CALL-ANALYSIS PILOT KIT Five fictional calls for an existing-phone-system comparison Created by Call Nerd, 11 September 2026 Guide: https://callnerd.xyz/blog/best-call-analysis-tools-existing-phone-system/ PURPOSE Use the same staged material to check a tool's import path, preservation of meaning, evidence access and usefulness. This is an original rehearsal kit, not a benchmark run against any named product or a claim of validated accuracy. HOW TO USE 1. Record each short script separately with a willing colleague. Follow the recording permissions that apply to your situation. Do not use real customer identities, phone numbers, account records or payment details. 2. Use the recording route you intend to keep: an existing provider export, a supported connection, or the actual controlled device capture workflow. A direct file upload does not test an integration or cellular capture. 3. Keep the original call labels and dates with each file. Use distinct test customer labels A-E. Only calls A and B share a topic; all five are separate fictional customers and cases. 4. Upload only the recorded conversations for an audio test. Keep the answer key below separate from the model input. If testing text-only analysis, supply the scripts alone and label the result as a text-only test. 5. Do not pad these short calls. Record their durations and whether summaries or other requested outputs are available at that duration on the tested plan. 6. Compare findings with the answer key. Then test additional authorised examples that represent your real languages, noise and recording conditions. CALL A - DELIVERY DATE NOT CONFIRMED Customer: Can you confirm delivery for Friday? Staff: Not yet. The warehouse has not confirmed the date. I will check and call you by two. Customer: Thank you. I need the date before I arrange time off. CALL B - ANOTHER CUSTOMER, RELATED QUESTION Customer: Has a delivery day been assigned to my order? Staff: I can see the order, but the delivery day is still blank. I will ask dispatch to check it and send you an update tomorrow morning. Customer: Please do. I can be available in the afternoon once I know the day. CALL C - DIAGNOSIS IS NOT THE WHOLE REPAIR Customer: Does the fifty-unit diagnostic fee pay for the entire repair? Staff: No. It covers the diagnostic visit. We will quote any parts and repair labour separately, and ask you before doing that work. Customer: Okay, please book the diagnostic visit. I am not approving a repair yet. CALL D - DO NOT CANCEL Customer: Please do not cancel my Tuesday appointment. I only want to change the service address. Is that possible? Staff: I need to check whether the new address is in our service area. I will check and call you back. I have not changed or cancelled the appointment. CALL E - REQUEST IS NOT APPROVAL Customer: Can the delivery team carry the item upstairs? Staff: I need the delivery manager to confirm that. I will ask, but I cannot promise an upstairs service or a price for it yet. Customer: Please check before you finalise the booking. OPTIONAL ANALYSIS REQUEST For each call, identify the customer's question, confirmed facts, unresolved items, promises and next actions. Cite the call label and supporting passage or timestamp. Do not turn a request into approval or fill in missing dates, prices or answers. Across these five calls, group related customer questions and count distinct calls supporting each group. Keep an individual unresolved case separate from a general topic. Say when the evidence is insufficient. ANSWER KEY - DO NOT INCLUDE IN THE TEST INPUT A: The customer wants a confirmed delivery date. Friday is a request, not a confirmed date. Staff promised to check and call by two; no day for that callback is explicitly stated in the script. Do not invent one. B: The customer wants an assigned delivery day. It is not available. Staff promised a dispatch check and an update tomorrow morning. Afternoon is the customer's availability, not a confirmed delivery slot. Without the call's actual date, do not invent a calendar date for tomorrow. C: The fifty-unit fee covers diagnosis only. Parts and repair labour require a separate quote and approval. The customer agreed to the diagnostic visit, not the repair. The script does not specify a currency, tax or refund policy. D: The customer explicitly wants to keep the Tuesday appointment and asks about changing the address. The change is unconfirmed pending a service-area check. The appointment has not been changed or cancelled. No exact callback deadline or new address is supplied. E: Upstairs delivery is requested, not approved. Both availability and price remain unknown. Staff must check with the delivery manager before the booking is finalised. No firm callback time is given. ACROSS-CALL CHECK A and B support one narrow theme: uncertainty about the delivery day (2 of 5 calls in this staged set). E is also about delivery but asks a different question requiring a different answer: upstairs service. C concerns fee scope; D concerns an address change while retaining an appointment. Other sensible group names are acceptable if their meaning, evidence and counts are correct. These are counts within fictional test material, not real-world prevalence. No call establishes that the underlying service issue was subsequently solved. RECORD YOUR OBSERVATIONS Tool and plan: Test date: Recording device/provider and input route: Call labels and actual durations: Settings/languages: Manual transfer and review time: For each call, mark usable / needs correction / missing / not tested: - Arrival: correct file, date and identity - Meaning: dates, amounts, negations, requests and promises - Evidence: supporting passage can be located - Across-call finding: grouping and distinct-call counts are reviewable - Action: what the next person can do, with uncertainty preserved Record exact errors and unavailable features. Do not collapse these five examples into a percentage claiming general accuracy. A file-upload test says nothing about whether a separate phone integration works. A staged English test says nothing by itself about noise, other languages or real call capture.