Sequential Liability

TripleLift is an exchange that helps various other parties transact programmatically. While our high-quality ad formats, non-intrusive user engagement, high quality publishers, and similar attributes are often the focus of our discussions with our partners in the ecosystem, broad industry trends are often shaping the discussions. The ad tech industry continues to push for “marketplace quality” - as evidenced by trends like TAG, mitigating ad fraud, and ads.txt, contracting around these points has significant impact on the economic relationships between parties.


In the halcyon days of early RTB, before ad fraud had risen to the level that it is today, the ad exchanges operated through the principle of caveat emptor (buyer beware) - meaning that once a buyer submitted a bid for an impression, if they won the impression they have to pay for it. So the buyer needs to be conducting its "diligence" on the impression before submitting the bid. This principle is used effectively in real estate, where the buyer has access to ensure the seller really owns the property and that there are no material defects - and the buyer can bring in 3rd party experts to verify the claims. Because of the nature of ad fraud and the relationship between buyers and sellers in the context of RTB, caveat emptor has slowly given way to various forms of sequential liability.

In its purest articulation, sequential liability means that the entity that needs to pay for a purchase is only obligated to do so to the extent that it has received payment from its own customers relating to the same purchase. This is a meaningful change from caveat emptor, especially if one considers that the buyer doesn't necessarily even need to bill its own customers based on certain exemptions (or the downstream customer is exempt from paying contractually based on certain carveouts), thus these exemptions will never be paid for.

The impact that sequential liability has can be considered in the context of a hypothetical car dealership. In the model of caveat emptor, the buyer walks into the showroom, has ample time to inspect the car and make their own purchase decision, and then ultimately negotiates some price with the car dealership and drives away with the new car. If the buyer gets home and realizes it's not a car, but a lemon, tough bananas. The upside is that this is predictable for all parties - the downside is the risk of lemons, meaning each buyer would theoretically price in some risk factor (states often have lemon laws for this precise problem).

In the world of sequential liability, things get significantly more complicated. Take the same car dealership, but imagine the buyer is actually an agent working on behalf of a mysterious, wealthy customer who never shows their face in public. The mysterious customer and the agent have a contract that stipulates the customer will pay the agent based on the delivery of a car with certain qualities, and with certain stipulations. The agent and the dealership then enter into a contract that says that the agent will pay the dealership upon receiving money from their customer. The dealership has already purchased the car from the manufacturer, and has to pay the manufacturer within 30 days of a sale (not how things work in reality). When the agent buys the car, the dealership's clock starts - it has to pay within 30 days. It also has to pay its sellers their commissions and all related costs. The mysterious customer receives the car and decides the engine doesn't meet the stipulations of the contract, and thus returns and doesn't pay for the engine. So the agent receives 90% of the purchase price from the buyer 90 days later, takes their cut, and pays the rest to the dealer.

The complexities introduced include: 1) the dealer has a fixed payment to its supplier and employees, but a conditional payment from its customer, 2) the dealer is, to some degree, operating at the whim of its customer and may not know on any given transaction whether it will get paid (it does not necessarily see the contract between the agent and the mysterious buyer), and 3) what was once a predictable float period (the time between paying and getting paid) becomes unpredictable (if the mysterious buyer declares bankruptcy between the purchase and payment times - the dealer might never get paid). In response, the dealer will take on the responsibilities of understanding the needs of its upstream customers and the contract stipulations that could result in the buyer not wanting to pay. It will begin to inspect the cars it receives from the manufacturer before obligating itself to pay for them, and thus change its relationship with the manufacturer. The dealer will also change its relationship with its sellers - to perhaps only paying once the dealership receives payment. It will need to pay for insurance (implicitly or explicitly) against bad sales like above or buyers simply failing to pay their agents, which in turn will be priced into the cars itself.

The advertising technology ecosystem is evolving from caveat emptor to sequential liability. The payments flow largely as shown below:


Two pressing changes in the ecosystem are ad fraud and viewability. In both cases, media agencies may - on behalf of the advertiser - decide they will not pay for fraudulent or non-viewable impressions ("invalid impressions" for the sake of this Lift Letter). Whereas previously all invalid impressions would be paid for, and one could assume the likelihood of an invalid impression would be priced in, now the ad exchange may not be paid for them. DSPs, trading desks, and media agencies are variously demanding contractual provisions that they not pay for these impressions. In turn, the sensible result would be for the exchange to similarly not pay publishers for the impressions. But things aren't so clean in the real world. Different agencies have different measurements for what counts as an invalid impression, and sometimes the publisher setup doesn't allow for it to not be paid for an impression, even if it is subsequently determined to be invalid (this is especially the case for impressions purchased via Google's EBDA - a major challenge in working with them).

Exchanges are now in the position of understanding what constitutes an invalid impression for what buyers and ensuring they do not purchase them, to the best of the exchange's abilities. Arguably, this is a good outcome. Invalid impressions will become less valuable to the extent that sequential liability truly flows to the publisher, decreasing the economic incentive for any actor in the value chain for invalid impressions. The value of valid impressions will increase, as the risk factor for invalidity has been removed from the buying entities. The challenges appear, however, when not all parties have sequential liability (e.g. EBDA) or when different parties have varying and inconsistent definitions of invalid impressions. This leads to a situation, just like the car dealership, where the cost of "insurance" against buying invalid impressions, as well as buyers going bankrupt, gets baked in to the cost. But overall, the quality of goods in the ecosystem should improve. The hardest party to convince of the merits of sequential liability, however, will be the publisher - which is the challenge faced by exchanges.

By Katie Glass