Also on Substack ↗

As a student at a four-year college in the United States, you can get a discount on pretty much everything from your computer to your shoes. In fact, the discounts are so good that I used my undergraduate ID for years after finishing school, thanks to a lack of any stated expiration date and my generally boyish features. But hey, such is the cost of investing in America's youth.

Students are valuable customers because, in the act of choosing their first bank account, music streaming platform, or car insurance plan, they are likely committing years of additional surplus value to their selected brand. However, because they tend to have less discretionary income than their credentialed peers, a discount is often necessary to achieve any kind of conversion. The student discount, therefore, is a classic example of price discrimination: a way to convert into paying customers those individuals who value a product below its listed price.

In general, price discrimination is a way for companies to solve for the multiplicity of demand curves in their target market. Common examples range from airline seating to premium shampoo, and are not worth exhausting. There are two main problems that companies have to be careful of when pursuing such a strategy:

The first is fairness. There is a risk that your higher-price-paying customers will see such differentiation as "unfair," and generally incongruous with a democratic society. Fairness is subjective, of course—demand-based pricing for rideshare and airfare is acceptable, while surge pricing for toilet paper is not—but generally, people anchor on what they consider to be normative, and thus feel okay with the price they're paying so long as they know that most people are paying the same. This is why providing discounts to minority groups with high social value like students, veterans, or the NHS in the UK does not meet with much resistance, whereas a surcharge based on gender (the "pink tax") does.

The second (and arguably bigger) problem is accuracy. This is the risk that you accidentally give out lower prices than necessary, due either to incomplete information about your demand curves or inefficient segmentation of your customer base. I believe this is what prevents most forms of price discrimination from being effectuated.

Students, for instance, have a relatively accessible form of physical verification which, while not ironclad, generally presents a high-enough bar to exclude most customers outside the intended demographic. However, for products and/or services delivered digitally, such as news or software subscriptions, verification is more of an issue. The only feasible proof is an "edu" email address, which a large share of the degree-holding population still has access to. This means a company like Bloomberg, who gives students 75% off their standard $40/month subscription, likely has additional costs to support some kind of verification, either in-house or through a third-party provider.

To sum: an opportunity for price discrimination will be pursued if (1) it is perceived as fair, and (2) the combined costs of verification and leakage risk do not exceed the value of capturing the target segment.

Discovery costs

In a market where firms are bidding for the most valuable customers, it makes sense to pay more for high LTV customers through a form of discounted pricing, so long as there is enough margin remaining in the customer's expected payback period. The same does not hold in monopolistic or oligopolistic markets, where the risk of not acquiring a customer due to price trends toward zero.

Being enrolled in a four-year degree program is a signal of potentially high LTV, but it's not the only signal. There is a wide range of factors that firms value and compete for, some of which are universal (anything correlated with higher discretionary purchasing power, such as income or profession) and some of which are industry-specific (if you love to travel, for instance, then you are a valuable target for hotel loyalty programs).

Today, firms utilize these signals in a way that I would describe as antagonistic. You get targeted by ads, discounts, and other aggressive tactics based on known or presumed characteristics—and you either ignore these messages and attempt to live your life under minimal influence from consumer brands, or else you capitulate and become a buyer or user. We're all familiar with this world and its annoyances (although, for better or worse, your attention on any given platform is not actually all that valuable: about $15-20 per year, going off Meta ARPU, just a fraction of the ~$20,000 the average consumer spends on discretionary goods each year).

But what happens if you flip the model? What happens if consumers offer up signals of their value only when they choose and only in situations where it will be effective?

After all, in many cases, student discounts are not advertised—they are revealed by inquiry.

Is it possible to redistribute discovery costs?

Enter the agent

Consider a world in which agents act as representatives of consumers. They have the ability to verify key pieces of information—like being a student—and instructions from their owner as to what to purchase. They can communicate this information directly and privately to sellers, and complete a transaction on behalf of their owner.

By itself, this doesn't do much beyond making shopping a bit easier. If I can log into ChatGPT and tell it to buy me an earthenware pot, and the next day I see it arrive at my doorstep—well, great, I've saved a couple minutes. For most consumers, the risks of a misjudged purchase will outweigh any time saved by delegation.

Now consider this same scenario, except this time, the agent has access to some form of negotiation. An agent that returns a selection of products to buy is one thing, but an agent that saves you money off the list price is something else altogether.

A student discount, for example, is kind of like an ex ante form of negotiation. If an agent were to handle the inquiry and verification, then, theoretically, it becomes more cost effective to provide such a discount, and the digital economy gets closer to the physical one, i.e. more like a world in which I can enter a store, inquire about a student discount, and flash my ID where applicable.

