5D Library › 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 · English

第三章 四大异议的贝叶斯解码

同一句话,不同的原因

「太贵了」「我再考虑一下」「要问问伴侣或合伙人」「怕没时间用」,这四句话是高客单销售中最常听到的异议。每一句背后都可能是不同的原因:真实的预算限制、信任还没有建立、价值还没有讲清楚、真的没有时间,或者确实要与他人共同决定。把原因当作隐藏状态,把异议当作观测,就可以用贝叶斯公式解码。

在示意模型中,提出异议的买家里,五种原因的先验分别是 0.25、0.25、0.25、0.10 与 0.15。每种回应的成交概率取决于真实原因:重新阐明价值对「价值未讲清」最有效;分期或降档对「预算限制」最有效,但只收回原价的 0.6;案例与保证对「信任不足」最有效;约三方通话对「共同决定」最有效;实施计划对「没有时间」最有效。

推导

# Chapter 3. p05: Decoding the four classic objections. Among buyers who
# object, the hidden cause is: a real budget limit, trust not yet built,
# value not yet clear, a real lack of time, or a genuine joint decision
# (prior 0.25 / 0.25 / 0.25 / 0.10 / 0.15). Each cause makes each
# objection more or less likely; Bayes gives the cause behind an
# objection. Each response closes with a probability that depends on the
# TRUE cause; a payment plan or downsell brings in 0.6 of the price.
import math
CAUSES = ("budget", "trust", "value", "time", "partner")
PRIOR = (0.25, 0.25, 0.25, 0.10, 0.15)
OBJ = ("too expensive", "think it over", "ask my partner", "no time")
LIKE = ((0.70, 0.15, 0.10, 0.05), (0.20, 0.45, 0.25, 0.10),
        (0.45, 0.30, 0.10, 0.15), (0.05, 0.20, 0.05, 0.70),
        (0.10, 0.20, 0.65, 0.05))
RESP = {  # close probability by true cause, revenue factor
    "reframe value": ((0.05, 0.15, 0.35, 0.10, 0.10), 1.0),
    "plan/downsell": ((0.30, 0.08, 0.12, 0.08, 0.10), 0.6),
    "proof+guarantee": ((0.05, 0.40, 0.20, 0.10, 0.15), 1.0),
    "joint call": ((0.05, 0.15, 0.10, 0.05, 0.45), 1.0),
    "time plan": ((0.03, 0.10, 0.12, 0.40, 0.08), 1.0),
}

def posterior(o, prior=PRIOR):
    w = [prior[c] * LIKE[c][o] for c in range(5)]
    z = math.fsum(w)
    return [x / z for x in w], z

def score(post, r):
    p, f = RESP[r]
    close = math.fsum(post[c] * p[c] for c in range(5))
    return close, close * f

if __name__ == "__main__":
    print("   objection       P(obj)  most likely cause   best response"
          "    value")
    total_best = total_one = 0.0
    for o, name in enumerate(OBJ):
        post, pz = posterior(o)
        c = max(range(5), key=lambda i: post[i])
        r = max(RESP, key=lambda x: score(post, x)[1])
        v = score(post, r)[1]
        total_best += pz * v
        total_one += pz * score(post, "reframe value")[1]
        print("   {:<14}  {:6.3f}  {:<7} {:5.3f}       {:<15}  {:.3f}".format(
            name, pz, CAUSES[c], post[c], r, v))
    print("   value per objector (share of full price):")
    print("     matched response {:.3f}; one script (reframe) {:.3f}".format(
        total_best, total_one))
#    objection       P(obj)  most likely cause   best response    value
#    too expensive    0.357  budget  0.490       reframe value    0.161
#    think it over    0.275  trust   0.409       proof+guarantee  0.249
#    ask my partner   0.215  partner 0.453       joint call       0.266
#    no time          0.152  time    0.459       time plan        0.236
#    value per objector (share of full price):
#      matched response 0.219; one script (reframe) 0.163

「太贵了」占异议的 0.357,其背后最可能的原因是真实预算(后验 0.490),但最好的回应仍是重新阐明价值(期望收回原价的 0.161),因为降档只收回 0.6 的价格。「我再考虑一下」最可能是信任不足(0.409),最好的回应是案例与保证(0.249)。「要问伴侣」最可能是共同决定(0.453),最好的回应是约一次三方通话(0.266)。「没有时间」最可能确实没时间(0.459),最好的回应是实施计划(0.236)。针对异议选择回应,每位提出异议的买家平均收回原价的 0.219;对所有异议都用同一套「重新阐明价值」的话术,只有 0.163。

四种异议背后各种原因的后验概率(p05)。

「太贵了」要先问清楚

「太贵了」是最含糊的一句:一半是真的付不起,其余多半是价值或信任的问题。与其猜,不如先问一个澄清问题:「如果价格在您的预算之内,这是您要的方案吗?」真实预算受限的买家多半回答「是」(0.85),价值没讲清楚的买家较少(0.25),信任不足的居中(0.40)。

推导

# Chapter 3. p06: "Too expensive" is ambiguous: half the time it is a real
# budget limit, otherwise mostly unclear value or missing trust. One
# clarifying question first: "If the price fitted your budget, is this
# the right solution for you?" A yes is likely from a real budget limit
# (0.85), less likely when value is unclear (0.25) or trust is missing
# (0.40); time 0.5, partner 0.5. Asking loses 5% of objectors.
import importlib.util

def load(name):
    spec = importlib.util.spec_from_file_location(name, name + ".py")
    m = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(m)
    return m

p05 = load("p05_objections")
YES = (0.85, 0.40, 0.25, 0.5, 0.5)
LOSS = 0.05

def best(post):
    r = max(p05.RESP, key=lambda x: p05.score(post, x)[1])
    return r, p05.score(post, r)[1]

if __name__ == "__main__":
    post, _ = p05.posterior(0)
    r0, v0 = best(post)
    print("   no question: {} -> {:.3f} of full price".format(r0, v0))
    v1 = 0.0
    for ans in (True, False):
        w = [post[c] * (YES[c] if ans else 1 - YES[c]) for c in range(5)]
        z = sum(w)
        q = [x / z for x in w]
        r, v = best(q)
        v1 += z * v
        print("   answer {:<3} (p {:.3f}): budget {:.3f}, value {:.3f}"
              " -> {} {:.3f}".format("yes" if ans else "no", z, q[0], q[2],
                                     r, v))
    v1 *= 1 - LOSS
    print("   with the question: {:.3f} ({:+.1%})".format(v1, v1 / v0 - 1))
#    no question: reframe value -> 0.161 of full price
#    answer yes (p 0.579): budget 0.719, value 0.136 -> plan/downsell 0.147
#    answer no  (p 0.421): budget 0.174, value 0.560 -> reframe value 0.241
#    with the question: 0.177 (+9.9%)

不问就回应,期望收回原价的 0.161。先问:有 0.579 的买家回答「是」,预算受限的后验升到 0.719,这时最好的回应是分期或降档(0.147);有 0.421 回答「否」,价值未讲清的后验升到 0.560,最好的回应是重新阐明价值(0.241)。即使问题本身让 5% 的买家离开,期望收回仍升到 0.177(+9.9%)。

注

真实预算限制要被尊重。 当澄清问题显示买家确实付不起,正确的回应是分期、降档或请他以后再来,而不是继续施压。下一章计算施压的代价。

小结

四大异议都不是一个原因的同义词。按贝叶斯后验选择回应,比一套话术用到底多收回三成以上(0.219 对 0.163)。「太贵了」最含糊,先问一个澄清问题再回应,收回提高 9.9%。

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 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).