Don’t Just Look at the Change – Look at the Change in the Degree of Change

Comparison to previous yearview at full size

Use date comparisons to identify not just a change, but a change in the change.

You updated your home page at the beginning of April 2013; otherwise, you made no other changes to your website. When you check Google Analytics a month later, you see that the conversion rate for your primary goal was 10% in April.

You know that this number is not very meaningful if not compared to another time period, but which time period should you compare it to?

Compare it to March 2013?

You see that the site had an 8% conversion rate in March, so the home page update was effective, right? Well, not necessarily. If conversion rate increased by 2% from March to April during previous years as well, 2013 would only be following the seasonal trend.

Compare it to April 2012?

You see that the site had a 9% conversion rate in April 2012, so the update to the home page must have driven those extra conversions, right? Well, again, not necessarily. Conversions could have already been trending up during the previous months, so the home page update could have had no effect or even a negative one.

What you really need to compare is conversion rate over the course of March and April 2013 vs. March and April 2012. In this way, you’re not just identifying a change relative to the previous month or the same month of the previous year – you’re identifying a change in the change.

  March April
2012 8% CR 9% CR
2013 8% CR 10% CR

To complete the picture, we see that the conversion rate in March 2012 came in at 8%, just like March 2012. That means that the increase from March 2013 to April 2013 was 2% – 1% greater than the increase from March 2012 to April 2012.

Now you can celebrate – the home page update very likely increased conversion rate. (But next time, test the new home page in a Content Experiment to be sure!)

Please share this post:
Facebook Twitter Plusone Linkedin Email

The Ratio of Unique Visitors to Overall Visits Decreases for Longer Date Selections

Unique Visitorsview at full size

The proportion of unique visitors to overall visits decreases as you lengthen the data selection.

For almost all websites, the ratio of unique visitors to overall visits decreases as you display longer time periods in Google Analytics.

To illustrate, let’s say that:

• your website receives 100 visits daily

• each visitor visits only once per day

• 50 of the 100 visitors each day are return visitors

If you report for a single day in Google Analytics, unique visitors and overall visits will each total 100, or a 1:1 ratio.

If you report for two days, however, overall visits will total 200, but unique visitors will total only 150 for a ratio of 3:4, since 50 of the visits on the second day will be attributable to return visitors from the previous day.

For a website that had no return visits, this ratio would not change for a longer time period. Since almost all websites do, however, have return visitors, it is normal for the proportion of unique visitors to decrease relative to overall visits as you select longer time periods to display in Google Analytics.

Please share this post:
Facebook Twitter Plusone Linkedin Email

For Average Time on Site and Pages per Visit, Is High or Low Better?

Average Visit Duration and Pages Per Visitview at full size

Depending on the type of site, higher Average Visit Duration and Pages Per Visit may not be desirable.

For most Web metrics, it is clear which direction you want the trends to point towards. A higher number of visitors is nearly always a positive indicator, a higher bounce rate nearly always negative.

Average Visit Duration and Pages Per Visit, however, require more interpretation. For a news and education site, an upward trend for these metrics would probably be desirable, as they would suggest a higher level of user engagement.

On an e-commerce or lead generation site, however, these metrics could represent user confusion as much as user engagement – especially if your conversion rates and revenues are declining.

Before placing great emphasis on the high-level metrics that Google Analytics provides, make sure to understand what they really mean for your type of website, and evaluate them in the context of other less ambiguous metrics.

Please share this post:
Facebook Twitter Plusone Linkedin Email

New “Compare to Previous Year” Feature – Don’t Use It

Custom data range comparisonview at full size

In many cases, it’s best to avoid the Compare to Previous Year selector.

“Don’t use it” might be a little too adamant, but the new built-in Compare to Previous Year date-range selection may not always be your best option for comparing date ranges in two consecutive years.


For the majority of websites whose traffic is dictated by day of the week, you should instead apply a custom comparison to correspond with the weekly cycle.


