plus
We are at the beginning of a shift. We realize that our analytics is giving us the wrong answers. And there is a trend forming in the analytics community to fix it
There is an interesting trend in the world of analytics. Just like the big trend in social media is now "return of investment", the big trend in analytics is that we are beginning to question the value and accuracy of our metrics.
What we are realizing is that our impressive web statistics are not matching reality. You see this with paywalls and how no newspaper are able to reach their projections. You see it when with digital advertising, which isn't providing a positive return.
Something doesn't add up.
It is the same with the new real-time analytics solutions. They are extremely dangerous to use because they are not aligned with what really matters.
In this article, I'm going show you what the problem is and what you need you to do instead. I cannot give you a final solution, because none exists, but I will point you in the right direction. You should actively use your analytics to design your article to help you grow. You don't want to give everyone the same thing. A subscriber is vastly different from a one-time visitor who just clicked on a link.
Web Analytics is not designed with publishers in mind.
Before we discuss the solution, let's take a look at the problem. The web analytics that you know today was never really designed with publishers in mind. It was designed for brand and ecommerce websites.
When you hear about things like absolute unique visitors, page views, bounce rates, and conversion funnels, they not really relevant. In most cases, the standard data and the standard reports are directly misleading.
There are two big problems with web analytics for publishers.
The first big problem is that the raw data is wrong. Let me give you a simple example:
The largest danish tabloid newspaper had a reported 6,639,565 absolute unique visitor (Danish IPs only) in May 2011. Compared to the New York Times this doesn't sound that impressive until you realize that the population of Denmark is only 5.5 million people.
So according to the analytics, this newspaper reaches 114% of the total population (in all age groups.) That is not actually possible. The same newspaper also measure their audience via phone surveys (calling people asking them if they have visited the site), and when you adjust the sample size to the total population, the result is that they only have 1.3 million readers-which sounds much more likely.
Their raw data is more than 500% off target.
The reasons for this inaccuracy can be by several things. There are an increasing number of automated requests. In the past we just had search engine bots etc, but today these automated requests also comes from many other services-like every time someone shares a link on Facebook.
Analytics tools like Google Analytics do a good job filtering these out, but can you say the same for your publishing system?
There is the increasing problem that the internet is now a multi-device environment-especially for newspapers. It is very likely that people check up on news several time per day ...meaning they use their phones, tablets, their work computer and their home computers. That is four devices, and thus four absolute unique visitors-which is the same person.
And then we have the growing problem that more and more people use different forms of blocking tools ...or browse incognito. With the addition of local legislations (like the European Cookie Law) which makes it harder for us publishers to track people.
Every time someone deletes a cookie, they show as a new person and drastically skews the results. One person blocking a cookie, but checking news several times per day, might show up as 50 different people over the cause of a month.
That is not good!
The other problem is much more deceptive. Almost all analytics systems are based on a fairly standard set of measurements-like page views, absolute unique visitors, visits, bounce rates, exit rates, pages per visit, time on page-and for the more the advanced users-conversion funnels.
All of those are designed for websites in which a conversion is different from a visit. For an ecommerce site, the conversion is getting people to not just look at a product, but also buy it. They need to checkout. For brands sites, it is all about maximizing exposure, so you want to encourage people to "explore."
With these types of sites, things like exit rates, and conversion paths are very important. You don't want people leaving your site before they have completed a purchase.
For publishers, however, your product is the page. You sell articles, so you have very different conversion goals. You need to convert the one-time visitors into readers, make them likely to come back, build up value over time and turn finally them into subscribers.
That is your conversion funnel.
Which means all the standards metrics has no relevance to us. Metrics like exit rates etc. are irrelevant if people have read the article. You have already converted them into readers. You have already sold your product.
Another example is page views. Newspapers seem to be in love with getting clicks and page views, and they are shooting themselves in the foot because of it. It is easy to get page views. Just divide your page up into a slide show, and suddenly one page is converted into 10 page views. The result is that your bounce rates improve dramatically, your exit rates drops significantly and your pages per visit increase.
It all sounds very good, until you look at how effective those types of articles are at turning one-time visitors into loyal readers. And then you find that you have suddenly become a spammer. Yes it attracts a lot of traffic but the conversion rates to real readers is appallingly low.
You are looking at the wrong metrics and thus making the wrong decisions.
In order fix the problem, you need to move away from looking at analytics as a single point of metrics.
A simple example. It is not relevant to look at time-on-site per visit. But it is highly relevant to look at time-on-site per person. With that you might find that people (in average) would not be turning into subscribers until they have spent 70 minutes or more reading your articles.
Base everything you do on *people*, not things, click, views, hits, or other technical do-hickeys.
But it isn't enough just to look at people, you also have to segment them into different types of readers. Specifically, you need to segment your audience into one-time visitors, loyal readers and subscribers. Each segment has different user behavior and conversion goals.
A one-time visitor is someone who just found a link somewhere, either on social channels, a link mentioned a blog post, on another site or simply via search.
