5D Library › Books › 5DL-B-084

对话式成交自动机、异议消解信息增益与高客单销售动力学

诊断式提问树、信任阈值停时规则与人机协同逼单协议

Conversational Closing Automata, Objection-Handling Information Gain and High-Ticket Sales Kinetics: Diagnostic Question Trees, Trust-Threshold Stopping Rules and AI Closer Escalation Protocols

  • 范雅福 Fan Yafu
  • Version 1.0.0 · 2026-11-09
  • CC-BY-NC-ND-4.0
  • 5DL-B-084
DOI10.5281/zenodo.24090221

摘要

一位潜在客户在私域询问 RM 9,800 或 RM 29,800 的高客单方案时,卖方(或替卖方说话的 AI)该先问什么?怎样回应「太贵了」?何时请求付款,何时交给人工?本书把这些经验法则写成可以运行的算术,再看它们在什么条件下成立。书中的先验、回答概率、成交率、退款率与成本均为示意或合成模型;全部数据由书中程序生成;本书不是法律意见,也不提供任何操纵买家的话术。

第一章把买家的预算与紧迫度写成部分可观测决策过程的隐藏状态,改正一个直觉:过早报价并非没有信息(0.229 比特),它的代价是被拒绝之后的流失。第二章用动态规划求解最优提问树:先问情境问题,平均只问 1.73 题,价值比直接报价高 9.0%;完全信息的上限是 RM 3,466。第三章用贝叶斯公式解码四大异议,按后验回应收回原价的 0.219,一套话术只有 0.163;「太贵了」先问澄清问题再回应,收回提高 9.9%。第四章发现高压话术多成交(27.0 单对 21.9 单),计入退款、拒付与差评之后每 100 位提出异议的买家净少 RM 113,605。

第五章用吉廷斯指数在广告来源之间分配人工时间(新来源的探索奖励 +0.102),并发现按「最热」交接只用上人工潜力的 55%。第六章求解信任阈值停时规则:信念低于 0.138 继续聊,0.138 至 0.500 交给人,0.500 以上由 AI 请求付款;并以爱尔朗 C 公式配置人工坐席。第七章与《鬼谷子》反应篇、内揵篇与《战国策·秦策一》并读,改正「因其言,听其辞」的引文(开放问题 0.509 比特)。第八章给出一万段对话的账本、概率校准的检查与一项预注册实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。

关键词:高客单销售;对话式商务;AI 销售智能体;部分可观测马尔可夫决策过程;信念状态;互信息;信息价值;SPIN 销售法;诊断式提问;提问树;动态规划;异议处理;贝叶斯解码;澄清问题;高压销售;退款;拒付;声誉;说服知识;吉廷斯指数;多臂老虎机;增量建模;人机协同;最优停时;爱尔朗 C 公式;客服坐席配置;概率校准;布里尔分数;鬼谷子;战国策;苏秦;预注册

Abstract

When a prospect asks in a private channel about a high-ticket offer at RM 9,800 or RM 29,800, what should the seller (or the AI speaking for the seller) ask first? How should "too expensive" be answered? When should payment be requested, and when should a human take over? This book writes these rules of thumb as runnable arithmetic and then asks when they hold. Priors, answer probabilities, close rates, refund rates and costs are illustrative or synthetic; all data are generated by its programs; the book is not legal advice and offers no scripts for manipulating buyers.

Chapter 1 writes the buyer's budget and urgency as hidden states of a partially observable decision process and corrects an intuition: an early pitch is not empty of information (0.229 bits); its cost is the buyers lost after a refusal. Chapter 2 solves for the optimal question tree by dynamic programming: ask the situation question first, only 1.73 questions on average, worth 9.0% more than pitching at once; the full-information bound is RM 3,466. Chapter 3 decodes the four classic objections with Bayes' rule: answering by the posterior recovers 0.219 of the price, one script only 0.163; asking a clarifying question after "too expensive" raises the recovery by 9.9%. Chapter 4 finds that high pressure closes more (27.0 sales against 21.9) but, after refunds, chargebacks and bad reviews, nets RM 113,605 less per 100 objectors.

Chapter 5 allocates human time among sources of leads with Gittins indices (an exploration bonus of +0.102 for a new source) and finds that handing over the "hottest" conversations uses only 55% of the humans' potential. Chapter 6 solves for a trust-threshold stopping rule: keep talking below a belief of 0.138, hand over between 0.138 and 0.500, let the AI ask for payment above 0.500; and staffs the human desk with the Erlang C formula. Chapter 7 reads the Guiguzi's "Fan Ying" and "Nei Jian" and the Zhanguo Ce's "Qin I", correcting the quotation to "follow his words, listen to his phrasing" (an open question, 0.509 bits). Chapter 8 gives a ledger of ten thousand conversations, a calibration check and a pre-registered experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses.

