Google Analytics, Mixpanel, Optimizely...

The broken or misleading metrics

Things that go wrong when general analytics is applied to conversion optimization.

Conversion funnels

The conversion funnels in general analytics are really nothing more than a bar charts with a flow-like look and feel.

Event funnel in Heap analytics

This funnel just shows the ratio of different event types. These manually configured steps have little (or nothing) to do with conversion optimization.

These steps won't tell you how people gradually build desire before taking action and they cannot uncover your actual engagement bottlenecks where most of your visitors are leaving your site.

However, they can help you discover bottlenecks in step-by-step flows or “wizards” where the order of steps is guaranteed.

Conversion rates

The general data model calculates conversion rates by taking a group of visitors within a time range and count how many of them had converted within the start and end date.

Time range takes a subset of events

This time range is essentially a snapshot: completely unaware of the events outside the range. You don't know whether the visitors are new- or returning, or whether they had already converted.

Time ranges make website optimization practically impossible.

For example, when you acquire new visitors, your conversion rates will drop no matter how qualified the new visitors are. That's because you now have more visitors in the early stages of the conversion funnel. They are just building interest and not taking any action.

I have the perfect advice on how to raise conversion rates significantly. All you have to do is stop marketing.

Jared Spool

Founder of User Interface Engineering

The correct way to measure a conversion rate is to take a group of new visitors, wait for them to do all their actions, and see how many had converted. This way, you can isolate the people from a marketing campaign and see how they behave. You can also compare them to any other visitor segment, and see if they did any better.

You cannot rely on snapshots — studying what the visitors did over the entire visitor lifetime is the only way to acknowledge success or failure.

Time metrics

Time-on-page is the time difference between two page view events. It doesn't matter what happened between those two events: did the visitor pay attention to the content or was the visitor making coffee?

And if the visitor only viewed only one page, the time information is not available at all.

Time on page is always zero for single-page visits, and it is always zero for the last page visited.

A visitor can engage for several minutes, even hours on a long-form landing page, and your analytics won't notice any difference: the time-on-page is zero. *

Moreover, an inactive or hidden browser tab can sometimes track up to 30 minutes of idle time, causing the metric to go more off-sync with the reality.

This is why both time on page and time on site cannot be trusted.

Bounce rate

Bounce rate is the ratio of visitors who visited one page. *

This metric is problematic because it fails to tell whether a page is engaging or not:

First — No matter how much content is consumed or how much time is spent on the landing page, the visit is treated as a bounce if the user won't continue to a new page.

Second — A bounce is sometimes tracked as non-bounce. For example, when someone opens many tabs but only visits one. This kind of “page parking” is becoming a typical pattern, especially for millennials’ *

You can interpret bounce rate to support any argument you want

Jared Spool

Founder of User Interface Engineering

Learn Volument

{"desc":"Things that go wrong when general analytics is applied to conversion optimization.","title":"The broken or misleading metrics","url":"/learn/the-broken-metrics","key":"the-broken-metrics","created":"2019-11-12T07:59:24.668Z","modified":"2020-02-08T09:39:02.040Z","createdISO":"2019-11-12","modifiedISO":"2020-02-08"}