High bounce rates consistent with a login that takes you to a 3rd party site?

[TheKatzMeow]

My firm has a credit union client whose bounce rates skyrocketed after implementing an online banking portal. Logging in to the online banking portal takes you to a 3rd party site. Would arriving at the site and immediately logging in be considered a bounce? And if so, would a high bounce rate actually correlate with a warm reception to their online banking tool?

[TheKatzMeow]

Paul,

Let me just say thank you for a thoroughly informed and thoughtfully written response. It gives me everything I need to guide my client. Really appreciate you taking the time.

[ThompsonPaul]

The technical definition of a bounce in Google Analytics is that the visitor arrived at the page and triggered a hit to the Analytics tracking code and then left the page without triggering any further interactive hits to the tracking. That's exactly what's happening when a visitor comes to the primary site's page, then leaves it to go on to the new 3rd-party page that doesn't contain the same tracking code.

The critical question here then is "would a high bounce rate actually correlate with a warm reception to their online banking tool?" Unfortunately, in GA's default tracking mode, there's no way to know definitively whether a visitor bounced because they didn't get what they wanted and left, or if they got exactly what they wanted and went on to the online banking tool. And ideally, we want more than just correlation, we'd like reliable data. Knowing the difference between these two site behaviours gives us the critical understanding of whether those bounces are happy visitors or not. The solution is going to require some enhancement of the default tracking.

Fortunately, once you understand that definition of a bounce above, the possible solutions to your challenge become more clear.

The ideal solution would be that the 3rd-party site has a system to allow you to add your own site's Google Analytics tracking code to your pages on their site. You could then set up cross-domain tracking, and your GA would then fire again on the 3rd-party site visit so you'd know for certain they got there and you'd be able to report on the users' behaviour as they went from your site to the other one and back again. Many of the better 3rd-party tools offer this if you check, as it's a common need.

If that's not possible, we still need to understand the behaviour of the visitor during the course of the bounced visit on the primary site. The way we discover and track user behaviour within a page (as opposed to when moving from page to page) is by adding Event tracking to the actions a user might take that are valuable or of interest to us. In this case, we can add event tracking to the link click that takes the visitor off to the 3rd-party site. That way, we'll know that for every bounce visit that also had that click event associated with it, we had a "warm reception", rather than an annoyed "I came, I puked, I left" visitor (to quote Analytics expert Avinash Kaushik).

And in the definition we have another clue - if a bounce is a visit with only one hit to the tracking code, we can decide that a successful one-page visit shouldn't be counted as a bounce. To do that, we can specify that the event tracking click should be classified as interactive and count as another hit to the tracking code. This means single pageviews with such clicks will no longer be counted as bounces. This is just a semantics change - we will already be able to tell in the GA reports that the event occurred. But if the data model for the site is that bounces are bad, this will make the data more properly reflect that this bounce visit shouldn't be counted as bad.

Final note - if you're not able to get your own tracking added to the 3rd-party site, you're going to need to add that site to the primary site's Referral Exclusion list. This way, if/when a visitor comes back to the main site from the online banking tool, they'll essentially "pick up from where they left off" in their original tracking, instead of just being (artificially) counted as a whole new session with a referrer as the banking tool site, which would badly skew the accuracy of your data.

Hope that all makes sense?

Paul

[Sidenote: This is why average bounce rate for a site is a totally worthless metric. Some pages are actually most successful if they are bounces - i.e. the visitor got exactly what they wanted in a single pageview. Bounce must always be segmented by page (and even by location/mobile/desktop etc) to avoid incorrect assumptions. And if you add behavioural understanding through event tracking, you can get a much fuller picture of what's actually happening during those bounces and where optimization is needed.]