Reading Reports In Google Analytics: Recency

December 3, 2007

Do you know how to read the recency report in Google Analytics? I sure didn’t.So I reverse engineered it — and here, come learn with me what it’s all about when they say, “Most people visited your site 0 days ago…”

Methodology. All great scientific experiments start with a boring chapter on methodology. (Right?) So here is mine: In order to test, I created a number of profiles where my home computer was the only “person” in the profile. Then, I was able to learn how GA computed recency when there was only one unique visitor, and where I knew exactly when she (that’s me!) had visited, and if she had wiped her cookies.

Here is what I learned:

Recency is about visits, not visitors.

When the report says, “Most people visited 1 day ago” — that’s just a typo. The report isn’t about people, or unique visitors. It is about visits. If you have any doubt, look at the column with the numbers in it. It says, “visits.” Then, whip out your calculator, add them up, and without changing the calendar, head on over to the Visitor > Visit Trending > Visits report. If they aren’t the same number, then comment here. Send me email. Call me on my cell phone.

I want to say this a different way. If you have 100 visits in the time period, there will be exactly 100 visits in the recency chart. Even if they all come from the same person’s computer.

Recency computes time (in seconds) between each visit from the same visitor. If the time between visits is less than 24 hours, it will show up as a Zero Days Ago visit.

Before I explain this one, let me show you the results of one of my tests. It will make it easier to explain.

recency-excel-luna.jpg

Without wiping my cookies, I set up a new profile for the LunaMetrics website, and created a user defined variable that had only myself in it. Then I visited. Here is when I visited — you can see the 24 hour clock down the left margin.

Now, I will take you on a tour of all these visits. At some level, this is incredibly boring, like all good scientific experiments. (I’m sorry. The truth is that I don’t have the definitive word on recency, so the best I can do is show you the evidence. Sort of like CSI.)

OK, for those of you who are still with me: All the visits that aren’t in a colored block had a visit right before it that was less than 24 hours earlier. So, for example, on November 16, I only visited once, at 7 am, but the visit before it was on the 15th at 19:00 — only about 10 hours earlier. Every single one of these non-colored visits will get a notch in the zero days ago category. (Right? If recency is time between visits, and there are only 10 hours between two visits, they are less than a day apart, so the earlier visit gets classified as zero days before the later visit.)

But what about those visits in the colored blocks? Let’s start with the two visits at 9 am on the 22nd (orange). I visited twice in one hour. One visit was probably 45 minutes or so before the other — clearly less than 24 hours before it, and it shows up as zero days. The other orange visit was preceded by the green visit. Notice that there are more than 24 hours between the one in green and the one in orange. So now we have a visit in the 1 Day Ago category.

Now look at Yellow, on the 17th. This is just like Orange: two visits in one hour, one of them clearly less than 24 hours before the other, and the other, more than 24 hours ago. Another 1 Day Ago visit.

OK, we will turn our eyes (if we haven’t gone to sleep yet) to Green on the 20th at 18:00 (6 pm).I didn’t even visit the day before. The prior visit was MORE THAN 48 HOURS before it. So that one will show up as Two Days Ago. (Go team!)

And finally, the first box, in blue. I don’t have records for when I visited before this date — I just started this profile on the day that I started the test, but I didn’t clear my cookies. This part is very important: The recency chart cares about visits outside the time you picked, as long as they are affiliated with a visit that is inside the time period. For example: If my most recent visit before the very first one, i.e. the blue one, had been 7 days ago, it would show up in the 7 day bar, even though that is outside the time period.

OK, let’s summarize. Of the 16 times in this period I visited, 12 of them had a prior visit that was less than 24 hours before it (0 days ago), two (an orange and a yellow) had a prior visit that was just over 24 hours before (1 day ago ), one had a prior visit that was two days ago (the green one) and one of them (the blue one) is in the “I’m not sure” category.

So if I pull the recency chart for the period of November 13-22, here is what it looks like:

recency-lunametrics2.jpg

So in this ten day period, we had one person (me) visit 16 times, and most of those visits were preceded by a more recent visit that took place less than one day before it. That’s what they mean when they say, “Most people last visited: 0 days ago.”

And moving right along: Very frequent visits are in the same zero days bucket with new visits. This is bad.

I set up about five different tests. Here is a different one that I set up on November 14. This time, I cleared all the necessary cookies, so that GA would think I had never been there before, but would still know to measure only me — and I visited on November 14, but not on the 15th or 16th:

recency-no-cookies2.jpg

The reason this is bad is that GA throws together visitors who visit the site often with visitors who are new. A content site might find that repeat customers are more likely to continue repeating (so they are the best visitors) and that new visitors may come back or bail (so they are question marks.) Those aren’t two categories we really want to combine.

So here are the takeaways and “to do’s” :

  • If you care about recency, create a returning visitors profile by using a filter to exclude new visitors (Need to learn how to create a custom filter?). If you see lots of visits in 0 days ago in this Returning Visitors profile, you are seeing visits from visitors who come back quickly. In fact, this chart is pretty hard to read, so the more you segment, the more you will learn.
  • Forget the line that says, “Most people last visited.” Just read the report and come to your own conclusions.

Let me close with two thoughts:

  1. I can’t stress enough how every visit during the period gets a matching visit that is the one that came just before it.
  2. I can’t stress enough how this is just my experiment. Go ahead, create your own. Prove me wrong. I would love to see it.

Whew. This took a long, long time to do. Many thanks to everyone who tried to help.