5D Library › Algorithmic Marketing, Multi-Touch Shapley Attribution and Stochastic Control of Acquisition Funnels · 中文

Chapter 6 Diffusion and Spend: the Bass Model, the Dorfman–Steiner Condition and Whole-Life Budgets

The Bass diffusion model

In 1969 Bass proposed in Management Science a model of the growth of first purchases of a new product (Bass 1969). Let MM be the number of potential buyers and F(t)F(t) the fraction who have bought by time tt; then

dFdt=(p+qF)(1−F).\frac{dF}{dt} = (p + qF)(1 – F).

pp is the coefficient of innovation, the external influence (advertising, media) independent of earlier buyers; qq is the coefficient of imitation, the internal influence of earlier buyers through word of mouth. The equation in the author’s framework checks out. Its solution is

F(t)=1−e−(p+q)t1+qpe−(p+q)t,F(t) = \frac{1 – e^{-(p+q)t}}{1 + \frac{q}{p}e^{-(p+q)t}},

and when q>pq > p new purchases peak at t*=ln⁡(q/p)/(p+q)t^* = \ln(q/p)/(p+q), where F(t*)=12−p2qF(t^*) = \tfrac12 – \tfrac{p}{2q}. It is the fully mixed version of the “simple contagion” of Book 25: each non-buyer is influenced by buyers with a probability proportional to the fraction who have bought, with no network structure and no complex-contagion threshold.

Evidence

The empirical range of the parameters. Sultan, Farley and Lehmann meta-analysed 213 applications of diffusion models from 15 articles, relating the parameters to the type of innovation, the country, the model specification and the estimation method (Sultan et al. 1990). Parameters differ greatly between products and markets, so the parameters in this chapter are only illustrative. Bass, Krishnan and Jain (1994) proposed the “Generalized Bass Model”, which puts decision variables such as price and advertising into the model, and explained why the original model without them still fits many data sets well: when the percentage changes in the decision variables are roughly stable, their effect is absorbed into pp and qq (Bass et al. 1994).

The Dorfman–Steiner condition

Dorfman and Steiner (1954) studied how a monopolist chooses price, advertising and quality together (Dorfman and Steiner 1954). The advertising part of their result is that at the profit maximum

APQ=βη,\frac{A}{PQ} = \frac{\beta}{\eta},

that is, advertising as a share of sales equals the advertising elasticity of demand β\beta divided by the price elasticity η\eta (taken as positive). The intuition: the more price-elastic demand is, the smaller the markup, the less profit each unit of sales brings, and the less advertising is worth.

Calculation

# A: Bass (1969) diffusion dF/dt = (p + qF)(1 - F): closed form, peak time
#    ln(q/p)/(p+q) and F at the peak (1/2 - p/(2q)); p, q and market size M refitted
#    by least squares to a FICTIONAL noisy monthly series, before and after the peak.
# B: Dorfman-Steiner (1954). With demand Q = k P^(-eta) A^beta and unit cost c, profit
#    (P - c)Q - A is maximised at A/(PQ) = beta/eta (advertising elasticity over price
#    elasticity); checked by a direct numerical search. The rule is STATIC: it says
#    nothing about how to spread spend over a product's life (see p12).
import math, random

def bass_F(t, p, q):
    e = math.exp(-(p + q) * t)
    return (1 - e) / (1 + q / p * e)

def fit(sales):
    """Least squares over p, q (grid, then local search); the market size M enters
    linearly, so for each (p, q) it is profiled out in closed form."""
    def sse(pq):
        g = [bass_F(t + 1, *pq) - bass_F(t, *pq) for t in range(len(sales))]
        M = sum(a * b for a, b in zip(g, sales)) / sum(a * a for a in g)
        return sum((M * a - b) ** 2 for a, b in zip(g, sales)), M
    best = min(((sse((i / 2000, j / 100))[0], (i / 2000, j / 100))
                for i in range(1, 60) for j in range(1, 150)))
    (p, q), step = best[1], 0.0005
    for _ in range(60):
        cand = [(max(p + dp, 1e-5), max(q + dq, 1e-4)) for dp in (-step, 0, step)
                for dq in (-10 * step, 0, 10 * step)]
        p, q = min(cand, key=lambda pq: sse(pq)[0]); step *= 0.8
    return p, q, sse((p, q))[1]