Keywords: high-ticket sales; conversational commerce; AI sales agents; POMDP; belief state; mutual information; value of information; SPIN selling; diagnostic questions; question trees; dynamic programming; objection handling; Bayesian decoding; clarifying questions; high-pressure selling; refunds; chargebacks; reputation; persuasion knowledge; Gittins index; multi-armed bandits; uplift modelling; human-in-the-loop; optimal stopping; Erlang C; call centre staffing; calibration; Brier score; Guiguzi; Zhanguo Ce; Su Qin; pre-registration

目录 · Contents

  • 序:先听,再报价English 中文Preface: Listen before You Quote
  • 第一章 买家隐藏的状态与每个问题的信息量English 中文Chapter 1. What the Buyer Hides, and How Much Each Question Tells
  • 第二章 诊断式提问树:问哪一题、问几题English 中文Chapter 2. Diagnostic Question Trees: Which Question, and How Many
  • 第三章 四大异议的贝叶斯解码English 中文Chapter 3. Decoding the Four Classic Objections with Bayes
  • 第四章 高压逼单的账:退款、拒付与声誉English 中文Chapter 4. The Books of High-Pressure Closing: Refunds, Chargebacks and Reputation
  • 第五章 AI 销售智能体何时交给人:吉廷斯指数与增量排序English 中文Chapter 5. When an AI Sales Agent Hands Over to a Human: Gittins Indices and Ranking by Uplift
  • 第六章 信任阈值停时规则与人工坐席规划English 中文Chapter 6. Trust-Threshold Stopping Rules and Planning the Human Desk
  • 第七章 《鬼谷子》反应与内揵、《战国策》苏秦说秦English 中文Chapter 7. The Guiguzi's "Fan Ying" and "Nei Jian", and Su Qin's Persuasion of Qin in the Zhanguo Ce
  • 第八章 一万段对话的账本、概率校准与预注册实验English 中文Chapter 8. A Ledger of Ten Thousand Conversations, Probability Calibration and a Pre-Registered Experiment

How to cite · 引用

Fan, Y. (2026). Conversational Closing Automata, Objection-Handling Information Gain and High-Ticket Sales Kinetics: Diagnostic Question Trees, Trust-Threshold Stopping Rules and AI Closer Escalation Protocols [对话式成交自动机、异议消解信息增益与高客单销售动力学] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090221

BibTeX
@book{fan2026conversational,
  title = {{Conversational Closing Automata, Objection-Handling Information Gain and High-Ticket Sales Kinetics}: Diagnostic Question Trees, Trust-Threshold Stopping Rules and AI Closer Escalation Protocols},
  author = {Fan, Yafu},
  year = {2026},
  publisher = {5D Library},
  edition = {Version 1.0.0},
  doi = {10.5281/zenodo.24090221},
  url = {https://doi.org/10.5281/zenodo.24090221},
  note = {Chinese title: 对话式成交自动机、异议消解信息增益与高客单销售动力学}
}
DOIhttps://doi.org/10.5281/zenodo.24090221
DOI(全部版本 all versions)https://doi.org/10.5281/zenodo.24090220
ORCIDhttps://orcid.org/0009-0004-3519-1203
5D Library ID5DL-B-084
版本 Version1.0.0 · 中英双语版 Bilingual edition
出版日期 Date2026-11-09
网址 Onlinehttps://5dlibrary.com/books/conversational-closing-automata-objection-information-gain-high-ticket-sales/
SHA-256 · conversational-closing-automata-objection-information-gain-high-ticket-sales_bilingual_v1.0.0.pdf3361ffcede82d9b07e8cfb744cca957acd87e0619418fcc8cbf11ac6bc0fbbab
SHA-256 · conversational-closing-automata-objection-information-gain-high-ticket-sales_en_v1.0.0.pdf7d25b0d8005739c9e00cf71e3d3ee4772a30a5700dce61130b66415357ea086b
SHA-256 · conversational-closing-automata-objection-information-gain-high-ticket-sales_zh_v1.0.0.pdf7e0e7fccec465031dc8b0547d6e83a5793eb86c9fd2c181434ecd9cf47e11380
Licence · 许可Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0). https://creativecommons.org/licenses/by-nc-nd/4.0/

Disclosure. The author uses the works of 5D Library, including this book, as the methodological basis of his community courses, the 5D Ascension Cards (Dimensional Shift Card Decks) and the Yafu AI knowledge base, and runs a number of websites, paid online reading software, subscriptions, reports, courses, business licences, advisory services and AI conversational sales systems whose private-channel conversations, objection handling and human hand-overs are exactly what this book studies, so he has a direct commercial interest in its subject. The book promotes no software, course, card deck or service, uses no user or conversation data from the author's businesses, and accepts no sponsorship. Its parameters are synthetic or illustrative. The experiment in Chapter 8 must follow a pre-registered protocol, and its results will be published whether positive or negative.