Report: The Editorial Paywall Optimization Tool, created with funding from the Google Digital Initiative Innovation Fund, and its use in the newsroom




Report:The Editorial Paywall Optimization Tool, created with funding from the Google Digital Initiative Innovation Fund, and its use in the newsroom







Facebook Twitter Del artiklen 

The Editorial Paywall Optimization Tool is a project created by Kristeligt Dagblad with funding from the Google Digital News Initiative Innovation Fund.

It is a simple-to-use editorial tool created to enable journalists to choose and dynamically test the betst paywall placements in articles.

Its purpuse: To increase the conversion rate per article compared to fixed paywall placements.



Table of Contents

1. Project Description

2. Summary

3. An Innovative Tool

4. How the Editorial Paywall Optimization Tool Works

5. The Timeline and Process: Milestone 1, Milestone 2

6. User Stories

7. Case study: “Erik Clausen: I min alder er døden ingen tragedie”

8. Analysis and Conclusions: The Editorial Paywall Optimization Tool Now and in the Future



1. Project Description

More sophisticated paywalls are necessary if the number of digital subscription conversions is to increase. The ambition of the Editorial Paywall Optimization Tool project was to give digital news journalists a simple-to-use tool that could enable them to choose and dynamically test the best paywall placements within a specific article body text in order to increase the conversion rate per article compared to fixed paywall placements.

This approach has several benefits; chiefly among them is empowering journalists and other editorial staff with the ability to increase conversions to subscriptions.

The number of log ins or conversions to subscribing customers is arguably the most important web metric for any digital news article. The Editorial Paywall Optimization Tool project insists that editorial staff can and should make a difference here.



2. Summary

The Editorial Paywall Optimization Tool is now a reality. It is in use and we consider it a success. Over the course of the last half year we have conceptualized, created, tested (and tested and tested) and finally implemented the Editorial Paywall Optimization Tool in the newsroom.

In conjunction with a total redesign of our article template both front and back end on k.dk we have in fact implemented the Editorial Paywall Optimization Tool to such a degree that not a single article, whether it’s created web or print first, is born without a dynamic paywall that editors and journalists can change and move around.

This publically available report is a part of our contract with Google DNI and will show how we created the Editorial Paywall Optimization Tool, how we use it and what we hope to accomplish in the future with the Editorial Paywall Optimization Tool.

The Editorial Paywall Optimization Tool would not have become a reality if not for the Google DNI Innovation Fund. The idea would in all likeliness have remained a dream, drowned along with many others in the daily rush of making a digital newspaper.



3. An Innovative Tool

The Editorial Paywall Optimization Tool is innovative in several ways:

Firstly, it takes a sophisticated and dynamic approach to relationship between article paywalls, paywall call to actions, and the article content which is the paywall context.

Secondly, it explores new forms of collaboration between digital journalists and marketing and sales on the basis of the idea that the journalist behind an article is best situated to suggest which potential paywall placements and paywall messages will work on a specific piece of content.

Thirdly, the tool insists that the expertise of content producers, sales and marketing must be challenged with continuous split testing and that the results of these tests must be presented to all team members in order to continuously refine paywall positions and call-to-actions.

Last but not least, this tool could and should strengthen the written journalistic product. It will continually remind the journalist to write good journalism: To keep it interesting, to make the reader keep reading and wanting more.



4. How the Editorial Paywall Optimization Tool Works

From the beginning we had four specific goals we wanted to achieve with the Editorial Paywall Optimization Tool, goals that have all now been met:

1) Editorial staff can place and customize paywall placements and call to action mesages within article body texts with tags with our Drupal installation.

After several test rounds with hand coded html articles containing a prototype Editorial Paywall Optimization Tool, we now have a fully implemented and easy to use tool.

It works like this:




In this video you will get a short tour through the workings of the Editorial Paywall Optimization Tool. How is works is further explained below.


If an online journalist creates the article, he or she can place a paywall with a single click on a button.




