/conv/ - Conversion Rate

i stopped using placeholder text as labels because it's total garbage for accessibility once someone starts typing. i always keep the label visible above the box so people don't lose context mid-flow. anyone else still relying on placeholders for simple forms?

just saw something about how 5th-7th grade reading levels convert at like 11.1%. it's basically just saying don't use big words, which i've known for years.
>small, evidence-based improvements
it's a nice way to rebrand the same old a/b testing we've been doing since forever. calling it an "ongoing cycle" doesn't change the fact that most of us are just fixing leaks as they happen anyway. marketing gurus love new names for old habits i don't see how this stops you from swinging for the fences and failing, though. does anyone actually have a workflow for this that isn't just a mess of spreadsheets?

more here: https://unbounce.com/conversion-rate-optimization/iterative-testing/

tried removing it but users keep scrolling back up to check their subtotal.
>the extra screen real estate is worth the distraction
it's killing my conversion rate because they lose track of the actual buy button lol.

i hate seeing clients run a single variant across both desktop and mobile when traffic is lopsided like this. if you don't separate your segments it's basically just noise that hides the real winners. i saw one test last week where mobile was tanking but desktop was up 10% and the aggregate result looked flat. has anyone else had to kill a winning test because of bad segmentation?

found this thing about 'ask advisor' filtering out all that junk. it's wild that 40% of traffic is just bad bots now, especially with cloudflare seeing 50 billion ai crawler requests a day. i'm skeptical about how much this actually changes my actual test results or if it's just cleaning inflating the numbers.

found this here: https://conversionsciences.com/filter-bot-traffic-google-analytics/

Using AI to skip designers for mockups sounds like a recipe for disaster.
>I can generate a mockup with AI and show the client exactly what I mean
It's fine for a quick visual, but how do you handle the actual implementation? If the final build doesn't match the AI-generated version, you're just setting yourself up for false expectations with the client. Plus, I'd need to see some data on whether these fast-tracked tests are actually running long enough to reach significance.

article: https://www.convert.com/blog/optimization/think-like-cro-pro-buse-hizarci/

another jungle scout piece talking about how 73% of people click the top result. basically just common sense. they're pushing this idea that you don't need to be twice as good as competitors, just slightly better through small tweaks. i've seen those 'bundle changes' tactics before and it's a mess for data integrity. if you're testing too many things at once on amazon's built-in tool, the results are useless. ime, waiting 2-3 months for an experiment to finish is just a massive waste of time. harder to scale when you can't even get a clean test result without waiting a whole quarter.

https://conversion-rate-experts.com/amazon-cro-playbook/

I'm looking at this one client's landing page right now and it's driving me crazy. They added a sticky nav with a big CTA button that stays stuck to the top of the viewport while you scroll through all those long feature descriptions. It looks okay for branding, but every time I check the heatmaps, people are constantly clicking it by accident when they just wanted to click a product detail expander. It's creating a huge mess in our click event data. We're seeing a massive spike in "navigation" events that don't actually lead to any real intent. The conversion rate on the main CTA looks high, but if you dig into the session recordings, half those clicks are just frustrated users trying to interact with the content underneath. I've seen this happen before where a clunky UI ruins your ability to see what's actually working. It makes it impossible to tell if people like the offer or if they're just bumping into the header. We might have to strip it back to a standard static header for the next test. **The click map looks like a crime scene

i've been thinking about that mckinsey stat about personalization driving 40% more revenue because it feels so circular like we're just seeing the result of having a massive budget for data. it's easy to get caught up in segmenting everyone but if version b wins for both new and returning users then personally there's really no reason to build two separate flows. i wonder how often we're actually wasting dev time on custom experiences that don't even change the underlying decision?

found this here: https://wingify.com/blog/the-crossover-test/

i'm seeing this massive jump in conversion on the control but i can't tell if it's just because of the holiday promo or the actual variant. we've been running the test for three weeks now and the delta is still moving around way too much to trust.
>the conversion rate is up 5%
i tried splitting the sessions by device type to see if it was a mobile-only thing, but it didn't clear anything up. i'm worried we're just looking at seasonal noise instead of a real effect from the new layout.