Pitches try a different pitch on the calls you already have, before adopting it

What this can and cannot tell you. The model's per-step judgements are already stored, so a different set of weights and thresholds can be applied to the same calls for free — no re-transcription, no tokens. What it cannot tell you is how a rep would have behaved under a different pitch, or how the model would judge a step the live pitch never asked about: a variant that adds a new step has no evidence behind it, and that is reported as unjudged rather than counted as a failure.

Nothing here changes live scoring until you press Adopt.

Compare

No proposed pitch yet. Paste one below — start from the live rubric and change what you want to test.

Add a proposed pitch

Same shape as the live rubric, and checked the same way: valid JSON, numeric weights, unique step ids, and weights totalling 10 per pitch type — because a score is the sum of the weights earned and is reported out of 10, never scaled.

Name this pitch

{
  "meta": {
    "name": "JSW One x Ozonetel Inside-Sales Call Analysis Rubric",
    "version": "1.0.0",
    "source_brd": "JSW One IS Call Quality BRD (Modules 1-3)",
    "scoring_philosophy": "moderate",
    "notes": "A step is PRESENT if its intent is executed, not if the exact scripted phrase is used. Ordering and phrasing are flexible; core steps must be present.",
    "band_thresholds": {
      "green_min": 8,
      "amber_min": 5,
      "red_min": 0
    },
    "partial_credit_fraction": 0.5
  },
  "pitch_type_detection": {
    "priority_order": [
      {
        "type": "customer_led",
        "rule": "opportunity_id is present / non-empty"
      },
      {
        "type": "new_registration",
        "rule": "registration_date within 30 days of call_datetime AND no prior opportunities"
      },
      {
        "type": "supply_led",
        "rule": "campaign_type == supply_led OR campaign flagged as supply-led (business-named campaigns)"
      },
      {
        "type": "baseline",
        "rule": "redial case OR dormant customer (redial_flag true / dormant_flag true) and none of the above"
      },
      {
        "type": "undetermined",
        "rule": "none of the above can be confirmed -> EXCLUDE from pitch scoring"
      }
    ],
    "excluded_from_scoring": [
      "undetermined"
    ]
  },
  "pitch_rubrics": {
    "supply_led": {
      "max_score": 10,
      "steps": [
        {
          "id": "SL1",
          "weight": 2,
          "name": "Opener & permission",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Agent greets, identifies as JSW One, uses customer name, asks if it's a good time. If declined, does not push; asks for callback time."
        },
        {
          "id": "SL2",
          "weight": 2,
          "name": "Hook (supply/price signal)",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Leads with ONE benefit (price OR availability, not both). Hook under ~30s. No immediate product dump."
        },
        {
          "id": "SL3",
          "weight": 2,
          "name": "Discovery (6 Qs)",
          "mandatory": true,
          "output": "full_partial_fail",
          "threshold_min": 3,
          "threshold_full": 5,
          "of_total": 6,
          "checks": "Q1 what customer manufactures, Q2 end applications, Q3 monthly MT, Q4 current brands/suppliers, Q5 JSW% of procurement, Q6 payment terms."
        },
        {
          "id": "SL4",
          "weight": 1,
          "name": "Pitch (2-3 USPs)",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "USPs selected relevant to what was discovered. Did NOT pitch all USPs indiscriminately."
        },
        {
          "id": "SL5",
          "weight": 2,
          "name": "Qualify",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Asked if current requirement exists. If yes: captured grade, thickness, width, qty, location, date. Confirmed stock on the spot if possible."
        },
        {
          "id": "SL6",
          "weight": 1,
          "name": "Closure with next step",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Explicit committed next step: quote with agreed turnaround OR specific callback date logged."
        }
      ]
    },
    "customer_led": {
      "max_score": 10,
      "steps": [
        {
          "id": "CL1",
          "weight": 2,
