(When this guy looks straight down his face looks almost like a funnel.)

Fun With Funnels

Improving the IntroCave Sales Funnel

One of the projects I've been working on since taking over IntroCave is cleaning up the analytics a little bit (the site uses mostly Google Analytics along with some homebrew dashboards). There was rudimentary goal tracking in place already, but the sales funnel was itty bitty—the only two steps were Users and Sales... which means I could tell that roughly 99.4% of site visitors came to the site and didn't buy anything, but I couldn't tell you much more than that. (A separate dashboard actually tracked number of previews which makes for a pretty good middle tier, but I'm specifically talking about tracking funnels in GA where the data source is the same.)

I ran a few A/B tests (using Google's Optimize service), which increased both page views and total sales... but it was hard to see the impact on conversion rate. 0.6% is a pretty small number and moving it signficantly in any direction takes a LOT of data. I now track 7 steps in our sales funnel:

1. USERS: how many users visit IntroCave (typically tracked daily).
2. INTRO: how many users visit an intro page (you can't very well purchase an intro video if you don't find one you like.
3. CUSTOMIZE: how many users start customizing a video using our free intro video preview tool?
4. PREVIEW: how many users finish customiznig a vidoe and wait to view the free preview?
5. ORDER: how many users click on the "download" link after viewing their preview video?
6. CHECKOUT: how many users choose their download settings and visit the checkout page?
7. COMPLETE: how many users actually purchase a video?

Here's a quick visualization of how those tiers currently break down over an arbitrarily-chosen chunk of recent time:

A few things jump out.

USERS -> INTRO PAGE: -30%

Around 30% of users never visit an intro page. This is a pretty good proxy for the site's bounce rate and is actually pretty good! I think we could get back a few percentage points by having a deeper catalogue of intro videos or offerering free intro videos (at lower render quality), but in the overall funnel there are bigger fish to fry.

INTRO PAGE -> CUSTOMIZE: -48%

This is a pretty big drop for being so high up in the funnel. I chalk this up to three major problems. Some chunk of those users are visiting on mobile or don't have their logos immediately at hand—there's not much I can do to capture those users. The second issue is overall page usability. The Customize button used to be lower on the page, and just by moving it higher I've already increased the percentage of users who click on it. I have a few other ideas for the page that I still need to A/B test but decided to put on hold until funnel tracking was in a better place. The third and possibly largest issue is actually sticker shock. In the current flow this is the first page that puts the prices front and center, and there are a large number of users who will see those prices and bail out immediately.

I indirectly wrote about this discovery a few weeks back. Removing the prices increases the number of previews created by a whopping 40-50%, but my current render servers actually can't handle that much load. At some point after I get the render servers settled I'd like to experiment with offering free renders. I don't know exactly what shape that might take yet, but—if I can figure out a way to do it that doesn't cost me a ton of money in render costs and lost sales—I think there's a lot to be gained from the goodwill that comes from offering a useful service.

CUSTOMIZE -> PREVIEW: -25%

I'm pretty happy with this step. The customization flow is a multi-step process. I may tinker with condensing this to a single step, but I think the majority of the drop-off here is users who don't have the proper logos at hand or don't want to think up the text that fills the text areas.

PREVIEW -> ORDER COMPLETE: -63%

In writing this post, I actually just realized that I missed a step. Users get to the order page after the video is done rendering, but I'm not actually tracking that as a step in the funnel right now. Anecdotally sales are much worse on days where the servers have hiccups or get too bogged down. I don't know how close to 0% I can get this number (the render time is never going to be zero seconds, and there are always going to be users who can't or won't wait the 2-3 minutes it takes to render a preview video), but this step in the funnel seems like something I can move dramatically with improved reliability on the render servers.

There's another big test that could be run here, but it seems a bit crazy—do users actually need to see a preview before making a purchase decision? Server costs aren't so high right now that I'm champing at the bit to test this, but if time-to-preview is really the limiting factor in this step of the sales funnel it seems logical to try it out. Another perhaps-more-challenging test would be to render just enough keyframes in our templates to show a preview sequence.

ORDER -> PAYMENT: -94%

I'm actually squashing two steps together here that are both around 75%. Looking at published abandoment rates show that the average is somewhere in the 70% range on typical ecommerce sites. IntroCave doesn't have a true "shopping cart" (most visitors only need a single intro video), so I'm not sure how comparable the stats really are... but I feel like a tremendous amount of work would be needed to push these two steps down into the 70% range. That doesn't mean it's impossible, but for the moment my focus is on expanding the top of the funnel.

Closing Thoughts

While overall visitor-to-conversion-rate is still an important metric to keep an eye on, I'll now be able to run A/B tests with an eye at improving specific steps in the funnel instead of only considering the bookends.