Your goal for these readers is not to turn them into subscribers, because nobody is going to pay for a newspaper they just found. You have to establish a relationship with them, meaning that your conversion goal is solely to turn them into loyal readers.
Forget about trying to sell them something, it won't work. Focus on showing them what you have to offer and give them a good reason to come back.
Your loyal readers are different from your one-time visitors. They have already decided that you are interesting, now your conversion goal is to provide value. It is no longer about what you have to offer. It is about how valuable it is. You have to build up your loyal readers perception of value, so that after a while they decide to turn into subscribers.
For subscribers it is different. They have decided that you are worth paying for. You no longer to convince them to buy anything-they are already paying. For subscribers you have to focus on providing in-depth value.
This is not just important in terms of analytics, also think of segmenting your audience when you design your newspaper. You need create different designs for each audience. One-time visitors need more background info, because they don't know the story, and they need more information about what else you have to offer.
Subscribers need a page design with more perspective and opinion. They already know what you have to offer. For subscribers it might also be relevant to expand beyond your newspaper-link more aggressively to other sites. This is dangerous for one-time visitors (because they don't come back), but it is highly valuable for subscribers.
It is not enough just to measure people in different segments. You have to look at what these *people* are doing. What are their actions and pre-actions?
For publishers the most important action is "how many people read the article?" Take this simple example:
In the table below you can see two highly popular articles (IKEA+viral) about marketing videos, followed by two less popular articles about media analysis. Based the page views, should you focus on "marketing videos" or "media analysis" in the future?
If this is the only data available to you, you would say "marketing videos." It is clearly the most popular type of articles. The IKEA article is more than 600% more popular than the analysis of how Mythbusters formats a broadcast hour.
But you are not looking at what actions people take. You are just looking at clicks. Instead, Iif you were to look at how many actually read an article, you see a very different picture.
Now the Mythbusters analysis is almost twice as effective at converting people into readers, than the one about IKEA. This is vital information to have.
Always compare your analytics to what action you need to take.
Pre-actions is equally important. It should answer why people are doing what they are doing. Why are they coming to this page? Where are they coming from?
Most analytics include referrer information, and you have probably added campaign variables to track traffic from news feeds, or email newsletters, or maybe traffic from iPad apps.
Here is the thing. You need to include referrals from your own content as well. If one article is leading many people to another article, then that article is just as important as a referral source as a 3rd party site.
This is information that you do not have today.
Finally, you have all the varies forms of reactions. What do people do after they have read the article. How many people share it? How many comment? Is it positive or negative? How many turn into loyal readers or subscribers? How many do you reach in term of exposure?
Most analytics systems will only tell this for the site as a whole, but that is not really useful. You need to know it specifically for each article and for each segment of readers. How else would you learn why that articles worked while others didn't?
And more important, what kind effect did the reactions create? Just knowing how many people share an article is irrelevant if you don't know how many new readers you got because of it.
It leads back to pre-actions. It is a never ending cycle. What is the effect of sharing? Who are sharing? Who are generating the highest return of readers?
In order to put this into perspective, here is a concept page of what an analytics page could look like for publishers (big version here.)
Note: This is just a concept. Not a real analytics page and not real data. It is designed to help you understand what to aim for.
As you can see, this page is very different from the usual page reports your get with traditional analytics. Gone are metric like page views, clicks, hits, time on page, bounce rates etc. Instead every single metric is based on either *people* or *real readers*.
You do not want the technical metrics, because they are misleading.
It starts with a graph showing how many people this articles has attracted since it was published, segmented into different types of readers.
This is actually very interesting. It shows is that on some of the days where the traffic was spiking traffic from loyal readers and subscribers dropped. There can be many reason for this. E.g. if you get traffic for one-time readers shortly after it a spike from subscribers, it would indicate that your subscribers had been sharing the articles.
Next is the real-time box. Real-time analytics is very interesting new area of analytics, but most tools out there are doing to wrong-they are tracking views and clicks.
What you want to do is to track real-time readers. And you want to know how that impacts the performance of your article. And, you want to know why.
Next is how good you are at converting people into readers. This is where it starts to get interesting. By segmenting your audience you get a completely new level of understanding.
If you look at all your visitors, you learn that 78% of your audience clicked on the link but left without reading the article (which is pretty bad.) But by segmenting, you learn that the drop rate is primarily caused by your one-time audience, while your subscribers have a pretty good read rate.
This drop rate is your new bounce rate metric. Instead of measuring bounce rates, based on how many pages a person views, measure bounce rate based on whether or not people reads the article.
The arrows underneath indicates how efficient this article is at converting people into loyal reader or subscribers. In this example, 1,366 people decided to come back and read another article *after* reading this article.
Of the loyal readers, no person decided become a subscriber just because of this single article, but 229 people decided to subscribe *after* reading this and other articles (multi conversion funnel.)
As for reactions, again we see a huge difference between each segment of readers. One-time readers performed rather badly, but 78% of your subscribers shared the article.
