View-through Conversion

In a calculation that counts clicks, an ad nobody clicks scores zero forever. People don't behave that way. Seeing a banner, scrolling past it, and typing the product name into a search box two days later is ordinary.

Pagedeck found the gap almost by accident. Ravi ran video and banner placements for two months, saw almost no clicked signups, and switched them off at the end of the month. The following month, visits from people searching the product name directly fell 18 percent. Until those two numbers were placed next to each other, that campaign was on record as having produced nothing.

Keep a record of the ad appearing on screen, and if that person converts inside a set window, credit the ad without a click.

1. Define an impression first. Count a banner that passed through the bottom of the viewport for one second and the number inflates; count only cases where more than half the ad was visible for over a second and it starts to resemble something a person saw. Without that written down, the platform's number and yours differ by more than a factor of two.
2. Keep the window short. One day is typical for a view; stretch it to a week and people who arrived for unrelated reasons flood in. The value gets set alongside the click side in the attribution window.
3. Report it separately from click conversions rather than summing them. The two numbers mean different things, and somebody who saw one ad and later clicked another can land in both columns, so adding them overstates the total.
4. Verify by switching things off. Split by region, take the ads down in one, and compare direct traffic across the two. Counting impression records alone proves no causation, because the people who saw the ad were selected by the platform for being likely to convert anyway.
5. That makes this a weak basis for moving budget and a reasonable device for not recording unclicked ads as zero. Teams working only from last-touch attribution, where brand advertising always finishes last, use this as the counterweight.

"Click conversions are 3. With view-through it's 41."

Put both numbers side by side and read the impression's share too.

"A seven-day view window means most of that is coincidence."

The window is long enough to absorb unrelated conversions.

"Split by region, switch one off, and watch it for two months."

Establish the effect by comparison rather than by impression records.

When Ravi rebuilt the reporting template, it got three columns: click conversions, view-through conversions, and visits from people searching the product name directly. The third column went in because the first two are numbers the platform supplies. In the month the banners went back up, that third column rose with them, and that was the argument for continuing — one number from outside the platform carrying more weight than the two inside it.