Anuj Kapoor and Ekansh Chaturvedi

Anuj: Assistant Professor (Marketing), IIM-Ahmedabad, India & Ekansh: Assistant Manager, Bajaj Auto Ltd., Pune, India.

Anuj can be reached at anujk@iima.ac.in

An increasing number of marketing decisions are made using machine learning algorithms. Algorithms decide what price we pay for our products, what recommendations we are exposed to and what ads we see.

Algorithms need to be user centric and focus on user needs, interests, behavior, interactions and are personalized for each user. Are we missing anything? We talked about the user centric nature of the algorithms - but we must not forget that firms and sellers have objectives (usually profit maximization) and the two objectives (buyer and seller) are not necessarily aligned.

Therefore, we must account for conflicting objectives and multi-objective optimization accounts for such scenarios. Formally, multi-objective optimization considers optimization problems involving more than one objective function to be optimized simultaneously.

We deliberate on many used cases encountered by marketing managers where multi-objective maximization has been implemented with success.

1. Digital Platforms and Recommendation Systems:

Digital platforms involve efficient matching of buyers with sellers - the two prominent stakeholders on the platform. Both the stakeholders are interested in optimizing multiple objectives e.g. user-based behavioural metrics (click-through rate, time spent on the app) and supplier exposure objectives (that different sellers get equal representation and are not discriminated against).

Platforms needs to jointly optimize both these objectives. Researchers have successfully implemented multi objective maximization on Spotify - music streaming platform. They jointly optimize - user centric objectives (satisfaction with using the app) as well as supplier centric objectives (providing equal opportunity to all the music artists on the platform and not favoring only the top ones) while designing their recommendation system.

2. Advertising:

Advertising agencies face the similar issue of multi-objective optimization problem in TV advertising. The advertising agency's dilemma is to decide on which commercial breaks to air the ads of various brands to jointly maximize reach or gross rating point (GRP) for the different brands while minimizing budget, brand competition, and scheduling issues.

3. Marketing campaigns and Fashion Industry:

While optimizing a marketing campaign, factors that cause an increased level of activation among the campaign recipients could in fact reduce the profit, i.e., these conflicting factors need to be jointly optimized, rather than optimized individually.

Using multi-objective optimization, researchers leveraged campaign data from a well-established company within the fashion retail industry and have demonstrated that activation and profit can be simultaneously targeted, using algorithms as well as human-controlled visualization.

4. Influencer Marketing:

Marketers need to effectively target \textit{influencers} - prominent members of social networks who have the power to influence individuals in their circle. They can advertise products or services with free items or discounts to spread positive opinions to other consumers (i.e., word-of-mouth). However, choosing the best influentials to target is single-objective and mainly focused on maximizing sales revenue. Research has documented a multi-objective approach to the influence maximization problem i.e. one that increases the revenue of viral marketing campaigns and at the same time reduces the costs.

5. Corporate Social Responsibility:

Firms can define their corporate social responsibility policies as optimizing both the private objective (such as profit) and a public objective (such as social welfare).

6. Sustainable Supply Chains:

Sustainability hinges on three prominent pillars —economic, environmental, and social. To build a sustainable supply chain, marketers need multi objective model which jointly minimizes costs, emissions, and employee injuries in a supply chain. This multi objective optimization will balance the social, environmental, and economic performance.

We emphasize that a marketing manager relying on algorithmic decision making needs to account for multi-objective versus single objective optimization to come up with fair, equitable and accurate managerial decisions.