          "name": "Opener acknowledges platform submission",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Acknowledges customer raised a requirement on platform. Does NOT restart from scratch. References what customer did."
        },
        {
          "id": "CL2",
          "weight": 3,
          "name": "Revalidate every field",
          "mandatory": true,
          "output": "full_partial_fail",
          "threshold_min": 3,
          "threshold_full": 5,
          "of_total": 5,
          "checks": "Reads back and confirms: product/grade, thickness x width, quantity (MT), delivery location/pincode, timeline. Changed fields updated before pricing."
        },
        {
          "id": "CL3",
          "weight": 2,
          "name": "Discovery (5 Qs, woven naturally)",
          "mandatory": true,
          "output": "full_partial_fail",
          "threshold_min": 2,
          "threshold_full": 4,
          "of_total": 5,
          "checks": "Q1 what customer manufactures/end application, Q2 monthly MT, Q3 current brands, Q4 JSW% of procurement, Q5 payment terms. Conversational, not interrogative."
        },
        {
          "id": "CL4",
          "weight": 2,
          "name": "Closure with price & timeline",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Committed to specific delivered price OR confirmed stock and pushed for immediate closure. Timeline for price revert stated. If stock confirmed, attempted immediate close."
        },
        {
          "id": "CL5",
          "weight": 1,
          "name": "No unnecessary re-pitching",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Did NOT re-pitch entire portfolio. Acknowledged customer already chose platform. Reinforced only 1-2 relevant advantages."
        }
      ]
    },
    "new_registration": {
      "max_score": 10,
      "steps": [
        {
          "id": "NR1",
          "weight": 1,
          "name": "Warm welcome opener",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Warm, welcomes new registration, positions as helper not pusher. No hard sales pressure in opener."
        },
        {
          "id": "NR2",
          "weight": 1,
          "name": "Verify basic profile (4 Qs)",
          "mandatory": true,
          "output": "full_partial_fail",
          "threshold_min": 2,
          "threshold_full": 4,
          "of_total": 4,
          "checks": "Q1 name/company, Q2 procurement city, Q3 involvement in procurement decisions, Q4 best contact number/email for quotes."
        },
        {
          "id": "NR3",
          "weight": 3,
          "name": "Full discovery (8 Qs)",
          "mandatory": true,
          "output": "full_partial_fail",
          "threshold_min": 4,
          "threshold_full": 6,
          "of_total": 8,
          "checks": "Q1-2 business type/end applications, Q3-4 monthly MT/brands, Q5 payment terms, Q6 current JSW buying, Q7 JSW% of procurement, Q8 what stopped direct JSW buying earlier."
        },
        {
          "id": "NR4",
          "weight": 1,
          "name": "Platform education",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Explained platform simply: transparent delivered pricing, Test Certificate, stockyard availability. Tailored to discovery."
        },
        {
          "id": "NR5",
          "weight": 2,
          "name": "First opportunity identified",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Asked if current material requirement. If yes: full spec captured. If no: specific callback date scheduled (not 'sometime')."
        },
        {
          "id": "NR6",
          "weight": 2,
          "name": "Closure - one of three outcomes",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Ended with exactly one of: (1) quote sent + follow-up time agreed, (2) specific callback date confirmed in CRM, (3) platform walkthrough scheduled with time agreed."
        }
      ]
    },
    "baseline": {
      "max_score": 10,
      "inferred": true,
      "inferred_note": "BRD names a Baseline pitch for redial/dormant cases but does not define its rubric. Derived here from the shared core of the three defined pitches. FLAGGED as an assumption for business sign-off.",
      "steps": [
        {
          "id": "BL1",
          "weight": 2,
