Field notes · 12 Feb 2026

Waterfall myths that still show up in weekly reviews

Waterfalls did not vanish when bidding arrived. They thinned, hid inside “hybrid” setups, and left behind a set of sentences that still get said in Monday reviews as if they were physics. This note is the version we use in Mediation Stack Literacy and in the fourth module of the Revenue Signal Lab.

Myth one: a higher floor is a strategy

A floor is a filter. It says which demand you are willing to hear. Raising it can make eCPM look composed while fill quietly leaves the room. In Thailand evening hours we still see teams celebrate a floor change that coincided with a session dip. The chart was prettier. The brief was wrong.

In the lab we date every floor. If you cannot say when it expires, it has become folklore.

Myth two: more instances mean more coverage

Instance maps grow because someone was afraid to delete. Duplicate instances for the same network and format do not create new advertisers. They create race conditions and reports that no longer match the UI. Coverage is a demand question. Instance count is an ops question. Mixing them is how weekly reviews run long.

Myth three: the waterfall is “done” once bidding is on

Many stacks in SEA still keep leftover waterfall instances for geos or formats where bidding match rate falls off a cliff. That is not failure. It is a sentence you should be allowed to say in a product review. Auction vs Waterfall Decisioning exists because teams were deleting those instances to make a slide simpler.

What we ask students to bring

A week of instance-level revenue, not a network rollup. If your vendor only gives the rollup, say so early. The myth-busting still works; the homework just gets slower. We would rather you leave with three dated floors than with a new religion about waterfalls.

If this is the argument you are currently having, sit Revenue Signal Lab or write the desk from contact.