Five prelanders walked in, one paid the rent

The problem

This account was the opposite of the usual mess: clean whitelist, honest tracking, stable $0.55 CPM inventory in two Tier-2 GEOs, and a subscription content offer converting 0.9% from click to paid subscription. ROI held at 31%, real but thin, and every attempt to scale volume pushed the blended CPM up faster than revenue followed. The traffic was fine. The page catching it was the constraint.

The approach

We froze the media side completely: same zones, same bids, same caps for eight weeks, so the prelander ladder would be the only moving part. Five angles went live behind an even rotation in RedTrack, each a static page loading in under 1.4 seconds on a mid-range Android: a news-style advertorial, a three-question picker quiz, a price-anchored comparison table, a plain benefits page, and the wildcard, a deliberately sparse page built around one testimonial-style paragraph and a single button.

Judging was by verified subscription conversions per angle, pulled from S2S postbacks, with a minimum of three times payout spent per angle before any verdict. CTR was recorded and ignored: the quiz angle produced the highest click-through of the five and the second-worst paid conversion, which is exactly why ladders are judged at the payout event and not the click.

The result

The advertorial that convention said would win came third. The sparse wildcard won by a distance that surprised both of us: 2.2% click-to-paid once it took 100% of rotation in week six, at the same $0.55 CPM, moving ROI from 31% to 118%. Volume then scaled 40% inside the existing whitelist before CPM pressure appeared, because the offer could now afford bids the old page could not.

The four losing angles are archived with their numbers in the client readout, and the follow-up ladder starts from the winner's skeleton. That is the whole prelander service in one file: five opinions, one survivor, receipts for everything.