A structural subsidy for product quality

previously: The CAC crunch goes mainstream

Tyler Cowan quotes Patrick Collison:

In corporate strategy, people have long debated the relative importance of distribution and product quality. Perhaps you’ve made something better, but does anyone know about it? Won’t people just continue to buy the ACME Corp Mousetraps? Personal agents will, I think, act as a kind of structural subsidy for product quality. How would markets look different if individuals always spent at least 10 hours researching their purchases? On the margin, focusing on making a product better is, I think, going to become a more effective strategy for companies. This is good!

Better shopping bots (ok, fine, fine, “agents.” Your site has bots, my site has agentic AI early adopters) tend to lower the returns to adversarial personalization, because a good bot/agent can take on a variety of customer qualities to try to get the best possible deal.

Adversarial personalization: who it works for, who it doesn’t

Adversarial personalization works great for Big Tech companies, which provide services to both big-budget and small-budget users and to both honest and dishonest sellers. The biggest example of adversarial personalization is the Big Tech senior fraud situation. Many older people are spending less money on discretionary purchases but are sitting on retirement savings, so the platform ML “learns” to match them not with legit products and services, but with fraud operators.

Adversarial personalization is not just a Big Tech thing, though. It’s also being offered to normal companies, in the form of

• offering lower-quality products and services to customers with fewer options
• price discrimination

For normal companies, though, adversarial personalization is not such a good deal. Hardly any firms try to serve as wide a customer budget range as Big Tech, most firms offer products within a narrow quality range, and almost no firm accepts the level of crime by its counterparties that Big Tech does.

Collison makes the good point that

If true, a corollary is that information about product quality will become more important and impactful. Today, most people have very limited ability to aggregate and integrate such data, but agents will have insatiable thirst for such signals. (It’s often said that data is becoming more valuable because it helps in the training of models. That’s obviously true, but I think there’s a secondary way in which it’s more valuable simply as a complement to cheaper intelligence.)

The ROI on “the more you tell, the more you sell” goes up—and it also goes up relative to the ROI of adversarial personalization.

In a pre-agent shopping environment, a company has to balance investment in adversarial personalization against investment to possible core competencies. Those might include

• offer good service
• make better stuff
• keep prices down

The ROI of adversarial personalization is fairly predictable, because consumer information-gathering time is limited.

Add agents, though, and the alllocation decision becomes tougher. The company has to not just predict that the adversarial personalization will work, and the profit from doing it won’t be captured by the vendors and employees who do it, but that the company’s own “AI” will beat whatever de-personalizing “AI” bots the customer can bring on their shopping journey.

With more uncertain returns to adversarial personalization, the quality, service, and price options look a lot more attractive. A company can’t git gud at everything, and management is more likely to be able to hire the right people to pursue an industry-specific quality, service, or price advantage than to be able to outbid everyone else in every industry for adversarial personalization services and experts.

Impact on marketing projects

For quite a while, one of the most underrated but impactful marketing projects has been the photo gallery site. Hire a photographer who’s willing to license under Creative Commons, stage a bunch of photos related to the products or services, and put them up with informative captions and all the right Creative Commons incantations to make it clear that they’re available to Wikipedia, and for use on news sites, with credit.

Another useful project, sometimes pigeonholed under SEO, is the knowledge base. Go through the support tickets, find the ones with generally useful info, strip out any customer-specific material, and put them up on the web.

Open source software that enables use of a product (such as device drivers, or clients for an API) is another good example, and already a checklist item for some markets.

All three of those can be lumped under “enable a high-information customer to make a good decision” and, in the increasingly common case where (1) more customers are acting like high-information customers because of agent help and (2) the company is focusing on service, quality, or price instead of adversarial personalization, so is more likely to make a sale when the customer makes a good decision.

Bonus links

Privacy win: Norwegian Authority fines Microsoft’s Xandr € 8.5 Million [T]he Norwegian Data Protection Authority has decided that Microsoft’s Xandr violated the GDPR. It processed personal data, including sensitive information on sexual orientation, without a valid legal basis.

Decade-old RAM is making a comeback by Stevie Bonifield. Intel and AMD are making new CPUs that’ll support the older, more affordable type of memory.