Maximizing Time-discounted Influential Sustainability: Balancing Speed and Longevity in Social Marketing

TL;DR

In the world of social marketing, being a "one-hit wonder" is a failure. While advertisers want their products to spread fast to beat competitors, they also need sustained engagement. This paper proposes a new mathematical model—Time-discounted Influential Sustainability—and a greedy algorithm to determine exactly who to target and when to activate them to ensure a marketing campaign stays relevant over time without losing the "first-mover" advantage.

Problem & Motivation: The "Flash-in-the-Pan" Paradox

Traditional Social Marketing research focuses on Influence Maximization (IM): finding a set of $k$ nodes to reach the maximum number of people. However, researchers have observed a critical flaw: a topic can go viral and reach millions but disappear from public attention almost instantly.

The authors identify two conflicting needs:

1. Influential Sustainability: Maintaining a long-term impact where each "wave" of the campaign reaches at least $\mu$ new users.
2. Time-Discounting: Recognizing that information loses value over time; spreading early builds competitive barriers.

Most prior work treats these in isolation. Furthermore, they often assume seeds are activated simultaneously at $t=0$, which is rarely optimal for maintaining a long-burning flame of influence.

Methodology: The Math of Strategic Timing

The paper defines a utility function $U$ that rewards the campaign only when the number of newly activated nodes $R(t)$ exceeds a threshold $\mu$. This reward is then multiplied by a discount factor ${\gamma }^t$ (where $\gamma <1$).

The Complexity Challenge

The authors prove that this objective function is:

• NP-Hard: Computing the exact spread is computationally intensive.
• Non-Monotonic: Adding more seeds doesn't always increase sustainability (it might "burn out" the network too fast).
• Non-Submodular: The "diminishing returns" rule doesn't strictly apply here, making standard IM algorithms ineffective.

The Greedy Hill-Climbing Approach

To solve this, they propose a Greedy Algorithm that doesn't just pick a node, but picks a (node, time) pair. It looks for the seed that provides the best "marginal sustainability" at the best possible starting time.

Fig 1: A conceptual example showing how staggered activation timings can prevent the "exhaustion" of reachable nodes.

Experiments: Topology Matters

The researchers tested their method on three classic network types using real-world census data from Beijing:

1. ER (Random): Uniform connection probabilities.
2. WS (Small World): High local clustering (like local friend groups).
3. BA (Scale-Free): "Rich-get-richer" dynamics (like Twitter/X followers).

Key Findings:

• Algorithm Superiority: The Greedy algorithm consistently beat "Synchronous" activation. In WS networks, the utility was nearly 30% higher than activating all seeds at once.
• Seeding Strategies: Selecting High-Degree nodes (hubs) remains the "gold standard" for most scenarios. However, in Small World networks, Low-Degree nodes (the fringes) occasionally perform better because they don't exhaust the local clusters too quickly, allowing for a slower, more sustainable spread.
• The Failure of Betweenness: Interestingly, "Bridge" nodes (High-Betweenness) performed poorly under Linear Threshold (LT) models, sometimes yielding zero utility because they couldn't trigger the threshold $\mu$ across different network components.

Table 1: Performance comparison showing the clear advantage of the proposed Greedy algorithm across different network models (IC and LT).

Critical Insight & Conclusion

The core takeaway is that timing is a resource as valuable as the seeds themselves. By staggering the activation of influencers, advertisers can "pulse" the network, ensuring the campaign remains above the visibility threshold $\mu$ for as long as possible.

Limitations & Future Work

• Heuristic Nature: Because the function is non-submodular, the greedy approach lacks the 63% (1-1/e) optimality guarantee familiar to IM researchers.
• Model Expansion: Future research will likely move toward the SIR (Susceptible-Infected-Recovered) model to better simulate how people "recover" from an ad and become immune to further influence.

This work provides a robust foundation for modern digital marketing strategies where the goal isn't just to "go viral," but to "stay viral."