if __name__ == "__main__":
    p, q, M = 0.01, 0.40, 100_000
    tp = math.log(q / p) / (p + q)
    print(f"A. Bass p {p}, q {q}, market {M:,}: peak adoption at month {tp:.2f},")
    fp = 0.5 - p / (2 * q)
    print(f"   F(peak) {bass_F(tp, p, q):.4f} (formula 1/2 - p/(2q) = {fp:.4f});"
          f" sales with 15 % noise")
    rng = random.Random(1969)
    sales = [max(0.0, M * (bass_F(t + 1, p, q) - bass_F(t, p, q))
                 * (1 + rng.gauss(0, 0.15))) for t in range(36)]
    print("   fitted on        p        q      market M   peak month")
    print(f"   (truth)       {p:.4f}   {q:.3f}   {M:9,.0f}      {tp:5.2f}")
    for n in (6, 9, 12, 36):
        pp, qq, MM = fit(sales[:n])
        print(f"   months 1-{n:<3d}   {pp:.4f}   {qq:.3f}   {MM:9,.0f}"
              f"      {math.log(qq / pp) / (pp + qq):5.2f}")
    print("B. Dorfman-Steiner: Q = k P^-eta A^beta, unit cost c = 10, k = 1e6")
    print("   eta   beta   optimal A/(PQ) by search   beta/eta   optimal price")
    for eta, beta in ((2.0, 0.10), (3.0, 0.10), (2.0, 0.25)):
        k, c = 1e6, 10.0
        prof = lambda P, A: (P - c) * k * P ** -eta * A ** beta - A
        best = max(((prof(P, A), P, A) for P in [10 + 0.05 * i for i in range(1, 600)]
                    for A in [10 * 1.03 ** j for j in range(500)]))
        _, P, A = best; R = P * k * P ** -eta * A ** beta
        print(f"   {eta:3.1f}   {beta:4.2f}          {A / R:6.4f}"
              f"               {beta / eta:6.4f}"
              f"     {P:6.2f} (eta/(eta-1) c = {eta / (eta - 1) * c:.2f})")
# A. Bass p 0.01, q 0.4, market 100,000: peak adoption at month 9.00,
#    F(peak) 0.4875 (formula 1/2 - p/(2q) = 0.4875); sales with 15 % noise
#    fitted on        p        q      market M   peak month
#    (truth)       0.0100   0.400     100,000       9.00
#    months 1-6     0.0165   0.433      66,359       7.27
#    months 1-9     0.0067   0.294     219,286      12.58
#    months 1-12    0.0122   0.390      96,003       8.62
#    months 1-36    0.0126   0.364     102,285       8.93
# B. Dorfman-Steiner: Q = k P^-eta A^beta, unit cost c = 10, k = 1e6
#    eta   beta   optimal A/(PQ) by search   beta/eta   optimal price
#    2.0   0.10          0.0497               0.0500      20.00 (eta/(eta-1) c = 20.00)
#    3.0   0.10          0.0333               0.0333      15.00 (eta/(eta-1) c = 15.00)
#    2.0   0.25          0.1240               0.1250      20.00 (eta/(eta-1) c = 20.00)

Part A takes p=0.01p = 0.01, q=0.40q = 0.40, M=100,000M = 100{,}000; the peak is at month 9.00, with F=0.4875F = 0.4875 at the peak, as the formula says. Then 15 % noise is added to monthly sales and pp, qq and MM are estimated by least squares. With only the first 6 months the market size is estimated at 66,359 and the peak month at 7.27; with the first 9 months (just up to the peak) the market size is estimated at 219,286 and the peak month at 12.58; only from 12 months on do the estimates settle near the truth. This shows a practical problem: before the peak, the three Bass parameters are hard to separate, and early forecasts (and budgets based on them) can be off by a factor of two or more.

