Affiliate marketing: from Tobin's 1989 patent to Amazon Associates and the cookieless reality of 2026

Affiliate marketing is an acquisition model in which a brand pays a commission to third parties (creators, websites, companies, influencers) for each sale, lead, or specific action attributable to the affiliate's effort. The brand gets qualified traffic while paying only when there is a result; the affiliate earns revenue without taking on the cost of inventory, support, or fulfillment. The idea sounds modern, but it has almost four decades of history and quite a few operational layers that are rarely explained well.

Before getting into how it works in 2026, it's worth placing it historically, because the model has a specific origin, a pivotal case, and a recent technological transition that is substantially changing it.

The origin: Tobin (1989), CDNow (1994), Amazon (1996)

The formal paternity of online affiliate marketing is attributed to William J. Tobin, founder of PC Flowers & Gifts, who in 1989 launched an affiliate program on the Prodigy network (an online service that predated the commercial internet). Tobin filed for U.S. patent 6,141,666 ("System and Method for Tracking Web-based Sessions"), which covered central aspects of session tracking for commission attribution. The patent was granted in 2000, but the claims covered the practice established during the 1990s.

In 1994, CDNow -- one of the first online music stores -- launched its BuyWeb program in collaboration with Geffen Records. The arrangement was simple: artists and labels could link to specific albums on CDNow from their websites, and CDNow paid a commission for each attributable purchase. It's one of the earliest documented cases of the model in real commercial practice on the web.

But the moment that turned the model into a category was the launch of Amazon Associates in July 1996. Amazon opened its program to the general public: anyone with a website, blog, or mailing list could sign up, receive unique affiliate links, and earn a commission when someone bought a book after clicking their link. The mass adoption of the model started there. Today, Amazon Associates is still one of the most widely used affiliate programs in the world, although its commission rates have been falling over the decades.

In the late 1990s the first specialized networks appeared: Commission Junction (1998, now CJ Affiliate), LinkShare (1996, now Rakuten Advertising), ShareASale (2000, acquired by Awin in 2017). These networks centralize the model: the advertiser brand signs up with the network, affiliates sign up once and can promote multiple brands, and the network handles tracking, attribution, validation, and payments.

Knowing this history helps you avoid the common confusion of treating affiliate marketing as a recent invention of influencer marketing. The model has been running for thirty-five years; influencer marketing is a modern variant with its own particularities.

The commission models: four main categories

There are four typical payment models in affiliate marketing, each with distinct operational dynamics:

CPS -- Cost Per Sale. The affiliate earns a commission on a confirmed sale. This is the classic Amazon Associates model. It's usually expressed as a percentage of the sale value (typically 1-15% in e-commerce, up to 50% or more in software/SaaS and digital products). The affiliate's risk is high (they don't get paid if there's no sale), but the commission per conversion is too.

CPA -- Cost Per Action. The affiliate earns money for a specific action that isn't necessarily a sale: a signup, a download, a completed form, a free trial started. Useful for products with a long sales cycle where the brand wants to pay for qualified leads even if the sale takes months to close.

CPL -- Cost Per Lead. A subcategory of CPA where the action is specifically a qualified lead. Common in B2B sectors with consultative sales, in mortgages, insurance, and education.

RevShare -- Revenue Share. The affiliate earns a recurring percentage of the revenue the customer generates over their lifetime, not just from the first sale. It's the dominant model in subscription SaaS and gaming. The commission per affiliate is lower in initial absolute value but grows with customer retention.

Some less common variants: CPC (cost per click, rare in modern affiliate setups), hybrid models (CPL + a sale bonus), performance-tier models (more commission the more volume you drive).

The choice of model dramatically affects what kind of affiliate you attract. CPS attracts disciplined affiliates with buying audiences. CPL attracts traffic affiliates in general. RevShare attracts long-term partners willing to bet on the product's retention.

The dominant networks in 2026

The affiliate marketing ecosystem has consolidated significantly over the last ten years. The main players:

• Amazon Associates -- still the program with the most active affiliates in the world, although its commission rates dropped significantly between 2019 and 2020.
• Awin -- the group that acquired ShareASale in 2017 and absorbed Affilinet after the merger. It is especially dominant in the European market.
• CJ Affiliate (Commission Junction) -- one of the oldest networks, strong in the U.S. market, with large brands.
• Rakuten Advertising -- the rebrand of the former LinkShare, especially strong in premium and major consumer brand programs.
• Impact -- emerged in the 2010s with more modern technology and has captured much of the SaaS and D2C market.
• PartnerStack -- specialized in B2B SaaS partnerships, with a SaaS-friendly model.
• Refersion -- popular with mid-sized D2C companies that run their own programs.

Beyond the networks, many brands run in-house programs with their own technology infrastructure (Shopify offers modules for this, as do platforms like FirstPromoter or Tapfiliate). The trade-off is: networks give access to an existing affiliate base but charge an additional commission; in-house programs give more control but require the brand to recruit its own affiliates.

The cookieless transition: the model's biggest change in 25 years

Affiliate marketing was technically built on third-party cookies. When someone clicks an affiliate link, the browser stores a cookie that associates that visit with the corresponding affiliate. If the person buys within the "attribution window" (typically 24h, 7 days, or 30 days), the cookie makes it possible to credit the sale to the affiliate.

That mechanism is degrading rapidly:

Safari has blocked third-party cookies by default since 2017 with ITP (Intelligent Tracking Prevention).

Firefox has blocked them by default since 2019 with ETP (Enhanced Tracking Protection).

