FullSession | How to reduce support tickets without hiding the problem | Roman Mohren | October 2026 Article: https://www.fullsession.io/blog/reduce-support-tickets REPEAT-QUESTION REVIEW WORKSHEET A reusable working tool for one repeat question at a time. Every field is blank on purpose: fill it in from your own conversations. Nothing here reports a completed review, a test or a result, and it contains no benchmark to aim for. Set targets from your own baseline. HOW TO USE THIS FILE Part A fixes the period and the cluster so the next period counts the same thing. Part B is the conversation review log. Read each conversation in full; do not sort by keyword alone. Part C tallies causes and separates avoidable questions from legitimate support needs. Part D records evidence for product friction and defects. Part E is the fix plan: one named owner per fix, plus the interim answer. Part F checks that the route to a person stays open while you work. Part G compares the next period with the baseline. Part H records the decision. Copy the whole file for each new cluster. PART A: PERIOD, CLUSTER AND BASELINE Reviewer: Date of review: Baseline period (start date, end date, length in days): Cluster name (what the customer is trying to do, in plain words): Cluster definition (what counts, in one or two sentences): What does NOT count (near-misses to exclude): Channels included (website chat, in-app chat, helpdesk email, other): Denominator for the rate (active customers, accounts, orders or visits) and its value for the period: Baseline counts for the period: Conversations in the cluster: Distinct customers in the cluster: Total conversations in the period (all topics): Cluster rate (cluster conversations per 100 of the denominator): Repeat contacts (same customer, same issue, within 7 days): Requests for a person within the cluster: Downstream signals citing the same cause (refunds, chargebacks, cancellations, reviews): PART B: CONVERSATION REVIEW LOG Sample size planned (20 is a reasonable starting point; read all if fewer): For each conversation, record: Ref = conversation reference Customer's words = the question as the customer wrote it Answer and source = what the AI agent or teammate answered, and which article, page or policy it came from Outcome = solved / not solved / unknown Person requested = yes / no Cause = see Part C codes Evidence available = transcript only / account data / recording captured / recording not captured / not used # | Ref | Customer's words | Answer and source | Outcome | Person requested | Cause | Evidence available 1 | | | | | | | 2 | | | | | | | 3 | | | | | | | 4 | | | | | | | 5 | | | | | | | 6 | | | | | | | 7 | | | | | | | 8 | | | | | | | 9 | | | | | | | 10 | | | | | | | 11 | | | | | | | 12 | | | | | | | 13 | | | | | | | 14 | | | | | | | 15 | | | | | | | 16 | | | | | | | 17 | | | | | | | 18 | | | | | | | 19 | | | | | | | 20 | | | | | | | Keep sensitive data out of this log: no payment details, passwords or personal identifiers beyond the conversation reference. PART C: CAUSE TALLY Cause codes: C1 Missing or unclear content (avoidable) C2 Correct content, but not where the question came up (avoidable) C3 Policy or rule customers did not know (avoidable if shown where they decide) P1 Product friction: the flow works, but people fail at it (needs a product fix; the customer may need a person meanwhile) P2 Defect: something did not work (legitimate; stays with a person) L1 Account-specific change or exception (legitimate; stays with a person) L2 Judgement call: refund outside policy, contract, security, data deletion (legitimate; stays with a person) L3 Complaint (legitimate; stays with a person) U Unclear after reading (note what is missing) Code | Count in sample | Typical words customers used | Notes C1 | | | C2 | | | C3 | | | P1 | | | P2 | | | L1 | | | L2 | | | L3 | | | U | | | Avoidable total (C1 + C2 + C3): Product total (P1 + P2): Legitimate total (L1 + L2 + L3): Did any customer move from an avoidable question to a legitimate need inside one conversation? Describe: PART D: EVIDENCE FOR FRICTION AND DEFECTS Fill this in for P1 and P2 conversations. The transcript says what the customer said; this part records what the page or product did. Page or step where the problem happened: Device, browser or app version, if known: Steps to reproduce on a test account: Reproduced? (yes / no / not tried): Session recordings (only with a capture product such as FullSession Core CX / UX, and only for sessions that were captured): Conversations with a captured session (count): Conversations without a captured session (count): What the recordings showed: Funnel step with the largest drop for this task, split by device (if you use funnels): Errors or failed requests linked to these sessions (if you use error tracking): Sensitive inputs masked in what you reviewed? (yes / no / not checked): Open question still unanswered: PART E: FIX PLAN Content fix What changes (article, page wording, policy text): Owner (one named person): Due date: Added to the AI agent's approved sources? (yes / no / date): Answer re-checked after the change? (yes / no / date and result): Product fix What changes: Owner (one named person): Due date: Ticket or issue reference: Policy fix (if any) What changes: Owner (one named person): Due date: Interim answer until the fixes ship What customers are told (approved wording): Which conversations stay with a person until then: Known-issue note added to approved content? (yes / no / date): PART F: CONTACT-ROUTE CHECK Run this before and after the fixes. A fall in contacts does not count if any answer below is "no". Can a customer ask for a person at any point in the conversation? (yes / no): Is the request for a person honoured without the customer having to rephrase it? (yes / no): Does the person take over in the same thread, with the earlier answers visible? (yes / no): Is the contact route visible on the pages where this question comes up? (yes / no): Outside office hours, can a customer still leave the question? (yes / no): Was any contact link, form, email address or phone number removed or hidden during this period? (yes / no; if yes, what and why): Observed on (date, device, website or in-app): PART G: BASELINE VERSUS NEXT PERIOD Next period must have the same length as the baseline and start after the fixes ship. Use the same cluster definition and denominator. Next period (start date, end date): Fixes shipped on (content / product / policy): Other changes in either period (releases, promotions, seasonal swings, pricing changes): Measure | Baseline | Next period | Read it with Cluster rate (per 100 of denominator) | | | Total contact rate Repeat contacts within 7 days | | | Cluster rate Successful help (confirmed or verified) | | | Requests for a person after the AI agent answered Requests for a person, by reason | | | Avoidable versus legitimate split Downstream signals citing the same cause | | | Cluster rate Fix lead time in days, per owner | | | Interim answer used meanwhile Sample of next-period conversations read in full (count and what changed in the customers' words): PART H: DECISION Is the cause fixed? (yes / partly / no): Evidence for that answer: Next step (close the cluster / revise the content / escalate the product fix / extend the period): Next cluster to review: Reviewed by and date: NOTES FOR FULLSESSION SUPPORT & SALES DESK USERS These describe published behavior on https://www.fullsession.io/solutions/support-sales-desk, checked 19 October 2026. Record what you observe in your own setup. Max answers from approved content (documentation, help center, pricing page, marketing site, support portal) and from data your MCP server is permitted to expose. In your web app it answers from your docs or co-drives the fix while the customer watches; personal next steps are remembered across visits. Customers always know when they are talking to Max and can ask for a person at any time. Your team takes over in the same thread from one shared inbox, and Max's replies are labeled in the transcript, which makes Part B faster to fill in. The Desk app, shared inbox, human chat and supported integrations are included with Max, with no seat fees. Max is billed per AI-assisted conversation: see https://www.fullsession.io/pricing?solution=desk for current terms. Session recordings, heatmaps and funnels come with the separate Core CX / UX plan. Recordings depend on your Core plan, capture settings and retention, so Part D has a recording only for sessions that were captured.