Figure 1: Our article template back end.



Figure 2: A close up of the body text field. The user simply places the curser wherever he or she wants a paywall, then use the button we created for the Editorial Paywall Optimization Tool.


When the paywall button is clicked a pop up window opens with an optional text field (see Figure 3 below). If the user does not write in the text field, a standard text is used. But this gives the editor/journalist/marketing department the opportunity to create more effective paywalls. If for example it was a feature article reconstructing a murder story, we could write, “Do you want to know who did it?” or some such story specific texts. It could be an article in a series highlighting a certain subject, and we could use the optional text field to tell readers that there is much more to read about the subject.




Figure 3: This pop up window opens and the journalist/editor/marketer can write an optional story specific text for the paywall.


Here is an example where we even played with some animated drawings and a story specific text (the front end result of the Editorial Paywall Optimization Tool:




Figure 4: The Editorial Paywall Optimization Tool is build to be story specific. This example is from https://www.kristeligt-dagblad.dk/historier/william


In the CMS back end of an article, paywall placements look like this:




Figure 5: The paywalls once placed in the body text field. Notice that thay get each their unique ID and because of the icon (same as the button icon) they are recognizable as paywall placements. As the text on the right side of these elements indicate, journalists can also chose to differentiate between paywall placements on mobile and desktop browsers.


To edit a paywall once placed, all you have to do is double click the element and the pop up opens again.

To delete a paywall placement, you can simply use backspace on the keyboard.

To move around with the paywalls: Either drag and drop or use Control + X, Control + V on the keyboard.

Finally, journalists can differentiate between paywall placements on mobile and desktop browsers (see Figure 4 above).

If the article is transferred from Newscycle Solutions (which is the system we use for the print paper) to our Drupal set up, the article is born with a minimum of two paywall placements. Later, an editor or an online journalist can move these automatically placed paywalls around as he or she sees fit.

(Sidenote: As more and more data is collected, we will continually review where the paywalls should be placed automatically – 5, 10 or 50 percent down the text? Or all three?)



2) Once the article is published the paywall placements will rotate. Caching- and security issues will be dealt with in order to optimize user experience and page loadtime.

Please, try it out for yourself. Make sure you are not logged in and then press the “refresh” button in your browser a few times. The placement of the paywall will change. Here is an article with three paywall placements: https://www.kristeligt-dagblad.dk/historier/erikclausen (this particular example is one of our prototypes, not our final article template, but the effect is the same)




Figure 6: Three different paywall placements in the same article. Example from the link above.


“Behind the scenes”, in the source code, users who are not logged in will be able to view the source code and read text in html markup down to the last paywall if they chose to inspect the code. However, people who inspect code will be relatively few. In any case, after the last paywall placement no text can be found in the source code.

This is because users who are not logged in actually receive a completely different html page.




Figure 7: No more text after PAYWALL C (https://www.kristeligt-dagblad.dk/historier/erikclausen again).


3) When statistically significant, conversion values will be calculated. Significance levels will determined pragmatically based on click, conversion and traffic levels.

And

4) The editorial staff and relevant management will be informed of winning paywall placements. Aggregated click and conversion data from all articles will be available on daily and weekly basis (using customized reports in Google Analytics).

For now, we use Google Analytics to monitor which paywall placement is the most effective and calculate its value.

Below is a screenshot showing that paywall placement “C” made a user sign up through the story “Erik Clausen: I min alder er døden ingen tragedie”, while paywall placement “A” made a user sign up through the story “Amerikanske Medier: Er Trump psykisk syg?”

(“A”, “B” and “C” simply means the first, second and third paywall placement in the body text field)




Figure 8: Screenshot from Google Analytics showing us which paywall placements create the most conversions.


Two screens in the Kristeligt Dagblad digital newsroom always show these numbers in real time so the editor can determine what paywall placements to keep.