          "name": "Opener & permission",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Greets, identifies as JSW One, uses customer name, asks if it's a good time."
        },
        {
          "id": "BL2",
          "weight": 2,
          "name": "Re-establish context",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "References prior interaction / dormant relationship; does not cold-restart as if never contacted."
        },
        {
          "id": "BL3",
          "weight": 3,
          "name": "Discovery / requirement refresh",
          "mandatory": true,
          "output": "full_partial_fail",
          "threshold_min": 2,
          "threshold_full": 4,
          "of_total": 5,
          "checks": "Refreshes: what customer manufactures, monthly MT, current brands, JSW% of procurement, payment terms."
        },
        {
          "id": "BL4",
          "weight": 1,
          "name": "Relevant value reinforcement",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Reinforced 1-2 relevant advantages; did not dump full portfolio."
        },
        {
          "id": "BL5",
          "weight": 2,
          "name": "Closure with next step",
          "mandatory": true,
          "output": "pass_fail",
          "checks": "Explicit committed next step: quote with turnaround OR specific callback date."
        }
      ]
    }
  },
  "score_bands": [
    {
      "band": "green",
      "label": "Strong",
      "min": 8,
      "max": 10,
      "action": "No action. Positive flag for scorecard."
    },
    {
      "band": "amber",
      "label": "Partial",
      "min": 5,
      "max": 7,
      "action": "Flag for coaching. Supervisor review within 1 week."
    },
    {
      "band": "red",
      "label": "Low",
      "min": 0,
      "max": 4,
      "action": "Urgent coaching. Supervisor listens within 48 hrs."
    }
  ],
  "customer_quality_flags": {
    "note": "Independent flags. 0-4 can coexist. Classification, not a score.",
    "dimensions": [
      {
        "id": "segment",
        "flag_field": "mfg_flag",
        "detail_field": "segment_detected",
        "values": [
          "manufacturing",
          "construction",
          "trader",
          "cement",
          "wrong_product",
          "unknown"
        ],
        "flag_true_when": "segment_detected != manufacturing",
        "checks": "Manufacturer -> MFG (target). Construction co/contractor -> construction. Cement dealer/trader/retailer -> trader/cement. Stainless/alloy/pipes only -> wrong_product. Unclear -> unknown."
      },
      {
        "id": "geo",
        "flag_field": "geo_flag",
        "detail_field": "geo_reason",
        "values": [
          "language_barrier",
          "policy_excluded",
          "unknown_area",
          null
        ],
        "flag_true_when": "agent states area not serviceable OR customer in known out-of-coverage state/city OR language breakdown",
        "checks": "Non-serviceable area, out-of-coverage geography (e.g. J&K, Kolkata for certain products), or language switch with communication breakdown."
      },
      {
        "id": "moq",
        "flag_field": "moq_flag",
        "detail_field": "moq_qty_detected",
        "severity_field": "moq_severity",
        "severity_values": [
          "standard",
          "critical",
          "none"
        ],
        "flag_true_when": "quantity <= 5 MT OR agent mentions MOQ constraint OR sub_lost_reason == 'A- Lower than MOQ'",
        "severity_rule": "critical if < 3 MT, else standard if <= 5 MT, else none",
        "checks": "<=5 MT flagged; <3 MT critical."
      },
      {
        "id": "material",
        "flag_field": "material_flag",
        "detail_field": "material_detected",
        "status_field": "catalogue_status",
        "status_values": [
          "available",
          "not_available",
          "unknown"
        ],
        "flag_true_when": "material outside catalogue or not available",
        "checks": "Stainless steel, alloy steel, pipes, cement (where unavailable), or a spec agent confirms cannot be sourced."
      }
    ]
  },
  "supply_signal": {
    "field": "supply_flag",
    "note": "Module 3 = OUR ability to serve. Tracked separately from Module 2 (customer quality). Single categorical value per call.",
    "values": [
      {
        "value": "out_of_stock",
        "checks": "Agent states material not available, or customer told to wait for replenishment."
