Conversion funnels that optimize themselves | Hacker News Reader

Conversion funnels that optimize themselves

subintent.com

33 pointsadityar19 comments

This is a very interesting idea because it drastically lowers the bareer to entry for someone wishing to try A/B testing.

I think business intelligence should be part of any company's core competences, but there is a market for this kind of automation too.

Some more general feedback:

   - big call-to-action link in the middle of the top goes nowhere!
   - better styling everywhere 
   - pricing should fit on one screen!
   - marketing: focus on GET RESULTS, and IMPROVE CONVERSIONS
     and less on the technical side---my guess is that your 
     main customers will be business people not technical folk,
     so talk about the RESULTS and not so much about the tech.
     keywords: ML, OPTIMIZE, BETTER, AUTOMATED, SMART, etc...
   - who are you targeting with this pricing?
     I think if your product really works it is delivering the 
     same value as a full time employee doing A/B testing and
     optimizing the website for you --- so you should charge more.
     Also -- the ppl who are most likely to benefit from this 
     are high traffic websites (1000+ visitors for stat. signif?)
     so you should focus on them as your target market. 
     I think it will be much better to get 2-3 BIG customers than
     chase mom-and-pop and their little website. 
   - maybe have a demo? (tell user to keep reloading a page/iframe)
     so they see the different versions.

On a separate note, I think the notion of an "interactive" sales pitch that is custom generated for the user is interesting in its own right. Check out the link to my book on my hn profile page --- I have little conversation-like bubbles which add a bit of interactivity. This concept could go so much further though... Think "webpage as a conversation."

Good job on the coding. Now go open the right .css file and make your site look pretty ;)

since I'm assuming they're eating their own dogfood isn't it just a matter of time before the page morphs into a version that addresses all your points? The big call-to-action that goes nowhere will be tested out... maybe the pricing page you get that doesn't fit on one screen is just one version that will get phased out :)

Maybe I'll check back in a few weeks, and if the site has morphed into the best sales page ever then it means the tech works

Thanks for the feedback. My initial target is small/medium web app developers and not so much the big guys (yet). At this stage, some A/B & design saavy is required and not fully eliminated. There's a lot of demo possibilities (one that comes up very often - eat your own dogfood much?) which I think will improve conversions.

This is a perfect example of building a technology, not a product. This is Optimizely without the amazing user interface that allows marketers to do this work.

is that good or bad?

I don't get your site. There's a BIG green button at the top and when I click it nothing happens.. Then there's an empty crystal ball and below that two other buttons that actually do something.. I have to click again to at least get a very basic idea on what you do.

Besides that the project actually sounds quite interesting!

I really should make that link go to a page that says "See what I mean?"

With this multivariate testing, how many visits does it take to get statistically significant data?

Calculating the number of tests needed to reach statistical certainty is not straightforward at all. The equation for simple A/B tests relies on the change of conversion rate from a control. More info here: https://developer.amazon.com/sdk/ab-testing/reference/ab-mat...

That's for simple A/B tests - this is not simple A/B tests, instead it is setting up a search space of possible variations of a page and comparing many to many to find which is best.

Splitting the data based on visitor referral, country, and more make the sample size needed even larger.

TLDR: This solution needs a decent amount of traffic to have reasonable results. The exact numbers are not definite.

Source: I co-created xander.io which is basically this concept but more open source.

that does look interesting.

Depends on the tests. However a point to note is that this is not classical testing followed by hard-coded values. It's more like a multi-armed bandit-type optimization.

Yeah, it sounds like the perfect kind of Q-learning problem.

are you using this to optimize your own site? if yes, do an open case study to show its power

if not, why not?

Came here to ask this. Your call to actions, etc are kind of lack-luster (although I have been waiting for a product like this for quite some time).

Are you optimizing your own funnel(s) with this? Please tell!

product is in development and testing at this time so,no. But, that's going to be a persuasive argument - we eat our own dog food.

wait, so you are not willing to use your own product now but you expect others to pay for it and use it now?

Site is off =(

This Site44-hosted website has been temporarily suspended due to excessive usage.

now fixed

down :(

Reply on news.ycombinator.com