Figure 9: Digital Editor at Kristeligt Dagblad Stinne Andreasen with Google Analytics on two screens behind her desk. The screen makes the entire editorial organization aware of the current rates of different paywal placements.


In the coming days we will create a setup where the most relevant data regarding paywalls and their respective values will be automatically scraped from Google Analytics and collected in a few daily e-mails sent to the digital editors.

A little bit further down the line, we will fully automate this process, so that the computer choses which paywall(s) to keep and which to discard. Work on this stage of the Editorial Paywall Optimization Tool is already in progress. Nevertheless, the screen with the raw data from Google Analytics will still have a prominent place in the news room.



5. The Timeline and Process: Milestone 1, Milestone 2

The Google DNI Innovation Fund requires certain milestone goals, which helped form our process. We had two milestones.

Our application to the Google DNI Innovation Fund was send in December of 2016 and selected by the fund in that same month. Both parties signed our contract in January 2017.

Milestone 1. Completion date: May 1st 2017. Goals: Implementation and testing of the Editorial Paywall Optimization Tool. Prototypes. Completed on time.

Milestone 2. Completion date: September 1st 2017 (originally proposed for July 1st, but postponed).

GOALS

Milestone 2 Goals


  • a. The Editorial Paywall Optimization Tool is in use.
  • This goal have been fully met.
  • b. All journalists within the digital department are familiar with the tool and are using it on relevant articles.
  • This goal have been fully met.
  • c.Data is collected and being analyzed.
  • This goal have been fully met.
  • d. Per article average conversion rate has increased at least 25 percent.
  • As of August 30th 2017, the increase in conversion rates have not fully been achieved.

Milestone 2 Goals


  • a. The Editorial Paywall Optimization Tool is in use.
  • This goal have been fully met.
  • b. All journalists within the digital department are familiar with the tool and are using it on relevant articles.
  • This goal have been fully met.
  • c.Data is collected and being analyzed.
  • This goal have been fully met.
  • d. Per article average conversion rate has increased at least 25 percent.
  • As of August 30th 2017, the increase in conversion rates have not fully been achieved.



The per article average conversion rate (d) have not yet increased as much as estimated. We are still in the process of our analyzing best performing paywall placement practices.

However, we are still confident that we will be able to make significant increases in overall conversion rates on kristeligt-dagblad.dk because of the Editorial Paywall Optimization Tool.

Early results and tests paint a very promising picture. We will further elaborate on our findings in the case study and analysis further below in this report.



6. User Stories

As part of our original proposal we submitted six user stories. These help to further explain the scope and use of the Editorial Paywall Optimization Tool works.

USER STORIES

Milestone 1 User Stories


  • User story 1: As a journalist, inside Kristeligt Dagblad's drupal CMS on the edit article page, I can place more than one paywall tag in the form containing the body text of my article.
  • This user story has been delivered.
  • User story 2: As an anonymous user on kristeligt-dagblag.dk's article page, I can experience a paywall placement and message, which is customized to the article that I am reading and placed according the body text form tag inserted by the journalist (user story 1).
  • This user story has been delivered.
  • User story 3:As another anonymous user on kristeligt-dagblad.dk's article page, I can experience another paywall placement and message (different from user story 2), which is customized to the article that I am reading and placed according to the body text form tag inserted by the journalist (user story 1).
  • This user story has been delivered.
  • User story 4: As a journalist, I can easily access key metrics related to different paywall positions. Also, I can see which placement had the strongest performance, and when the winning placement was determined. This data is also available inside Google Analytics on an aggregated basis.
  • This user story has been delivered.