This is not just for the benefit of the main over-time graph, which is much easier to interpret when days of the week are matching. Consider also that September 2012 included five weekends, while September 2011 included only four. If your traffic tends to vary significantly on weekends, a straight September-to-September selection without any offset will skew the comparison for any metric.


A previous post displays a date-range comparison aligned by day of week. This approach is still the most useful in many or most cases, despite the selectors for data-range comparison recently incorporated into the Google Analytics interface.

Please share this post:
Facebook Twitter Plusone Linkedin Email

Average Time on Page, Average Visit Duration, and Browser Timestamps

Browser history showing timestampsview at full size

Your browser records timestamps for page requests but not for page exits.

Every time you request a Web page – by clicking a link, accessing a saved bookmark, or entering a URL directly into the address bar – your browser records a timestamp, precise to a one-second increment.

It’s important to note that, technically, the browser does not record a timestamp when you leave the page. It does record a new timestamp for your next page request, but that timestamp is associated only with the page you’re accessing, whether or not it’s part of the same website, and not the page you’re exiting.

What this means is that Google Analytics can calculate Average Time on Page and Average Visit Duration based only on the difference between timestamps for successive page requests on your site.

Average Time on Page is calculated as the average difference between the request timestamp for that page and the request timestamp for the next pageview that occurred within your site. If only a single page is viewed during a visit, that pageview does not figure into Average Time on Page, since there is no second timestamp to subtract from.

Similarly, Average Visit Duration is calculated based on the timestamp difference between the first and final pageviews that occurred during a visit. Since the duration of the final pageview cannot be calculated, Average Visit Duration as reported in GA is always somewhat shorter than in actuality.

In an upcoming post, we’ll review the implications of timestamp-based metrics for a website, such as a blog, that normally experiences a high proportion of single-page visits.

Please share this post:
Facebook Twitter Plusone Linkedin Email

Align Date Comparisons by Day of Week

Date comparison offsetview at full size

Offset date comparisons to align days of the week.

For most purposes in Web analytics, the 31-day period most closely corresponding to October 2012 is not October 2011 but more specifically October 3, 2011 – November 3. 2011.

Because activity on most websites is influenced by day of the week, the one-day offset (or two-day for leap years) more closely aligns the time periods that you’re comparing.

As another example, July and August both contain 31 days, but July 2012 had 5 Sundays while August 2012 had only 4. If Sundays typically see the least traffic to your site, a comparison of several weeks in July and August starting on the same day of the week might provide better insights than a straight month-to-month comparison.

Please share this post:
Facebook Twitter Plusone Linkedin Email

Use a Custom Report to Aggregate Visits by Hour

view at full size

To aggregate metrics by hour, a Custom Report is required.

If you click change Visitors Overview report from the default Day view to the Hour view, you might not get what you expect.

What the report shows is not an aggregated view by hour for the time period selected, but rather the number of visits that occurred during each hour on each day.

Because this would equate to 720 individual data points for a 30-day month, this view seems to be practical only if you want to pinpoint the most active points in time over the course of that month.

This view can be more usable if you narrow the date selection down to a day or two; in this case, there are few enough data points on the main over-time graph to be intelligible.

But what if you want a consolidated view of hourly activity?

For this, you can easily set up a Custom Report using the Hour dimension. (The Hour of Day dimension, which provides the same non-aggregated data as the Day view in Standard Reports, won’t do the trick.)

It’s better to select Flat Table instead of Explorer format, since this suppresses the over-time graph, which would unhelpfully display only visits by day for the time period selected. If you want to chart the 24 aggregated hourly data points, you can export from the Custom Report into a spreadsheet and create a line or column chart there.

As a related note, if the time zone indicated in your profile settings is not correct for you, you can change it at any time, but keep in mind that this change will not apply retroactively to data that Google Analytics has already captured.

Please share this post:
Facebook Twitter Plusone Linkedin Email