Why were there so many form tests?

Because the form is the only surface of the layer the reader sees. For someone who does not sign up, the flow, the metadata and the reply loop behind it have no equivalent at all. So rather than making quick decisions on the form, I took each step separately: where it goes, what it looks like, what it says. In the copy test, a variant carrying four changes recommended in the sources at once took the start of form interaction from 4.28 percent down to 1.14 percent, meaning what it lost was attention, not sign-ups.

What evidence did you base the form format decision on?

On session recordings and the measured use of the extra elements. In the recordings, visitors passed the boxed mid-article form without slowing down; the form was visible on screen, it registered in the shown event, but no interaction followed. In the same period the extra panel inside the form took 10 clicks across those same 5,000 impressions, while form interaction started at around the 2 percent band. That measurement on its own does not give the panel's contribution to the sign-up rate; for that, two variants, one with the panel and one without, would have been needed. But the extra elements clearly were not forming an interaction surface of their own, so there was nothing to lose in removing them. The new format was tried on a few of the highest-traffic posts first, then rolled out generally.

Why is the form at roughly 40 percent of the post?

The number did not come from a source, it came out of my own scroll measurement. After bot traffic is filtered out, the average depth real readers reach sits around that band. The form is placed just before the heading closest to roughly 40 percent of the article body's height, and it only opens when that heading enters the screen. The reader has taken in part of the content but has not left yet; putting it at a heading is so that the offer lands at the topic's natural pause rather than in the middle of a sentence.

How do you separate a real reader and their intent?

A page view on its own says nothing, because model crawlers fetch the page too. The distinction comes out of reading behaviour: the moment reading starts, time spent on the page, the depth reached, and whether there is a slowdown while passing the form. None of these signals occur in automated visits. I described how it is built in [Scroll Depth Tracking: Measuring Content Engagement Beyond GA4 and GTM](/en/premium/scroll-depth-tracking-content-engagement).

Does the light and dark theme comparison produce no result?

It produces no behavioural result, it produces an audience result. Almost none of the visitors ever change the theme, and the theme they see is the equivalent of their operating system preference. On desktop the average duration of dark theme sessions is several times that of light theme sessions, while on mobile the relationship inverts. A colour scheme does not have an effect that changes direction by device; what changes direction is the audience composition. That is why the theme is useful as a targeting and content planning dimension, not as an optimisation variable.

The Newsletter Form as an Interaction Layer: Decisions and Tests

The form is the only surface of the newsletter layer the reader sees. This post takes each step of that form separately: where it goes, what it looks like, and what it says. There are three tests, and in all three the decision came from measured form events and behaviour in session recordings.

TL;DR

The form is the only surface of the newsletter layer the reader sees, so the whole layer comes down to the interaction with that single box. This post covers the placement decisions and three tests: copy, format and theme.

Create Free Account Sign In