Milestone 1 User Stories


  • User story 1: As a journalist, inside Kristeligt Dagblad's drupal CMS on the edit article page, I can place more than one paywall tag in the form containing the body text of my article.
  • This user story has been delivered.
  • User story 2: As an anonymous user on kristeligt-dagblag.dk's article page, I can experience a paywall placement and message, which is customized to the article that I am reading and placed according the body text form tag inserted by the journalist (user story 1).
  • This user story has been delivered.
  • User story 3:As another anonymous user on kristeligt-dagblad.dk's article page, I can experience another paywall placement and message (different from user story 2), which is customized to the article that I am reading and placed according to the body text form tag inserted by the journalist (user story 1).
  • This user story has been delivered.
  • User story 4: As a journalist, I can easily access key metrics related to different paywall positions. Also, I can see which placement had the strongest performance, and when the winning placement was determined. This data is also available inside Google Analytics on an aggregated basis.
  • This user story has been delivered.

USER STORIES

Milestone 2 User Stories


  • User story 1: As an editor, journalist or marketer, I can analyze the best performing paywall placements in Kristeligt Dagblad's archive of articles with paywall tags and tests.
  • This user story has been delivered.
  • User story 2: As a journalist on any media, I can download a report with the first findings concerning text around and within paywall positions including the raw data generated by the Editorial Paywall Optimization Project so far.

Milestone 2 User Stories


  • User story 1: As an editor, journalist or marketer, I can analyze the best performing paywall placements in Kristeligt Dagblad's archive of articles with paywall tags and tests.
  • This user story has been delivered.
  • User story 2: As a journalist on any media, I can download a report with the first findings concerning text around and within paywall positions including the raw data generated by the Editorial Paywall Optimization Project so far.



As the chapter “How the Editorial Paywall Optimization Tool works” shows, all of the above user stories have all been met. This publicly available report completes Milestone 2, User story 2.



7. Case study: “Erik Clausen: I min alder er døden ingen tragedie”

Let us take a closer look at one of our articles with three paywall placements, the story “Erik Clausen: I min alder er døden ingen tragedie”. It is a fairly typical Kristeligt Dagblad article with pretty average numbers.

Paywall A is placed about 10 percent down the article text, right after a big question: What does Erik Clausen (a famous Danish movie director and actor) mean, when he says “death is a beginning”?

The reader who meets Paywall A can read the beginning of his answer, the rest of which is blurred out:




Figure 10: Case study: Paywall placement A.


Paywall B is placed about 25 percent down the body text, and for readers meeting Paywall B it is “revealed” that the story contains video and other story enhancing material. The video alone invites these readers to spend more time on the article.

Again, the paywall is placed right after an interesting question about his rerlationship to death. Just visible, but blurred, the text reveals the beginning of his answer (“I’ll show death. I know it wins, but it’s going to have to run fast to get me”)




Figure 11: Case study: Paywall placement B.


Paywall C is placed almost at the 50 percent mark which is much further down than we would ever have tried without the Editorial Paywall Optimization Tool. The reader with paywall C gets even more of the text, an extra picture and meets the paywall at yet another interesting place in the text:




Figure 12: Case study: Paywall placement C.


Let us look at the numbers for these paywall placements.




Figure 13: The above discussed article in Google Analytics


First of all, Paywall A, B and C was seen by as about the same amount of readers, but with Paywall C being met by slightly more than Paywall A and B, 176 more page views than Paywall A:




Figure 14: Translation: "Sidevisninger" is "Page views" and "Unikke sidevisninger" is "Unique page views".


The amount of time spend on the page is clearly (and unsurprisingly) relative to how much of the article we have given them:




Figure 15: Translation: "Gns. tid på side" means "Average time spend on page"


However, the big question is of course, which paywall placement was the most effective? Which of the three placements worked the best? (And yes, why, but that is not something Google Analytics can tell us).

The answer to that question is Paywall A, which made three people sign up to our news paper. Paywall B had no success, and paywall C made just one person sign up for the newspaper.




Figure 16: translation: "Sideværdi" means "Page value".


