Price experimentation is an important tool for firms to find the optimal selling price of their products. It should be conducted properly, since experimenting with prices can costly. A firm, therefore, needs a pricing policy that optimally balances between learning and gaining revenue. In this paper, we propose such policy, called controlled variance (CVP). The key idea enhance certainty equivalent taboo interval around average previously chosen prices. width shrinks at appropriate rate as...

We consider a seller’s dynamic pricing problem with demand learning and reference effects. first study the case in which customers are loss-averse: they have price that can vary over time, reduction when selling exceeds dominates increase falls behind by same amount. Thus, expected as function of has time-varying “kink” is not differentiable everywhere. The seller neither knows underlying nor observes prices. In this setting, we design analyze policy (i) changes very slowly to control...

We study a dynamic pricing problem with finite inventory and parametric uncertainty on the demand distribution. Products are sold during selling seasons of length, that is unsold at end season perishes. The goal seller to determine strategy maximizes expected revenue. Inference unknown parameters made by maximum-likelihood estimation. show this satisfies an endogenous learning property, which means learned fly if chosen prices sufficiently close optimal ones. small modification certainty...

We study a dynamic pricing problem with multiple products and infinite inventories. The demand for these depends on the selling prices parameters unknown to seller. Their value can be learned from accumulating sales data using statistical estimation techniques. quality of parameter estimates is influenced by amount price dispersion; however, large variation in costly since it means that suboptimal are used. seller thus needs balance optimizing instant revenue, i.e., exploitation exploration....

We consider a dynamic pricing problem with an unknown and discontinuous demand function. There is seller who dynamically sets the price of product over multiperiod time horizon. The expected for piecewise continuous parametric function charged price, allowing possibly multiple discontinuity points. initially knows neither locations points nor parameters but can infer them by observing stochastic realizations time. measure seller’s performance revenue loss relative to clairvoyant underlying...

Problem definition: This paper addresses the question whether or not self-learning algorithms can learn to collude instead of compete against each other, without violating existing competition law. Academic/practical relevance: is practically relevant (and hotly debated) for regulators, and academically in area analysis multi-agent data-driven algorithms. Methodology: We construct a price algorithm based on simultaneous-perturbation Kiefer–Wolfowitz recursions. derive theoretical bounds its...

We consider dynamic pricing and demand learning in a duopoly with multinomial logit demand, both from the perspective where firms compete against each other aim to collude increase revenues. show that joint‐revenue maximization is not always beneficial compared Nash equilibrium, several axiomatic notions of collusion can be constructed are threat consumer welfare. Next, we construct price algorithm prove it learns charge supra‐competitive prices if deployed by firms, respond optimally class...

We examine recent claims that a particular Q-learning algorithm used by competitors 'autonomously' and systematically learns to collude, resulting in supracompetitive prices extra profits for the firms sustained collusive equilibria. A detailed analysis of inner workings this reveals there is no reason alarm. set out what needed demonstrate existence colluding price does form threat competition.

We study the problem of an agent continuously faced with decision placing or not trust in institution. The makes use Bayesian learning order to estimate institution's true trustworthiness and place based on myopic rationality. Using elements from random walk theory, we explicitly derive probability that such ceases at some point relationship, as well expected time spent conditioned their discontinuation thereof. then continue by modeling two truster agents, each own relationship consider...

The topic of dynamic pricing and learning has received a considerable amount attention in recent years, from different scientific communities. We survey these literature streams: we provide brief introduction to the historical origins quantitative research on demand estimation, point subfields area pricing, an in-depth overview available learning. Our focus is operations management science literature, but also discuss relevant contributions marketing, economics, econometrics, computer...

This paper contributes to the ongoing debate on plausibility of tacit collusion between sellers in algorithmic marketplaces, which can be detrimental customers and social welfare. We study a broad class assortment decisions routinely made by online platforms, including products are offered customers, at what price, how they displayed. In this context, decision-support tools extensively studied operations literature widely adopted practice. propose simple notions collusive outcomes describe...

We study the question whether giving discounts for perishable products on their expiry dates can simultaneously reduce waste and increase profit. In this paper, we consider a seller of single product who daily replenishes inventory up to certain order-up-to level, serves customers whose purchase probabilities both depend price remaining shelf life product. model dynamics as Markov process show that system admits unique stationary distribution. This distribution does not lead informative...

We study the interpersonal trust of a population agents, asking whether chance may decide if ends up with high or low trust. model this by discrete time, stochastic coordination game pairwise interactions occurring at random in finite population. Agents learn about behavior using weighted average what they have observed past interactions. This learning rule, called an “exponential moving average,” has one parameter that determines weight most recent observation and may, thus, be interpreted...

We consider a seller's dynamic pricing problem with demand learning and reference effects. first study the case where customers are loss-averse: they have price that can vary over time, reduction when selling exceeds dominates increase falls behind by same amount. Thus, expected as function of has time-varying "kink" is not differentiable everywhere. The seller neither knows underlying nor observes prices. In this setting, we design analyze policy (i) changes very slowly to control evolution...

In this note we study the behavior of maximum quasilikelihood estimators (MQLEs) for a class statistical models, in which only knowledge about first two moments response variable is assumed. This includes, but not restricted to, generalized linear models with general link function. Our main results are related to guarantees on existence, strong consistency and mean square convergence rates MQLEs. The obtained from principles stronger than known a.s. rates. find important application...

Abstract We study a single‐product fluid‐inventory model in which the procurement price of product fluctuates according to continuous time Markov chain. assume that fixed order price, addition state‐dependent holding costs are incurred, and depletion rate inventory is determined by sell product. Hence, at any controller has simultaneously decide on selling whether or not, taking into account current level. In particular, faced with question how best exploit random windows low. consider two...