed resolutions, possible resolutions, hand-offs and unresolved needs separately.
Start with a support need, not a page view
Define an eligible need: the customer had a question or problem the organisation would ordinarily help with, and the community offered a relevant route.
Choose a period in which to follow that need. Keep account-specific matters in the record even when a moderator correctly moves them to private support.
For each sampled need, record the community interaction, any later ticket, chat, call or other supported contact, and the outcome known at the end of the window.
Link repeated contacts about the same problem where the organisation can do so appropriately. If linking is incomplete, disclose that contacts may be duplicated or missed.
| Outcome state | Evidence to require | Reporting treatment |
|---|---|---|
| Resolved without an observed support contact | The member confirms the answer worked, or another agreed check establishes resolution; no related contact appears in covered channels within the stated window | Count as a confirmed community resolution under that rule, not automatically as a proved avoided contact. |
| Possibly resolved | A relevant answer was used and no related contact is observed, but resolution is unconfirmed | Report separately. |
| Handed to support | A support route accepted the issue | Count the handoff and inspect whether it reduced repeated explanation or created extra work. |
| Still open or unknown | No usable answer, incomplete handoff or insufficient outcome evidence | Keep visible in the same report. |
These are working categories, not platform labels. A later contact may show that an earlier answer did not work; update the case rather than preserving a favourable classification. Apply the same follow-up window to comparable needs.
Outcome States in Community Support Deflection Measurement
- Resolved without observed support contact
- Confirmed resolution; no related contact in covered channels within window
- Possibly resolved
- Relevant answer used, but resolution unconfirmed; no related contact observed
- Handed to support
- Support route accepted the issue; assess for reduced repetition or added work
- Still open or unknown
- No usable answer, incomplete handoff, or insufficient evidence
Audit what the platform calls a deflection
Zendesk's self-service dashboard includes ticket-deflection reporting. Zendesk notes that all ticket-deflection reports, including confirmed and assumed deflections, track only tickets submitted using the help-centre request form.
A customer who calls, chats or leaves without a solution may therefore be absent from that particular ticket comparison. These dashboard labels should not be renamed as confirmed community resolutions.
Searches with no useful result, questions with no substantive answer and requests that return after a hand-off belong in the review.
Zendesk's self-service dashboard can be filtered by sessions with search; treat search activity as a diagnostic signal, not an outcome verdict. If the team discourages tickets to improve a ratio, check the resulting customer experience.
Limitations of Zendesk’s Ticket-Deflection Reporting
- Pros
- Tracks only tickets submitted via help-center form; useful for baseline comparison
- Cons
- Excludes calls, chats, and unsolved requests; may overstate deflection if not cross-checked
Estimate avoided work conservatively
For a confirmed resolution, ask whether that type of need would normally have generated a support contact. Some members came to compare ideas and would never have opened a ticket. Others still needed staff to write the answer in the community.
Record that staff time and any later correction or escalation.
Use the organisation's own contact-handling costs and estimate only work plausibly avoided. A confirmed self-service resolution is not automatically a full ticket's saving if an agent spent similar time answering publicly. When the likelihood of a contact is uncertain, show a bounded estimate and keep it separate from the confirmed resolution count.
Report the period, eligible needs, sampling rule, counts in every outcome state, contact channels covered and staff work included.
Key Metrics for Measuring Community Support Deflection
- Eligible needs sampled
- Defined by customer support need and community relevance
- Confirmed resolutions
- Customer confirmed fix or peer validation; no follow-up contact
- Possibly resolved
- Answer used but resolution not confirmed
- Hand-offs to support
- Issue passed to ticket, chat or call channel
- Unresolved or unknown
- No answer, incomplete handoff, or insufficient evidence
