Increasing conversions

Context

Before I joined this company, there was no one who focused on UX, the company had never previously done any type of user research. The company is a news site who has three types of subscriptions available: Free, Standard and Premium. ‘Free’ is where the majority of their users were, very unwilling to pay a subscription fee. The company is successful and making money, but they didn’t know about their user’s behaviour and why they weren’t paying for subscriptions. 

I was brought in to conduct some UX research, my main job was to find why users weren’t upgrading their free subscriptions into paid accounts, and how to get them to pay for subscriptions. On previous projects, I mainly used qualitative and quantitative methods of user research, where I could speak to users directly. But I had constraints on the project where we weren’t able to speak to users. The research method that was available was analytics research. 

Research Plan

As the only UXer at this company, I worked independently. I constructed a plan for the next month.

1. Establish macro and micro conversions
2. Look at the different analytics tools being used (Google Analytics/Mixpanel) to see how they are being used currently 
3. Speak to marketing and sales about how they are using analytics tools
4. Create and validate user personas
5. Develop hypotheses/experiments
6. Define metrics to solve experiments and hypotheses
7. Solve hypotheses/experiments
8. Define possible solutions
9. Create segments for different subscribers: non-sub, free, standard and premium, then track their behaviour.
10. Explore exit rates of free users using their limit of their subscription and conversion rates 
11. Look at frequency/recency analytics

Starting the process

To start, I first needed to find out what the company wanted to find out. We established the macro and micro conversions. 

Macro conversion: Users upgrading their subscription from free to any paid account. 

The company’s subscriptions were; free account (where they could read 4 articles a month), standard (where they could read all articles) and premium (where they could read all articles and had access to data sets and graphs). When a user hit their ‘limit’ they hit a paywall, which prompted them to upgrade their account to access the page they want to visit. 

Micro conversions

Process Milestones (that lead to a macro conversion)

Users subscribing for free

Users reading articles

Users clicking on particular pages (like the subscription pricing page and pages only available to premium users)

Secondary actions

Users browsing the site

Users subscribing to newsletters 

Users clicking on marketing (like social media posts and links)

After this, we developed some questions, referring back to the macro and micro conversions, we came up with these: 

Engagement: Are more engaged users more likely to upgrade their subscription? 

Paywall hits: Which paywalls are the most effective at gaining conversions?

Triggers: Are there any other triggers which cause users to subscribe?

Regional differences: Are there regional differences in subscriptions?

Hypotheses

Then I developed the hypotheses in a structure that would be simpler to answer using analytics. So let’s focus on one hypothesis that showed some interesting results: 

We believe that more engaged users are more likely to upgrade their subscription. We will know we’re right when we see the segments that are ‘engaged’ upgrade their account from free to paid or standard to premium or from a monthly to annual subscription more than the unengaged segments.

From this, I started working out how to answer this from analytics data. Firstly, I needed to divide users by engagement. 

So I created segments within GA:

New engaged users

Returning engaged users

New unengaged users

Returning unengaged users. 

And to define condition to put particular users in each segment, I did some research and asked some colleagues. This is what I defined for the segment conditions. 

New engaged users 

Count of sessions =1

Session duration ≥ 60 

Unique pageviews ≥ 4

Returning engaged users

Count of sessions >1

Session duration ≥ 60 

Unique pageviews ≥ 4

Days since last session <30

New unengaged users

Count of sessions =1

Session duration ≤ 30 

Unique pageviews ≤ 2 

Returning unengaged users

Count of sessions >1

Session duration ≤ 30 

Unique pageviews ≤ 2 

Days since last session <30

This segment conditions defined for us what an ‘engaged’ and a ‘unengaged’ user is, because of the nature of our site (a news site) we needed users to stay for over a minute and visit at least 4 different pages to be ‘engaged’ as this is a sufficient amount of time to read at least one article and explore a few different pages on the site. I also made segments separating the new and returning users to see how their behaviour differs on the site.  

The results were interesting, the segment ‘New engaged users’ were the ones most likely to subscribe, proving the hypothesis was correct, but was strange because the numbers were lower than ‘Returning engaged users’. This showed me that analytics have many factors that can cause differing or unexpected results. And making educated guesses was okay. From the research, I had gathered that either these returning users may have wanted to be convinced more before signing up to the site, or because User ID within analytics is not set up (due to GDPR reasons) therefore someone might actually be a returning user but they are recognised as a new one. When making educated guesses, it is important to prove your theory, and in this case, I showed the reverse goal path (the path before hitting the goal of subscribing) and it showed that some users were going straight to subscribe or verifying their account, so they had the intention to subscribe from another session. 

From this research, I also found that engaged users were much more likely to upgrade their account rather than paying for a paid subscription straight away. Users were trying everything they could before hitting a paywall, they were limited, as they were on a free account. So I suggested moving the paywall limitations and having a free trial of standard or premium for new customers, as it could be a way of increasing conversions.

Further research of paywalls

This then led me into more research focusing on paywalls. And the other research from my first month led to more questions that needed to be investigated. Here I will focus on one.

Are all the paywalls having the desired effect?

Using mix panel I investigated the paywall hits. From previous research, I learned that the only paywall that was successful in gaining conversions was the paywall on articles. It makes sense, users browse the site, read a few articles, then they are engaged but then hit their limit of 4 articles, then this paywall gains a conversion because the user wants to continue reading. But I needed to find out if the other paywalls were also having the same effect, were they also gaining as many conversions? In short, no. The other paywalls weren’t effective, which was shocking and an unexpected result. But this led to a simple solution, not all the paywalls were needed, and some things had to change for there to be an increase in conversions. 

Suggesting solutions

After conducting my research and having all my collected findings, it was ready to present and suggest changes that would lead to a change in conversion rates. I presented my findings to the heads of departments and C-level executives. All of my findings came to two main conclusions. That most of the paywalls were ineffective and should not be on the website anymore, as they were too restricting of users wanting to explore the site. And a free trial of standard and premium subscriptions was needed to be available to new users so they could explore the site before making up their mind about purchasing a subscription.