local media insider
Case study

Personalized paywalls increase revenues

Here's how they did it



Key executive: Brandon Erlacher, VP of strategic Data, Sandusky Newspaper Group

Company: Sandusky Newspaper Group

Initiative: Increase subscription revenues by targeting


Gigya: User database and single sign on

Cxense: DMP, targeting, algorithm, recommendations

Chatterbot: Ecommerce, customer service

Mailchimp: Newletter delivery, customer service and retention

Technavia: Digital Editioin

ProCirc: providing the digital access via rate.


The Sandusky Newspaper Group operates 12 community newspapers, 31 websites, a variety of email newsletters and a full-service digital agency, in Utah, Michigan, Ohio and Tennessee. In 2017, the company decided to create its first paywall.


Erlacher and his team wanted to find the best approach to maximize circulation sales but have a minimal impact on traffic and CPM based online advertising dollars.

“Paywalls drive revenue but choke traffic,"  Erlacher told attendees of the 2018 Mega Conference.  

"Circulation managers want the paywall. But the reality is that we also get revenue from banner ads. It is a multimillion business that we do not do not want to lose.¨  

Studies show that 75% of adults will not pay for news. In addition, churn is on the rise and repeated discounts that keep subscribers onboard reduce the overall rate of gross subscription sales.

One problem is that current paywall vendors provide only complete, metered or micropayment solutions. Erlacher wanted to create a hybrid solution that could target the most likely subscribers using a variety of data, but let other users “pass through.” He needed a program designed to respond to a deeper understanding of reader behavior.


Instead of an out-of-the box solution, the team decided to use the tool Cxense, that can send any widget or designed page - in this case the registration and paywall offer -  to readers based on their behavior.

The next question was which readers to target first? The simplest threshold for registration is a sessions per time period model, so some experimentation with the starting point was needed.  

Looking at at the overall model, ARPU, or average revenue per user over time, divides into four groups:

  • Casual users

  • Loyal users

  • Brand lovers

  • Subscribers

Brand lovers, defined as those who visit the site 15 times in 30 days, view an average of 10 pages at 4 minutes per session, have the highest revenue value, 30% more than loyalists and about 100 times higher than a casual reader.

The highest conversions were from people with five sessions in the last 7 days. But that is a limited group.

Ultimately the team chose a middle ground that seemed low risk but had enough readers to generate real revenue.  Here´s the basic plan:

1. Users with 8 sessions in 16 days received a registration prompt.  

2. After a 30 day cool-down period was maintained  the subscription offer is delivered.

3. The offer was $9 a week or $99 a year, and $13 a year for print subscribers.

4. A group of ten “smart people” in the company test serving registration offers against other behaviors, such as interest in long form journalism, or location within the county, moving the frequency requirements up or down depending on the theory, and watching results.

5. Any reader who uses an ad blocker is served a registration right away.

6. Any reader who is registered formerly as a print subscriber is served a paywall right away.

The registrations are placed after the first two or three paragraphs of the article, blocking them from further view. Once registered, users see a ´read more’ button during the “cool down” period.

Gigya collects all the data on registered users as well in order to keep the offers separate and collect a shorter profile. When the 30 days is reached, non-subscribers were given an option of $9.99 a week or $99 a year. Print subscribers are charged a $13 annual upsell.


The combination of technology is critical to the process.

Cxense tracks unregistered users in real time, showing category, author, date, referrer, device, browser and sentiment. Look-alike modeling then allows them to serve any widget to a user with the same behaviors.

This means that multiple models for sessions/days - and even pricing offers - can be served to readers with different interests, including offers for vertical emails, event promotions and ecommerce down the line.

“It is watching the behaviors in real time, trying to anticipate the next step, of both a logged in and not logged-in user," Erlacher said, “At the basic level, it tracks people who like this kind of content, where did they come from (direct or social referral),  have they been here before, what they did they do last time. "We can do look alike modeling, based on propensity. If a person is starting to look like this, I'm starting to show them content that looks like this.¨

¨We can drive a better set of whatever content I want to show, a paywall, a story, a picture, a call to action for a newsletter or an event.

Gigya on the other hand collects the information on registered users and segments the audience by a variety of basic factors:

  • Log-in status

  • Paid status

  • Newsletter subscription

  • Behavior

  • Demographics

“We use it to create a profile of the user from what is online and in circulation and in the email system. It proves a little beyond registration. About 20 to 30 percent of users use social media to log in so it will put from whatever social network they have,” Erlacher said.

“Once you make the link between Cxcense and Gigya, you have a profile that is going to last through time.

ProCirc is the software used to manage delivery of the different subscription bundles.

Finally, once a month, the group will change the setting in Cxense to test new behavioral targets. “I have a meeting with the people I think are smart, and we make a list and select the audiences and see what happens.”

“Xcense DMP lets you select different variable audiences. Tell me who has looked at this columnist on a Friday, logged in using Facebook, or made a comment in the last 30 days, or who likes to read negative sentiment, came from google, and has had four sessions in the last five days.”

An example of some of the experiments include geo-fencing around the county, in which case “we can lower the threshold from 8 sessions in 16 days to 6 sessions in 16 days, both running simultaneously.”

“If we have users who follow investigative reporters who write deep or long stories, we can test asking for a registration or a wall sooner,” he said. "We are continually adding different audiences and things to test."


The first test at the Standard-Examiner showed an annualised increase of $60k in circulation revenue with no change in users session, referral traffic, or MUV (Monthly Unique Users). Conversion rates averaged at 22% over time.

There were very few complaints regarding payments, but requiring registration was more contentious. “Communication internally is key,” Erlacher said.  “We have scripts for all employees, especially customer service, sales and the newsroom."

One surprise was that many customers thought they were already paying, which resulted in more sales.

Closing free access to the Digital Edition also resulted in a heavy call volume to customer services. In response, customer service was given a web tool to allow customer service reps to provide immediate access to the digital edition on request and create a Gigya account for the customer.

Lessons learned

  • Driving new customers into the funnel is vital. A focus on breaking news and aggregation was key to getting potential new paid subscribers into the system. A social reporter, for example, writes social media round-ups of hot topics that appear in local social feeds to bring more people to the website. “You always need to key the top of the funnel full of new prospects.

  • Pricing is still a work in progress. “We had to create a new lower rate as a fall-back option for some customers,¨ Erlacher said. “We will use Mather to drive them to higher rates over time.”

  • Training, especially having sales scripts and good technical walk-throughs made a big difference.

  • “Many of those willing to pay expect a seamless system for linking their account... Patience and staying on the line with customers is necessary.
  • Employee accounts have actually been “the biggest headache”, many of our employees did not have a subscription.


Erlacher contends there is no ideal sessions/days system.

“A lot depends on what your revenue from banner ads is today. I would say, look what network revenue you are bringing in now and realize that some of that will be at risk. What percentage of risk are you comfortable with to start? Start there.

“People who come back time after time see some kind of value. Ask them to register first, cool them off, and if they are still at this high level, ask them for money. Because the algorithm checks them every time, if they don't come back it restarts."

“For me it is about risk, how do I mitigate risk? There are 100,000 ways to do it. This is how I decided to do it, with a lot of  smart people throwing stuff at the wall."

Erlacher plans to roll-out the system to other Sandusky websites, tweak the algorythms, and use AB testing to identify the best performing offers, rates and promotional messages. Also on the road map is deeper integration with Mailchimp to remarket to for abandoned shopping carts.


No comments on this story | Please log in to comment by clicking here
Please log in or register to add your comment
met-Media Execs Tech
Meet your peers

Please log in to join your group.

View all groups