What does Loyalty-Point Liabilities and Breakage conclude, in numbers, about “Estimating Expected Breakage: Matured Cohorts, Point Weighting and the Reversal Constraint”?

Answer: Chapter 3 gives Wilson intervals on matured cohorts, the gap between breakage by points (0.169) and by members (0.422), and a reversal constraint met by booking a posterior percentile.

Scope of the numbers: Parameters are illustrative and all calculations are generated by the book's programs; this is not individualised accounting, tax or legal advice, and a practising accountant's opinion prevails. Chapter 1 reworks the allocation and the cumulative catch-up with the numbers of the standard's illustrative example.

中文版

Data table

The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 0.169, 0.422

quintilemean.balshare.ptsp.redeembreakage.pts
142.20.0190.2070.761
2108.20.0500.4260.561
3206.40.0940.6020.390
4398.10.1820.7540.237
51429.00.6540.8990.071
breakage by members 0.422by points 0.169
liability error if member rate is used: -0.305
Show the full program
# Chapter 3. p32: breakage counted by members is not breakage by points.
# Synthetic members (illustrative): 5000 balances, lognormal with
# median 200 points and log-sd 1.2; a member's chance of ever redeeming
# rises with the balance, p = 1 / (1 + exp(-(0.4 + 1.1 * ln(b / 200)))).
# The liability is in points, so the breakage that matters is the
# point-weighted one. Table by balance quintile; expected values, no
# second random draw is needed.
import math
import random

N, MED, SD, A, B = 5000, 200.0, 1.2, 0.4, 1.1


def members(rng):
    return sorted(rng.lognormvariate(math.log(MED), SD) for _ in range(N))


def p_redeem(b):
    return 1 / (1 + math.exp(-(A + B * math.log(b / MED))))


if __name__ == "__main__":
    bal = members(random.Random(20261113))
    total = math.fsum(bal)
    print("quintile  mean.bal  share.pts  p.redeem  breakage.pts")
    for q in range(5):
        grp = bal[q * N // 5:(q + 1) * N // 5]
        pts = math.fsum(grp)
        brk = math.fsum(b * (1 - p_redeem(b)) for b in grp)
        print("{:8d} {:9.1f} {:10.3f} {:9.3f} {:13.3f}".format(
            q + 1, pts / len(grp), pts / total,
            math.fsum(p_redeem(b) for b in grp) / len(grp), brk / pts))
    by_member = math.fsum(1 - p_redeem(b) for b in bal) / N
    by_points = math.fsum(b * (1 - p_redeem(b)) for b in bal) / total
    print("breakage by members {:.3f}  by points {:.3f}".format(
        by_member, by_points))
    print("liability error if member rate is used: {:+.3f}".format(
        (1 - by_member) / (1 - by_points) - 1))

Single source and citation

Fan, Y. (2026). Loyalty-Point Liabilities and Breakage Actuarial Science: Issuance, Redemption Rates and the Insolvency Boundary of Deferred Revenue (Version 1.0.0), Chapter 3: Estimating Expected Breakage: Matured Cohorts, Point Weighting and the Reversal Constraint. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-315 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026315,
  author    = {Fan, Yafu},
  title     = {Loyalty-Point Liabilities and Breakage Actuarial Science: Issuance, Redemption Rates and the Insolvency Boundary of Deferred Revenue},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book