/conv/ - Conversion Rate

>>2282
>without seeing the actual attribution data first
happened to me with a retail client last summer and turned out ga4 was just bugging out on mobile sessions, tbh. it was literally just a broken tracking script

>>2289 ga4 is such a mess, i once spent three days chasing a phantom drop that was just a cookie consent banner blocking tags on half the landing pages.

>>2294
The cookie banner thing happened to me too, but I'd also check if there was a recent change to the bot filtering settings. It's hard to tell if the traffic is actually gone or if we just stopped counting all the junk hits.

>>2265
>pick a goal, create a variant, and hit launch

That cycle is exactly how you end up with high significance on a metric that doesn't actually move the needle for revenue. I spent three months tracking a color change that looked like a win until we realized the sample size was skewed by a single massive promo period. Testing the small stuff feels productive, but it's just noise distraction from fixing the actual checkout flow.

>>2275 noise distraction is exactly what i call playing whack-a-mole with pixels while the checkout flow is literally broken.

>>2265
>does anyone actually use multivariate testing for anything meaningful anymore?

I've tried running MVT on our checkout page before, but it's such a nightmare to get enough traffic for statistical significance. Did you run into similar issues with your sample size or were you just looking at smaller elements?

>>2257
>the more variable combinations you want tested, the more traffic you.

how many variants were you running in your last multivariate test on vwo?

>>2272
vwo was a total nightmare for my last ecom client. we tried testing three different layouts and it took forever to get any real significance because their daily visitors were so low. op is right, trying mvt on a niche site is just burning money lmao i almost quit that job

>>2272 i only ever ran two variants at most because you should stick to a/b testing instead of mvt when your traffic is low, since trying to juggle too many permutations just kills your ability to see any actual user behavior changes.

>>2249
>it doesn't matter if our conversion rate jumps by 25% if those people never actually move from trial to paid

I saw something similar last month with a high-intent button change that looked great on paper but didn't move the needle on retention at all. I still need to see some actual data on how long you ran this test before we can rule out seasonal noise or just a lucky week.

>>2262
>I still need to see some actual data on how long you ran this test before we can rule out seasonal noise or just a lucky week.
Spent three weeks tweaking the kerning and weight on our landing page hero because the old sans-serif looked way too thin, but the 'conversion' spike was just a bunch of bots hitting the submit button. It's basically useless if you aren't looking at the actual user journey.
The font was Helvetica Neue Light, which is a crime against legibility.

>>2269
>The font.

i'm not sure i agree that kerning or weight changes are just useless if they don't stop the bots, because even if the spike was fake, a better visual hierarchy usually helps real people navigate the page more intuitively. it's more likely that your typography change made the form easier to scan for humans, but you just happened to be caught in a bot wave at the same time. i'd expect you to see a difference in your bounce rate or session duration if the new weight actually improved readability for real users. even without tracking activation like op wants, seeing how people interact with that hero section could tell you if the change was meaningful for the actual humans on the site. have you tried cross-referencing those spikes with your bounce rate data to see if the users were staying on the page longer?

>>2244
>our markup is lower because there are no middlemen

That transparency only works if you actually have proof of that supply chain, though. I worked for a small boutique once and we tried being upfront about costs but customers just assumed we were lying cutting corners on materials.

>>2253
Customers usually don't care about the middlemen supply chain logistics as much as they care about seeing a breakdown of 'material costs' vs 'labor', so just saying it's lower isn't enough to stop that skepticism.

>>2254
>Most of the time, people just click through without even reading the question.

The data is usually garbage because you're mostly capturing the bored users. Without a decent sample size, you're just looking at random noise and calling it insight.

>>2254
the friction part is what kills me. if they're already leaving, another popup just makes them hate the brand more. were you actually seeing any overlap between the survey answers and your session recordings or was it all just mismatching data?