[Thesis Review] Personalized pricing and advertising: Who are the winners?

This post is a review of a thesis Personalized pricing and advertising: Who are the winners? by Rosa-Branca Esteves and Joana Resende. There will be parts deleted and elaborated according to my understanding. If you want more details and equations please consult the paper.

Personalized pricing and advertising: Who are the winners?

I. Introduction

First of all, what is personalized pricing and advertising? After a few searches in Amazon, you might have seen a commercial of that product or similar products on your instagram feed. In my case, I do follow some sports/yoga brands like lululemon or manduka. So, I very often see commercials from sports brands.

Sadly, we see some irrelevant commercials, from a firm’s point of view it failed in efficient private advertising. In my case, I have no idea what ganesha_yogaprops are up to (have never heard of it) or sorry for Under Armour fans, but their products do not match my type of exercise nor my body shape. Although I brought failed examples of instagram advertising, it is a big deal in the market. To put it another way, the market is deviating from its classic TV commercials or mass advertising to personalized advertising methods.

If ‘personalized’ advertising were possible, ‘personalized’ pricing would also be possible. We often observe this pricing strategy in flight ticket sites or hotel booking sites. It is exceptionally prevalent in flight industry.

In my example of booking.com it is more of a VIP discount strategy, however, this also can viewed as a personalized pricing. It is naturally difficult to find an example of personalized pricing because the key point of the strategy is that everyone thinks that he/she is buying at a low price. When your lover cheats on you he or she will keep that a secret. Then what is your strategy? We go incognito 😎 in chrome, delete search records or turnoff cookies, etc.

Enough of interesting examples, what made this possible? We are going back to our beloved machine learning. However, as you already know, even a simple image analysis was 9 min read, if so, how much time and energy would it take for a firm/employees to analyze consumers’ search and consumption data and to expose us to certain commercials and prices.

As a student of economic department, we do not want to waste our $ 💸if it does not retrieve profit or welfare. So, in this thesis review, we will compare profits of a duopoly firms before and after personalized advertising and pricing strategies. Also, we will also think of consumers’ welfare. We do not want consumers’ welfare to be harmed by a new technology. (b/c we can never ‘not’ be a consumer.) This kind new method of pricing is controversial since if it does increase the total welfare for sure, it contradicts with the bible of competitive pricing. On top of that, it is also connected with antitrust issue of an oligopoly (whether or not it intensifies the oligopoly) and privacy issues of normal people.

II. Model setup

This is not a machine learning modeling so don’t freak out! 🕊
The main goal of this paper is to analyze the competitive (will new firms enter and end the duopoly?) and welfare effects of price discrimination with targeted advertising and pricing in comparison with the classic mass advertising and pricing.

• Static game of duopoly competition with 2 new differentiated products.
  →decision on personalized strategy are taken simultaneously. (non-cooperative)
• Each firm faces two types/segments of consumers; consumers with stronger preference to my firm vs. to rival firm with equal mass. (brand preference degree parameter)
• Through advertising, firms can differentiate consumers who are captive (partially informed) and selective (fully informed)

Side note) If you need some notations and longer version (copied from the thesis)

Consider a market with two firms, i = A, B. Each firm is launching a new good, produced at a constant marginal cost, which is assumed to be zero without loss of generality. Firms need to invest in informative advertising to generate demand, meaning that consumers are only aware of the existence and the price of the new products when exposed to firms’ ad campaigns. There is a unit mass of consumers with a common reservation price v. Each consumer buys at most a single unit of either good A or B. The set of consumers is divided in two segments with equal size: segment a and segment b. Consumers in segment i prefer product i over the other product by a degree equal to γ > 0, which can be interpreted as the degree of consumer’s preference towards its favorite brand.

III. Mass advertising and no price discrimination.

Mcdonalds’ 1987 commercial on value pack

• Firms only have access to mass advertising technologies and uniform pricing.
• It costs A(φi) = λφi² for a firm i to reach φi fraction of consumers. Same for firm j.
  and this advertising creates φi(1 − φj ) fraction of captive consumers(partially informed) for firm i and φiφj fraction of selective consumers (fully informed) and (1 − φi)(1 − φj ) of consumers who do not even know that such product exists.
• And the demand function will be given by considering the utility from the product (v ), price, brand preference parameter (γ) and whether or not a consumer is aware of the product (exposure to an advertisement)

• Firm i’s expected profit is Eπi = piDi(pi, pj ) − A(φi).

Equilibrium analysis:

Symmetric NE, Symmetric subgame NE (Propositions 1 to 4)

• Firm i’s expected profit given the rival’s decisions.
  Eπi = pφi(1 − φj ) + pφiφj[ 1 − 1 /2 F(p + γ) − 1 /2 F(p − γ)] − A(φi).
• Each firm serves its group of selective customers with probability q (proposition4)

• We say that firms have an incentive to serve all segment consumers. (v is sufficiently high, and A(φi) is not too high)

IV. Targeted advertising and price discrimination

• Advertising technology of A(φo i , φr i ) = λ(φo i )²+ λ(φr i )², here o and r are abbreviation for own and rival.
• Firm i’s expected profit with targeted advertising
  Eπt = (2λ + v − γ /2λ ) A(φo∗) + A(φr∗)
• Proposition 5 derives under perfect targeted advertising and price discrimination, a probability z that each firms wins the group of selective consumer in its own market. (prob that consumers buy what they prefer)

Some conclusions on comparing z and q.

• For sufficiently low v and high λ, the probability of a consumer choosing wisely is higher under uniform pricing than under price discrimination
• If the advertising costs are small or v is high enough the reverse happens.
• q is less responsive to the advertising cost λ than z.
• In targeted discrimination case, as λ goes down φo i increases. So, in segment i, the set of consumers who are potentially captive to the firm j becomes narrower, intensifying price competition. IOW, z goes up with λ going down.

V. Compare Mass advertising v.s. targeted Advertising

1. The two firms will compete in both market segments only when the ad costs are not excessively high and the consumers’ willingness to pay for the products is sufficiently high in relation to the brand preference degree parameter. (UNDERARMOUR)
2. When both firms compete in both market segments, average prices with mass advertising and uniform pricing are BELOW their counterparts with personalized strategy regardless of market segments. — Firms have motive for deviate to personalized pricing to sustain higher prices.
3. Consumers can be worse off.
4. In conclusion, compared to mass advertising and uniform pricing, firms’ expected profits can be increased though targeted strategy, especially when the advertising costs are low enough. So the information technology can boast firms’ profit and the prices that consumers face.

Thank you for reading it so far.
I am not sure how you felt on this non-computer related post.
This is not yet fully written, sorry for any related confusions.

Some proofs and propositions were not adequate for explaining in the post and sometimes they were beyond my own knowledge and understanding. So please do scan through the original paper.

If time permits, I will come back with some interesting paper reviews or sometimes non-academic writings. So please stay tuned.

Thank you as always and hope your day is not rainy.
October has just begun and it is starting to be chilly outside with some spits here in Seoul. Hope you are all well. 🤧