Part B searches directly for the optimal price and advertising under constant-elasticity demand Q=kP−ηAβQ = kP^{-\eta}A^{\beta} with unit cost 10; the advertising-to-sales ratios found (0.0497, 0.0333, 0.1240) agree with β/η\beta/\eta (0.0500, 0.0333, 0.1250) up to the grid spacing, and the optimal price equals ηc/(η−1)\eta c/(\eta – 1), the standard markup rule.

Note

The Dorfman–Steiner condition is static. It compares marginal revenue and marginal cost within one period, with no “today’s advertising brings tomorrow’s customers”. The author’s framework hopes to “combine the Bass diffusion equation with the Dorfman–Steiner theorem to find the whole-life optimal allocation of acquisition spend”; applying a static condition period by period to a dynamic problem amounts to assuming each period’s advertising affects only that period. The dynamic problem requires advertising in the state equation: Nerlove and Arrow (1962) treated advertising as investment in a “goodwill” stock that depreciates over time (Nerlove and Arrow 1962); Vidale and Wolfe (1957) gave an early differential equation for the response of sales to advertising (Vidale and Wolfe 1957); Horsky and Simon (1983) put advertising into the innovation term of the Bass model and studied the optimal advertising path over a new product’s life (Horsky and Simon 1983).

The author’s whole-life spend model

What follows is the author’s model, in the spirit of Horsky and Simon but simpler, with illustrative parameters. New buyers in month tt are

nt=M(p0+g(At)+qFt)(1−Ft),n_t = M\bigl(p_0 + g(A_t) + qF_t\bigr)(1 – F_t),

where AtA_t is acquisition spend in month tt and gg the advertising response function; the aim is to maximise discounted profit ∑tdt(mnt−At)\sum_t d^t (m\,n_t – A_t) over 36 months, with gross margin m=60m = 60 per new customer, monthly discount factor d=0.98d = 0.98, M=100,000M = 100{,}000, p0=0.002p_0 = 0.002 and q=0.20q = 0.20. The optimal spend path is found by month-by-month golden-section search and compared with three rules using the same total budget: even monthly spend, a fixed share of sales (the Dorfman–Steiner rule carried over period by period) and no advertising.

Calculation

# THE AUTHOR'S MODEL of "whole-life optimal acquisition spend" (a Bass model with
# advertising in the innovation term, in the spirit of Horsky & Simon 1983). Monthly:
#   new adopters n_t = M (p0 + g(A_t) + q F_t)(1 - F_t),  profit = sum d^t (m n_t - A_t)
# M 100,000, p0 0.002, q 0.20, margin m 60 per adopter, 36 months, d = 0.98.
# A: concave response g(A) = 0.02 ln(1 + A / 20,000). Optimal schedule by coordinate
#    search; compared with even spend and a Dorfman-Steiner-style constant share of
#    revenue, each with the same total budget, and with no advertising.
# B: S-shaped response g(A) = 0.03 A^2 / (40,000^2 + A^2): pulsing (twice the monthly
#    budget in alternate months) against even spend of the same total, months 1-12,
#    at several budgets, to locate where the better policy switches.
import math