These numbers might not sound big, but that is actually pretty o.k. for a single article. Also, should we judge from this single article, giving away just a little seems to be slightly more effective than giving away a lot. But that can hardly be concluded form this example alone - or even from the accumulated data we have so far. Overall the spread is almost equal, though giving away “too much” (maybe. +50 % of an article) does not seem ideal for business over all, especially not in standard news articles which tend to be written from the old journalistic principle of “tell the most important stuff first”. With some of our feature journalism that is written with a more dramatic structure, however, we are beginning to see a different trend where the later paywalls are the most effective.



8. Analysis and Conclusions: The Editorial Paywall Optimization Tool Now and in the Future

It is too early for us to draw any final conclusions, but preliminary results does at least indicate that different types of articles are best served with different paywall placements. The ideal paywall placement is, it would seem so far, determined by the content, style and dramaturgy of an article.

Taking all of that into consideration is something a machine cannot do. The true power of the Editorial Paywall Optimization Tool lies in the fact that it gives us the human control and a sharp focus on ideal paywall placement solutions – build into the daily work flow of journalist and editors in our digital news room.

In time, we will learn more and more from studying and analyzing the effectiveness of paywall placements in different types of content – and learn to use the Editorial Paywall Optimization Tool to an even greater effect than today, where we are just beginning to use it.

The Editorial Paywall Optimization Tool has already had a substantial effect on how individual digital journalists and editors work at Kristeligt Dagblad, it has increased our focus on converting readers into subscribers, it has further strengthened the collaboration between departments – journalists, editors, developers and marketers.

In an era where the work of a typical journalist is changing, our Editorial Paywall Optimization Tool is proving to be natural and easy to implement continuation of that trend.

In former times, a journalist wrote and delivered an article. Someone else set it up on a page. Someone else published it. Someone else marketed it. Someone else sold it.

No longer. Today, most journalists’ work is not finished when the article is. Journalists all over the world market their own content through social media, they are in direct contact with readers and can themselves update and change published work.

It is in this reality that the Editorial Paywall Optimization Tool shows its true worth and empowers journalists to sell their work – and perhaps even help them to write better.

Better, because the Editorial Paywall Optimization Tool by its very existence amplifies the focus on creating “cliffhangers” for the individual journalist using it in our newsroom. They want to place the Editorial Paywall Optimization Tool in the most optimal place they can. Early user tests among journalists show us that this is how the Editorial Paywall Optimization Tool actually effects the workflow. Sometimes the text dictates where a paywall should be placed; sometimes it works the other way around:

The possibility of choosing a paywall placement (or more) encourages the journalists to write with possible placements in mind. In effect, this can actually make for better writing; articles with more interesting dramaturgy and cliffhangers.

Of course, this could have resulted in bad journalism, “cheap click bait”, but the fact of the matter is – it does not. Perhaps this is because Kristeligt Dagblad is a legacy paper with strong ethical standards and traditions; we have strong policy against such writing and simply will not make set-ups that do not pay off reasonably, neither in headlines nor in these new custom paywall texts. Nevertheless, it is a danger, one that we are aware of. And one others should keep in mind, if they want to use the Editorial Paywall Optimization Tool.

This journalistic impact will be interesting to follow. We believe no one is better equipped to place paywalls in an article than the writers and editors themselves. This philosophy was the impetus behind creating the Editorial Paywall Optimization Tool in the first place. So far, it is exceeding our expectations.

A key factor in the success of the Editorial Paywall Optimization Tool is how easy it is to use. A great part of the funding provided by the Google Digital Initiative Fund has therefore gone into rethinking, developing and testing our redesigned CMS. It is now just as easy for our journalists and editors to create a paywall placement, as it is to add a picture, a hyperlink or a fact box.

How successful is the Editorial Paywall Optimization Tool? The truth is that it is too early to say. Editorially it has impacted our workflow positively; that we can conclude. But does it give us more subscribers? Can we learn from the data collected through Google Analytics on the success rates of different paywall placements – and become better at determining where and how to place our paywalls?

These questions will be answered in time. Therefore, we will update this online report on September 1st 2018 when we will have learned more, created improvements and further developed the Editorial Paywall Optimization Tool.