Problem 01
I was wrong about my own users
My hypothesis was elegant, humane, and completely false.
I argued for the polite version. Put the signup inline in the sidebar, let people find it, and you self-select for genuinely interested users — a smaller list, but a better one that converts harder later. The intrusive modal, I said, buys volume from people who never wanted it.
It is a reasonable argument. It is the argument most designers make. I ran it as a proper A/B test across markets because I wanted the data to back me up.
It did the opposite. The blocking modal beat the inline widget by 400% in France, and the pattern repeated everywhere we ran it. The polite variant was pulling a trickle; the modal was pulling dozens a day in the same markets, on the same traffic, with the same creative.
The uncomfortable part was the downstream data. The "low-quality" subscribers the modal brought in did not behave like tourists. They opened, they clicked, they bought. My filter-for-intent theory had no support at all — the inline widget was not filtering for interest, it was filtering for whoever happened to look at the right corner of the page.
The test that changed my mind
Inline, in the sidebar
My hypothesis
Polite. Waits to be noticed. Only genuinely interested users sign up.
Single digits per day
Modal, over the page
The control
Blocks the page. Interrupts the task. The thing I argued against.
+400% signups (France)
The finding: the polite variant did not filter for intent. It filtered for attention — and almost nobody was paying any.