      },
      {
        "value": "lead_time_issue",
        "checks": "Customer needs material within a timeline agent cannot meet from stock/standard mill lead times.",
        "extra_fields": [
          "customer_req_date",
          "agent_committed_date"
        ]
      },
      {
        "value": "not_in_stockyard",
        "checks": "Material in catalogue but not in JSW One stockyard; longer mill lead time."
      },
      {
        "value": "spec_not_available",
        "checks": "Exact spec (thickness, width, grade) not carried; customer spec cannot be met.",
        "extra_fields": [
          "spec_details"
        ]
      },
      {
        "value": "none",
        "checks": "No supply constraint mentioned."
      }
    ]
  },
  "output_schema": {
    "note": "Per-call analysis object Claude returns (BRD 8.1). Deterministic scorer emits the same shape.",
    "fields": [
      "call_id",
      "agent_id",
      "pitch_type_detected",
      "pitch_adherence_score",
      "pitch_step_results",
      "pitch_missed_steps",
      "pitch_score_band",
      "pitch_coaching_note",
      "segment_detected",
      "mfg_flag",
      "geo_flag",
      "geo_reason",
      "moq_flag",
      "moq_qty_detected",
      "moq_severity",
      "material_flag",
      "material_detected",
      "catalogue_status",
      "supply_flag",
      "supply_details",
      "quality_flags_total",
      "call_summary"
    ]
  },
  "supplementary_metrics": {
    "note": "Agent-performance metrics beyond the 10-point rubric. Measurements (not part of the score), derived deterministically from turn-level timings + content. Added per business request.",
    "conversation_dynamics": [
      "talk_ratio",
      "agent_talk_sec",
      "customer_talk_sec",
      "dead_air_sec",
      "longest_monologue_sec",
      "agent_wpm",
      "interruptions"
    ],
    "sales_effectiveness": [
      "opener_compliant",
      "discovery_qs",
      "qualify_reached",
      "next_step_committed",
      "objection_raised",
      "objection_handled"
    ],
    "fairness_segmentation": [
      "adherence_by_persona",
      "adherence_by_language",
      "adherence_by_segment"
    ],
    "data_quality": [
      "undetermined_rate",
      "very_short_call"
    ],
    "monologue_heavy_rule": "talk_ratio > 0.72 OR longest_monologue_sec > 55 (upper-tail flag, not a per-call scored step)"
  },
  "objection_handling": {
    "note": "From the revised IS pitches (07.07) objection playbook. Detected as a supplementary metric, NOT a scored rubric step (keeps the 10-point weighting BRD-locked). An objection raised but not handled is a coaching flag.",
    "source": "Revised Pitches IS 07.07 — per-pitch OBJECTIONS tables",
    "examples": [
      "price higher -> compare landed cost",
      "already have supplier -> trial order on running size",
      "need credit -> structured credit after two transactions",
      "remote site -> logistics partners"
    ]
  },
  "provenance": {
    "pitch_scripts": "Revised Pitches IS 07.07 (Supply-Led / Customer-Led / New-Registration). Rubric structure confirmed unchanged; script language adopted for detection markers.",
    "personas": "Customer Personas for JSW One's B2B Buyers in India — 5 personas (Fabrication MSME, Light Engineering SMB, Infra Contractor, Real-Estate Builder, Retailer/Stockist) drive segment, product, MT range, language, geography and objections in the synthetic data."
  },
  "acceptance_criteria": {
    "note": "BRD 10 - used by verify_sample.py to compute agreement vs ground truth / live-Claude.",
    "targets": [
      {
        "id": "AC1",
        "metric": "pitch_type_accuracy",
        "target": 0.9
      },
      {
        "id": "AC2",
        "metric": "pitch_score_agreement_within_points",
        "target": 1.5
      },
      {
        "id": "AC3",
        "metric": "customer_quality_flag_accuracy",
        "target": 0.85
      },
      {
        "id": "AC4",
        "metric": "supply_flag_accuracy",
        "target": 0.85
      },
      {
        "id": "AC5",
        "metric": "vernacular_transcription_quality",
        "target": 0.8
      },
      {
        "id": "AC7",
        "metric": "call_summary_usefulness",
        "target": 0.8
      }
    ]
  }
}

Saved beside the live rubric, not over it. Compare first, adopt only if you want it.