M, P0, Q, MARGIN, T, D = 100_000, 0.002, 0.20, 60.0, 36, 0.98

def path(A, g):
    F, prof, sales = 0.0, 0.0, []
    for t, a in enumerate(A):
        n = M * min(P0 + g(a) + Q * F, 1.0) * (1 - F)
        F += n / M; sales.append(n); prof += D ** t * (MARGIN * n - a)
    return prof, sales, F

def optimise(g, sweeps=8):
    A = [20_000.0] * T
    for _ in range(sweeps):
        for t in range(T):
            lo, hi = 0.0, 400_000.0                    # golden-section on A_t
            for _ in range(40):
                m1, m2 = lo + 0.382 * (hi - lo), lo + 0.618 * (hi - lo)
                f1 = path(A[:t] + [m1] + A[t + 1:], g)[0]
                f2 = path(A[:t] + [m2] + A[t + 1:], g)[0]
                lo, hi = (lo, m2) if f1 > f2 else (m1, hi)
            A[t] = (lo + hi) / 2
    return A

def share_of_revenue(g, budget):
    lo, hi = 0.0, 1.0                                   # constant A_t = s * revenue_t
    for _ in range(50):
        s = (lo + hi) / 2; F, A = 0.0, []
        for t in range(T):
            a = 0.0
            for _ in range(30):                         # fixed point A = s * m * n(A)
                a = s * MARGIN * M * min(P0 + g(a) + Q * F, 1.0) * (1 - F)
            n = M * min(P0 + g(a) + Q * F, 1.0) * (1 - F); F += n / M; A.append(a)
        lo, hi = (s, hi) if sum(A) < budget else (lo, s)
    return A

if __name__ == "__main__":
    g1 = lambda a: 0.02 * math.log(1 + a / 20_000)
    A = optimise(g1); B = sum(A)
    print("A. concave response; profit in thousands, budget fixed at the optimum's total")
    print(f"   optimal schedule (thousands): months 1-6 "
          + " ".join(f"{a / 1e3:.0f}" for a in A[:6])
          + f"; 12: {A[11] / 1e3:.0f}; 24: {A[23] / 1e3:.0f}; 36: {A[35] / 1e3:.0f}")
    stop = min(t for t in range(T) if A[t] < 1000) + 1
    print(f"   spend falls below 1 thousand in month {stop};"
          f" {sum(A[:6]) / B:.0%} of the total falls in months 1-6")
    rows = (("optimal", A), ("even", [B / T] * T),
            ("constant share of revenue", share_of_revenue(g1, B)),
            ("no advertising", [0.0] * T))
    for lab, sched in rows:
        pr, sales, F = path(sched, g1)
        peak = max(range(T), key=lambda t: sales[t]) + 1
        print(f"   {lab:26s} spend {sum(sched) / 1e3:7.0f}  profit {pr / 1e3:8.1f}"
              f"  peak month {peak:2d}  F(36) {F:.3f}")
    g2 = lambda a: 0.03 * a * a / (40_000 ** 2 + a * a)
    print("B. S-shaped response, first 12 months, same total: even against pulsed")
    print(f"   inflection k/sqrt(3) = {40 / math.sqrt(3):.1f} thousand; static break-even"
          f" k/sqrt(2) = {40 / math.sqrt(2):.1f}")
    print("   monthly budget   even profit   pulsed profit   better")
    for mon in (15, 20, 25, 30, 32.5, 35, 40, 60):
        e = path([mon * 1e3] * 12 + [0.0] * 24, g2)[0]
        pu = path([2 * mon * 1e3, 0.0] * 6 + [0.0] * 24, g2)[0]
        print(f"   {mon:6.1f} thousand   {e / 1e3:9.1f}     {pu / 1e3:9.1f}"
              f"       {'pulsed' if pu > e else 'even'}")
    print("   same comparison under the concave response of A:")
    for mon in (15_000.0, 60_000.0):
        e = path([mon] * 12 + [0.0] * 24, g1)[0]
        pu = path([2 * mon, 0.0] * 6 + [0.0] * 24, g1)[0]
        print(f"   {mon / 1e3:3.0f} a month: even {e / 1e3:8.1f}, pulsed {pu / 1e3:8.1f}")
# A. concave response; profit in thousands, budget fixed at the optimum's total
#    optimal schedule (thousands): months 1-6 74 58 45 34 25 18; 12: 0; 24: 0; 36: 0
#    spend falls below 1 thousand in month 10; 93% of the total falls in months 1-6
#    optimal                    spend     273  profit   4493.2  peak month 13  F(36) 0.995
#    even                       spend     273  profit   4147.9  peak month 17  F(36) 0.988
#    constant share of revenue  spend     273  profit   3860.9  peak month 20  F(36) 0.977
#    no advertising             spend       0  profit   3548.1  peak month 25  F(36) 0.921
# B. S-shaped response, first 12 months, same total: even against pulsed
#    inflection k/sqrt(3) = 23.1 thousand; static break-even k/sqrt(2) = 28.3
#    monthly budget   even profit   pulsed profit   better
#      15.0 thousand      3946.7        4109.9       pulsed
#      20.0 thousand      4062.3        4182.0       pulsed
#      25.0 thousand      4138.9        4207.9       pulsed
#      30.0 thousand      4185.3        4206.1       pulsed
#      32.5 thousand      4199.5        4198.5       even
#      35.0 thousand      4208.9        4187.6       even
#      40.0 thousand      4215.4        4158.4       even
#      60.0 thousand      4140.7        3991.5       even
#    same comparison under the concave response of A:
#     15 a month: even   4361.4, pulsed   4317.3
#     60 a month: even   4267.6, pulsed   4152.0

