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
第四章 高压逼单的账:退款、拒付与声誉
高压话术多成交了几单
「死缠烂打」的高压话术(假截止时间、一再催促、不给考虑的余地)确实能让当下的成交更多。问题是这些成交之后发生什么。买家一旦意识到自己被施压,就会把整段对话重新解读为操纵(说服知识模型)(Friestad and Wright 1994),后悔、退款、信用卡拒付与差评随之而来。
推导
# Chapter 4. p07: High pressure against matched answers, per 100 buyers who
# object to a RM 29,800 offer. High pressure (deadline, repeated push)
# closes by true cause 0.25 / 0.30 / 0.30 / 0.20 / 0.25; matched answers
# are those of p05. Of buyers who were pushed, 45% regret the purchase;
# of buyers whose objection was answered, 8%. Of regretful buyers 60% get
# a refund, 15% file a card chargeback (amount lost plus a RM 60 fee) and
# 50% post a negative review, each costing RM 4,000 in lost future leads.
# Delivering costs 20% of the price, half of it already spent on refunds.
import math
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")
PRICE, N = 29800, 100
HP = (0.25, 0.30, 0.30, 0.20, 0.25)
def matched():
close = rev = 0.0
for o in range(4):
post, pz = p05.posterior(o)
r = max(p05.RESP, key=lambda x: p05.score(post, x)[1])
c, v = p05.score(post, r)
close += pz * c
rev += pz * v * PRICE
return close, rev
def ledger(close, rev, regret):
sales = N * close
gross = N * rev
sad = sales * regret
avg = gross / sales
refund = 0.60 * sad * avg
charge = 0.15 * sad * (avg + 60)
review = 0.50 * sad * 4000
deliver = 0.20 * (gross - 0.60 * sad * avg * 0.5)
net = gross - refund - charge - review - deliver
return sales, gross, refund + charge, review, net
if __name__ == "__main__":
hp_close = math.fsum(p05.PRIOR[c] * HP[c] for c in range(5))
rows = (("matched answers", *matched(), 0.08),
("high pressure", hp_close, hp_close * PRICE, 0.45))
nets = []
print(" per 100 objectors sales gross RM refunds+CB reviews"
" net RM")
for name, c, r, reg in rows:
s, g, rc, rv, net = ledger(c, r, reg)
print(" {:<17} {:5.1f} {:9,.0f} {:10,.0f} {:7,.0f} {:8,.0f}"
.format(name, s, g, rc, rv, net))
nets.append(net)
print(" switching to high pressure: {:+,.0f} RM per 100 objectors"
.format(nets[1] - nets[0]))
# per 100 objectors sales gross RM refunds+CB reviews net RM
# matched answers 21.9 653,291 39,213 3,508 483,047
# high pressure 27.0 804,600 271,662 24,300 369,442
# switching to high pressure: -113,605 RM per 100 objectors在示意模型中,每 100 位对 RM 29,800 方案提出异议的买家:按第三章选择回应,成交 21.9 单,毛收入 RM 653,291;高压话术成交 27.0 单,毛收入 RM 804,600,确实多了。但被施压的买家有 45% 后悔,被认真回答的只有 8%。扣除退款与拒付(RM 271,662 对 RM 39,213)、差评对未来获客的损失(RM 24,300 对 RM 3,508)以及交付成本之后,净收入是 RM 369,442 对 RM 483,047。改用高压话术,每 100 位提出异议的买家净损失 RM 113,605。
注
这也是法律问题。 马来西亚《1999 年消费者保护法》第二部分禁止误导与欺骗行为、虚假陈述与不公平做法 (Parliament of Malaysia 1999)。用不存在的截止日期或名额制造紧迫感,除了上述的财务代价,还有法律风险。本书不是法律意见。
声誉会累积
差评不会只影响当月。下面把差评写成一个会逐月衰减的存量:每条差评使下个月的询盘与成交率都打折,存量每月衰减一成。每个做法模拟 400 段 24 个月的历史。
推导
# Chapter 4. p08: Reputation over 24 months. Each month 120 qualified
# leads arrive; 40% of them object. Close rates, regret and refunds as in
# p07 (non-objectors close at 0.30 either way). Every negative review
# lowers next month's lead flow and close rates by the factor
# exp(-0.03 * R), where R is the stock of negative reviews, which fades by
# 10% a month. 400 simulated histories per playbook, seed 84; money in
# RM, net of refunds, chargebacks and delivery, no discounting.
import importlib.util
import random
from math import exp
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
p07 = load("p07_pressure")
P = p07.PRICE
def run(rng, close_obj, rev_factor, regret, months=24):
stock, total, path = 0.0, 0.0, []
for _ in range(months):
f = exp(-0.03 * stock)
leads = sum(1 for _ in range(120) if rng.random() < f)
neg = 0
for _ in range(leads):
objects = rng.random() < 0.40
pc, fac = (close_obj, rev_factor) if objects else (0.30, 1.0)
if rng.random() >= pc * f:
continue
amt = P * fac
total += 0.8 * amt
if objects and rng.random() < regret:
u = rng.random()
if u < 0.60:
total -= amt * 0.9
elif u < 0.75:
total -= amt * 0.8 + 60
if rng.random() < 0.50:
neg += 1
elif not objects and rng.random() < 0.08:
neg += rng.random() < 0.50
stock = 0.9 * stock + neg
path.append(total)
return path
if __name__ == "__main__":
rng = random.Random(84)
mc, mr = p07.matched()
hc = sum(p07.p05.PRIOR[c] * p07.HP[c] for c in range(5))
books = {"matched": (mc, mr / (mc * P), 0.08),
"pressure": (hc, 1.0, 0.45)}
runs = {k: [run(rng, *v) for _ in range(400)] for k, v in books.items()}
for m in (3, 12, 24):
a = sum(r[m - 1] for r in runs["matched"]) / 400
b = sum(r[m - 1] for r in runs["pressure"]) / 400
print(" month {:2d}: matched RM {:>10,.0f} pressure RM {:>10,.0f}"
.format(m, a, b))
ahead = sum(1 for a, b in zip(runs["matched"], runs["pressure"])
if b[-1] > a[-1])
print(" histories where pressure is ahead at month 24: {}/400"
.format(ahead))
# month 3: matched RM 2,103,137 pressure RM 1,765,073
# month 12: matched RM 7,001,337 pressure RM 4,923,635
# month 24: matched RM 12,739,837 pressure RM 8,441,151
# histories where pressure is ahead at month 24: 0/400第 3 个月时,按异议回应的累计净收入是 RM 2,103,137,高压是 RM 1,765,073;第 12 个月是 RM 7,001,337 对 RM 4,923,635;第 24 个月是 RM 12,739,837 对 RM 8,441,151。400 段历史中,高压话术在第 24 个月领先的次数是 0。
假说 · 尚未验证
作者的假说(尚未验证): 在高客单私域销售中,禁用虚假截止日期与反复催促的团队,十二个月后的退款与拒付合计占收入的比例,比使用这些话术的团队低至少一半。什么会证明它错: 在按销售顾问随机分配话术规范的试行中,两组的退款与拒付比例相差不到一半。
小结
高压话术能多成交几单,但在示意模型中,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;把声誉的累积算进去,差距逐月扩大,24 个月后高压话术在所有模拟历史中都落后。
参考文献
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).