Of course, student discounts by themselves are not substantive enough an opportunity for a digital seller to integrate with such a technology, in which the seller would still need to be able to "read" the verifications provided by the agent and provide lower pricing only when those verifications "pass" the agreed-upon standard, but the student discount is just an example here. Agentic commerce opens the door to a much broader array of price discrimination.

Examples

Let's say that I recently came into some money, and am looking for the services of a financial advisor. I may be hesitant to have a direct conversation with these advisors, during which I could unintentionally reveal my naïveté about pricing norms within the industry, or else get talked into unnecessarily aggressive investment strategies, and so I instead utilize the service of an agent. I give my agent the authority to reveal my age, profession, education, and net worth, and have it request as much information as advisors are willing to provide about their credentials, performance, and pricing.

Somewhere out there is the advisor whose performance and expertise I value at the right level, compared to the cost of managing my account. However, in a standard model, the effort of full price discovery—that is, the effort the advisor would need to take in order to effectively gauge my value, and the effort I would need to take to benchmark the offerings—results either in a lengthy process of research and negotiation, or else the use of a "sticker" price to shortcut the conversation. Since I am not willing to negotiate in-person, and the "sticker" price would, by necessity, be above my optimal willingness-to-pay, we do not find each other.

In an agentic model, on the other hand, the advisor is able to effectively adapt his "sticker" price to one that encapsulates the specifics of the prospect in question. This introduces a form of formalized negotiation to the process, based on the parameters set by the advisor. For instance: clients with more than $1m in assets could have their management fees reduced by 10 basis points, with an additional 5 basis points provided to clients under 40 and introductory pricing of zero fees for one year given to clients moving at least $5m from a set of selected competitors. Maybe there is even a pricing engine that can calculate a custom discount based on the client's past portfolio history. In either sense, the advisor now has a more efficient means of achieving price discrimination, which, in a competitive market where the consumer is equipped with good information (like quotes from other advisors), should allow the consumer to get more value for their dollar.

I selected an example from financial services for a reason, because it may seem crazy to suggest adding more levers of pricing differentiation to an industry that is already particularly susceptible to discriminatory and harmful consumer practices, but I would argue that it is better to formalize such discounts into transparent and auditable rules than leave firms to develop their own, messy proxies for these signals—which is exactly what happened when the pricing structure set up by Ally for its auto-loan partners resulted in a $98m judgement from the CFPB.

A lighter example might be one in which I am walking by a store in SoHo, and I see a shirt that I like in the window. I take a picture and ask my agent to contact the seller about pricing, leveraging the information I have already authorized my agent to share. The agent tells me that the seller is willing to offer a one-time discount of 18%, largely based on my purchase history of clothing from a similar brand, which is queryable and verifiable because I also made those purchases through my agent. So I tell my agent to share my location and buy the shirt, and it arrives at my apartment three days later.

There are a lot of possibilities here, many of which probably wouldn't work. You could break them down into three categories:

1. Sympathetic: exemplified by discounts provided to students, seniors, veterans, and low-income individuals. For this category, agentic commerce would serve to bring down the costs of verification and hopefully expand access to these kinds of discounts, although admittedly neither the verification costs nor the market opportunities of this category are expected to be significant.
2. Neutral: discounts based on loyalty, timing, bulk purchasing, etc. This is probably the broadest and most interesting category, as a multitude of discrete, idiosyncratic signals across industries could potentially be aggregated into a standardized method of verification and negotiation (such as with the shirt in SoHo).
3. Questionable: pricing related to income, profession, or network. There is a lot of favorable pricing already in existence for those who are well-off (thus the maxim, it's expensive to be poor) and formalizing these kinds of arrangements is admittedly risky. Confidence in regulatory schema to protect against unfair or discriminatory business practices would be important here, as would inherent limitations on any signals closely related to protected characteristics.

Not all of this would necessarily lead to better outcomes, but right now, it is expensive and inefficient for companies to run any kind of pricing experiment. Giving firms the flexibility to experiment on the metrics of their choice while leaving the decision to claim those metrics in the hands of consumers presents, at the very least, an interesting opportunity for more efficient pricing dynamics in competitive markets, and puts much more value at play than do existing platforms. With negotiation starting to happen via agents already, as exemplified by tools like CarEdge, firms will be looking for ways to redistribute the surplus that in the past was provided to consumers who placed less value on their time (e.g. the coupon-clipper), presenting an opportunity to those who can send the right signals.

And if nothing else, we should at least be pushing for more student discounts.