Chrome, after several delays, announced in 2024 that it was abandoning the plan to eliminate third-party cookies entirely and moving to a user-choice model. Even so, in 2026 the effective restrictions are high and the trend toward less tracking continues.

iOS App Tracking Transparency has significantly reduced cross-app attribution since 2021.

The affiliate marketing industry has had to redesign attribution. The solutions that are consolidating:

Server-side tracking. Attribution doesn't depend on the browser but on events the website sends directly to the network's server. More robust, but requires technical implementation.

First-party cookies with redirects. Some networks redirect the click through the advertiser's own domain, avoiding the third-party cookie category. It works but adds latency.

Unique discount codes. Each affiliate has a code the customer enters at checkout. Direct attribution, no tracking. It's the model that best survives the cookieless era, especially in e-commerce and D2C.

Direct attribution via partner link. When there's a contractual relationship with large affiliates (major publishers, comparison sites), sales information can be shared directly without relying on cookies.

Models based on the advertiser's own data. If the customer has a CRM account and is logged in on the advertiser's website, attribution can be done server-side without involving third-party tracking.

The practical consequence is that affiliate marketing in 2026 is operationally more complex than in 2018, but also cleaner in terms of privacy. Programs that rely exclusively on third-party cookies are losing effectiveness; those that have migrated to modern models keep working well.

When affiliate marketing is the right play

There are contexts where the model fits particularly well:

E-commerce products with an average ticket in a comfortable range ($50-500). Below that, commissions are too small to incentivize serious affiliates. Above it, buyers research more and conversion via an affiliate link is low.

SaaS with high LTV and low churn. The RevShare model rewards retention, and affiliates specialized in SaaS tend to generate high-quality leads.

Products with obvious educational value that benefit from creators explaining them in depth. Software, productivity tools, books, courses.

Categories where social proof and recommendation carry a lot of weight. Cosmetics, supplements, technical equipment, financial services (with regulatory care).

Companies with a good CRM and analytical capacity. A serious affiliate program requires lead-quality validation, fraud detection, and per-affiliate optimization. Without that operational capacity, the program attracts a lot of noise.

When it's not the right play

• Products with very low margin. If your gross margin is 15%, paying a 10% commission leaves very little to cover operating costs.
• Early-stage brands with no clear reputation. Affiliates promote products with already-validated conversion better; convincing them to promote something unknown usually requires very high commissions that wreck unit economics.
• Highly regulated categories (strict financial sector, health, alcohol, gambling) where the affiliate's communication can create legal risk for the brand.
• Products that require complex explanation in consultative selling. The typical affiliate can't consult; a human sales channel is more appropriate.

Operational risks of the model

Fraud. It's one of the central challenges. Common types: click fraud (bots clicking links to inflate traffic), cookie-stuffing fraud (inserting cookies without the user consciously clicking), self-purchase fraud (affiliates buying their own products to collect commission, a detectable but annoying technique), and return fraud (purchases that are returned after the commission has been paid).

Traffic quality vs. volume. An affiliate that brings a lot of low-quality traffic (high return rate, low LTV) can be harmful even if their commission per sale seems acceptable. Validating quality per affiliate is necessary discipline.

Brand risk. What the affiliate says isn't what the brand says, but the customer associates the two. Affiliate programs without clear guidelines produce contradictory or problematic messages associated with the brand.

Attribution vs. cannibalization. If a customer who was going to buy anyway passes through an affiliate link at the end, the brand pays a commission for a sale it already had. Removing commissions for these sales requires serious incrementality analysis.

Regulatory compliance. In 2026, regulations on affiliate disclosure (the FTC in the U.S., equivalent European directives) are strict. Affiliates who don't clearly mark their links as sponsored generate legal risk for the brand as well.

Affiliate marketing and creative operations

For a brand that runs an affiliate program, producing assets that affiliates can use is regular creative work: updated banners, product images, copy for landing pages, demo videos, case studies, email templates. Without a coordinated creative program, affiliates create their own materials that may or may not be aligned with the brand.

That coordination fits into creative operations: the editorial calendar coordinates the production of assets for campaigns and key moments in the program, brand management defines what affiliates can and cannot do with the brand, and content production sustains the regular creation of materials that quality affiliates demand.

At Polimake, that logic lives on three surfaces: Studio to coordinate campaigns with specific affiliates at moments on the calendar, Studio to produce banners, videos, copy, and consistent brand templates, and Media as a repository where affiliates access up-to-date materials without requesting them by email.

---

If you manage growth, partnerships, or marketing at a company with a product that can be sold online and you've landed here looking for an answer about affiliate marketing, the most useful thing you can take from this article is probably the combination of three ideas: the model has been running for 35 years (it's not a recent invention), cookieless attribution has changed the rules (the programs that survive are those that have migrated to modern tracking), and quality matters more than volume (an affiliate program without per-affiliate quality validation is expensive noise disguised as a scalable channel).

To round this out, direct advertising covers the broader context of performance marketing where affiliates fit, CAC as a diagnostic covers how to evaluate the real profitability of affiliates versus other channels, and LTV covers the counterweight that decides whether RevShare models are sustainable.

Quick references

• Direct advertising -- the broader context of performance marketing.
• CAC as a diagnostic -- to evaluate affiliates as a channel.
• LTV -- key for RevShare models.
• Conversion funnel -- how affiliates connect with the rest of the funnel.
• Lead scoring -- to validate the quality of affiliate leads.
• Smarketing -- the sales-marketing alignment that a serious affiliate program needs.