More importantly is the effect of sharing. Your 9,184 subscribers, who read this article, caused 12,273 new readers to come to the site. That is impressive.
Curiously enough, sharing done by your one-time readers caused 11 subscribers to read the article. So people unknown to your site, influence people who are already subscribing. It is just simple example of our interconnected social world.
We also need to look at the economics of the article. And again, segmenting the data is vital to understanding what really goes on.
In this example, the site is monetized by both advertising, additional products (like books, reports, guides etc), and subscriptions (which is free of advertisements.)
What we can see is that our one-time visitors are actually a significant source advertisement income, exceeding that of your loyal readers. But your main source of income is from your subscribers.
This a bit of a problem, because it means you have an economic dead-zone in-between your one-time visitors and your subscribers. But without segmentation, and focusing on people rather than clicks, you would never know.
The page also highlight if the article is profitable or not based on how much it costs to make it (time to edit/write + other expenses.)
It is the same when it comes to where people are coming from. In this example, one-time visitors are mainly coming from social+news aggregators (like Stumbleupon) and Facebook, while your subscribers are mainly coming from source where they follow you.
This is great news, because it means you have a close connection to your regular audience. If it had been reversed you would have a problem.
Again, this is measuring people, not views. 26% of your subscribers came from Twitter, not that 26% of the clicks. Clicks are irrelevant.
And finally, segmenting your demographics might illustrate a number of very important issues. Here we are looking at countries, but you might want to look at more than that.
In this example, there is a clear disconnect between who are subscribing (mostly from Europe) and your one-time visitors (mostly from North America.) This might partly explain the economic dead-zone or why you are struggling to convert people into subscribers.
All of this is just an example illustrating how important it is to base your measurement on people and readers. And how important it is to segment it.
We have just been looking at a single page, but imagine if you applied the same principles to your newspaper, magazine or blog as a whole. Measure how people you reach, then segment them and measure things like "most popular articles for one-time visitors" vs. "most popular for subscribers or loyal visitors."
You need look at each section as a whole, each topic as a whole, and your analytics needs to be able to group articles together as well. When covering events like the "Hurricane Irene" or the "Terror in Oslo," having this level of insights for all the article about will give you a much greater understanding of your coverage as a whole.
For each segment and group you can then place the articles in a graph like this one:
Articles with a high popularity and a high read-rate are obviously what you want. And articles with a low popularity and read-rate are not what you want.
But if an article has a low popularity rate but high read rate, you need to ask yourself, why isn't this reaching more people? Are they not sharing it because they disagree with it? Or is it simply because you are not doing enough to facilitate and encourage sharing (e.g. is locked behind a paywall? Is it formatted for only one device?.)
On the other hand, if the article is popular, but with very few readers, you have to figure out why. Is it because you cheated your audience with a deceptive title (link-baiting) ...or is it simply because it is too simple or complex a topic for your audience.
Also remember segmenting. Low value articles might be good for creating exposure for your one-time visitors, but you don't want to create these articles for your subscribers.
It is the same when it comes to sites monetized by advertising. Instead of read rate you would measure the CPA rate. Are some article generating a higher CPA rate than others, and why?
The obvious question now is, how can you do all this today? The main problem is that the traditional analytics services are based visits, views, and non-publisher's needs.
You can get a long way by setting up custom segments, customs reports and adding custom analytics variables. E.g. you could create a custom report that would list unique visitors per page, instead of page views. You can do the same with most other metrics, but it wont tell you have many readers you have (which is critical.)
It might be possible to create custom triggers on your site to ping Google Analytics when a person has read an article. I'm still looking into that.
Many newspapers and magazine have build their own systems or they are using custom publishing systems with varies forms of analytics. If you are one of those, check if it is measuring people or views?
We are basically at the beginning of a shift. We realize that our analytics is giving us the wrong answers. And there is a trend forming in the analytics community to fix it.
My guess is that, in a year or two, we will start to see new types of analytic systems where you can track people and readers in a useful way. Up until then, my best advice is to create custom reports for each key metric.
Creating a propensity model is one of the most important tools publishers can have.
Free for subscribers
...or full access for $12
Free for subscribers
...or full access for $12
Free for subscribers
...or full access for $12
Several publishers have found that reducing volume leads to an increase in revenue
Free for subscribers
...or full access for $12
Free for subscribers
...or full access for $12
Time is such a critical metric for publishers, but it's also a very complicated one.
Free for subscribers
...or full access for $12
Free for subscribers
...or full access for $12
Free for subscribers
...or full access for $12
Free for subscribers
...or full access for $12
Publishers who start their own data studios need to take extra steps to identify real people.
Free for subscribers
...or full access for $12
Founder, media analyst, author, and publisher. Follow on Twitter
"Thomas Baekdal is one of Scandinavia's most sought-after experts in the digitization of media companies. He has made himself known for his analysis of how digitization has changed the way we consume media."
Swedish business magazine, Resumé
plus
free
plus
plus
plus
free