Part A uses the concave response g(A)=0.02ln⁡(1+A/20,000)g(A) = 0.02\ln(1 + A/20{,}000). The optimal path is front-loaded: 74 thousand in month 1, falling every month, below 1 thousand from month 10, with 93 % of the budget in the first 6 months. The reason is that customers won early bring more customers through word of mouth, so the earlier the investment, the more of this “interest” it earns; once the market approaches saturation, advertising merely brings forward purchases that would have come anyway. Spending the same 273 thousand, the optimal path earns 4,493.2 thousand, even spending 4,147.9 thousand (the optimum is 8.3 % higher), and a fixed share of sales only 3,860.9 thousand, a little better than no advertising (3,548.1 thousand) but far worse than front-loading. A fixed share of sales does badly because it concentrates spending at the sales peak, exactly when word of mouth is strongest and the marginal effect of advertising smallest. Figure 5 shows the spend and new-buyer paths.

The author’s whole-life spend model (concave response). Left: optimal, even and share-of-sales spend paths (same total budget). Right: the corresponding monthly new buyers; dotted, no advertising (illustrative parameters).

Part B turns to pulsing. Mahajan and Muller (1986) analysed how pulsed and even advertising policies affect awareness of new products (Mahajan and Muller 1986); Feinberg (1992) showed that in common aggregate advertising models pulsing cannot be optimal, and that periodic pulsing can be optimal for some parameter values only when the response function is S-shaped and combined with a smoothing filter (Feinberg 1992). The program uses the S-shaped response g(A)=0.03A2/(40,0002+A2)g(A) = 0.03A^2/(40{,}000^2 + A^2) and compares even with alternate-month pulsing (same total) over the first 12 months at several monthly budgets: at 15 thousand a month pulsing earns 4,109.9 thousand, more than even spending’s 3,946.7; at 20, 25 and 30 thousand pulsing is still better (4,206.1 against 4,185.3 at 30); at 32.5 thousand even spending is slightly better (4,199.5 against 4,198.5), and at 35, 40 and 60 thousand clearly so (4,140.7 against 3,991.5 at 60). The switch lies between 30 and 32.5 thousand a month. It is not the inflection point of the response curve: that is at k/3≈23.1k/\sqrt3 \approx 23.1 thousand, and pulsing still wins at 25 and 30 thousand, above it. In a static one-month comparison, g(2m)/2>g(m)g(2m)/2 > g(m) exactly when m<k/2≈28.3m < k/\sqrt2 \approx 28.3 thousand; the dynamics of diffusion (word of mouth from early customers) push the switch somewhat higher. Under the concave response, pulsing is worse at either budget. Whether “pulsing” helps therefore depends on the shape of the response function and on the relative size of the budget, and both have to be measured by experiment; they cannot be derived from the model.

Note

This model is not budget advice. It assumes that the advertising response function is known and constant, that competitors do not react, that the word-of-mouth coefficient qq does not depend on advertising, and that all customers have the same margin; real media environments violate almost every one of these. It shows only a directional conclusion: in markets with word-of-mouth diffusion, a static “share of sales” rule can put money where it is least needed.

Summary

The Bass equation dF/dt=(p+qF)(1−F)dF/dt = (p+qF)(1-F) checks out, with peak time ln⁡(q/p)/(p+q)\ln(q/p)/(p+q); before the peak, market size and the imitation coefficient are hard to separate, and early forecasts can be off by a factor of two or more. The Dorfman–Steiner condition (advertising/sales = advertising elasticity/price elasticity) checks out, but it is a static condition and cannot by itself give a whole-life allocation. The author’s whole-life model shows that with concave response and word-of-mouth diffusion the optimal spend is front-loaded and earns 8.3 % more than even spending of the same budget, while a fixed share of sales does worst of the three policies with that budget (though still better than no advertising); under the concave response pulsing is always worse than even spending, and under the example’s S-shaped response pulsing helps only at relatively low monthly budgets (up to 30 thousand in this example), with a switch that does not coincide with the inflection point.

Hypothesis · not established · The word-of-mouth multiplier of early acquisition spend

The author’s hypothesis (not established): in new categories with clear word-of-mouth diffusion, regions randomly assigned higher acquisition spend in the first three months after launch have a larger share of “non-paid-channel” buyers among their 12-month cumulative buyers than control regions, i.e. early advertising has a multiplier effect through word of mouth. What result would show it to be wrong: in a pre-registered regional randomised experiment (run by independent researchers), the 95 % confidence interval for the ratio of 12-month cumulative non-paid buyers in treated and control regions contains 1.

References

Bass, Frank M. 1969. “A New Product Growth for Model Consumer Durables.” Management Science 15 (5): 215–27. https://doi.org/10.1287/mnsc.15.5.215.
Bass, Frank M., Trichy V. Krishnan, and Dipak C. Jain. 1994. “Why the Bass Model Fits without Decision Variables.” Marketing Science 13 (3): 203–23.
Dorfman, Robert, and Peter O. Steiner. 1954. “Optimal Advertising and Optimal Quality.” American Economic Review 44 (5): 826–36.
Feinberg, Fred M. 1992. “Pulsing Policies for Aggregate Advertising Models.” Marketing Science 11 (3): 221–34. https://doi.org/10.1287/mksc.11.3.221.
Horsky, Dan, and Leonard S. Simon. 1983. “Advertising and the Diffusion of New Products.” Marketing Science 2 (1): 1–17. https://doi.org/10.1287/mksc.2.1.1.
Mahajan, Vijay, and Eitan Muller. 1986. “Advertising Pulsing Policies for Generating Awareness for New Products.” Marketing Science 5 (2): 89–106. https://doi.org/10.1287/mksc.5.2.89.
Nerlove, Marc, and Kenneth J. Arrow. 1962. “Optimal Advertising Policy under Dynamic Conditions.” Economica 29 (114): 129–42.
Sultan, Fareena, John U. Farley, and Donald R. Lehmann. 1990. “A Meta-Analysis of Applications of Diffusion Models.” Journal of Marketing Research 27 (1): 70–77.
Vidale, M. L., and H. B. Wolfe. 1957. “An Operations-Research Study of Sales Response to Advertising.” Operations Research 5 (3): 370–81.

Fan, Y. (2026). Algorithmic Marketing, Multi-Touch Shapley Attribution and Stochastic Control of Acquisition Funnels: From Thompson Sampling Bandits and Absorbing Markov Chains to BG/NBD Customer Lifetime Value, Read Beside the Strategic Dispositions of the Sunzi [算法营销、多触点沙普利归因与获客漏斗的随机控制] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23836264 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).