Scrolller Weekly Update on Shareable Collections, Content Recommendation, and More

Scrolller Weekly Update September 15th, 2021

Hello and Welcome to our weekly update. This week's main highlight is on resolving blockers for shareable user collections, prioritization of user experience issue tickets, pushing faster hosting for premium users one step closer to reality and content recommendations. If you are new to the newsletter, subscribe to keep up with our journey of building a 100 million MAU web application.

User Highlights

Here's a recent feature suggestion by Urcheon on User Highlight this week. A function that indicates a collection's popularity in the search results so that users can quickly determine which collection is the best. The feature would be useful when browsing a list of similar collections, and more. Thanks for the suggestion @Urcheon. The dev team has it in consideration and working towards ensuring all users have great experiences with Scrolller.

Don't hesitate to let us know if you would like to share any feature feedback and contributions. We greatly appreciate you taking the valuable time to offer us feedback. Many thanks!

Go Here to help us with Scroller Features testing.

Also, we'd like to meet you if you can help us translate Scroller into Russian. Please get in touch by signing up below. Thank You in advance!

Sign Up To Help → **https://airtable.com/shrrRkFqMSZBqF24Q**

User Accounts

257447 → 262422

User Accounts Growth

We observed a definite increase in user account creation, but we are still only halfway to our prior level, and we are continually working hard to get back to where we were before the significant dip. While effective, account locking categories negatively affect users’ experience. We discovered that users that arrive at the categories page and encounter the account lock are bouncing at a high rate. We would like to unlock categories but in a way that preserves the experience.

Unlocking Categories plan

In last week's newsletter, we discussed a few methods for unlocking the categories that will not only prevent users from being bounced when they try to immediately land on the page but will also not have a detrimental influence on new user account creation on the site.

We shared various mockups of our account lock popup with our community last week to see which ones they wanted to see in the Scrolller, and we found a clear winner. In the coming days, the categories page will be unlocked according to the following criteria and users will be able to see the design they choose implemented on the site.

• After X number of visits from a user, lock other visits and show a CTA popup to educate users that they need an account to unlock this feature.

• The winning mockup for unlocking categories CTA is displayed below.

In the next newsletter, we will show you some mockups of these potential ideas. Please let us know in the comments what do you think about these ideas to keep the categories page unlocked.

Impact of Pro-Metric Ads

In the last newsletter, we talked about the steps we're taking to evaluate the impact of pro-metric ads in terms of dollars earned from conversions via these ads. Our developers have added the necessary labels in Google Analytics and we have started to see some specific data about the return on our investment in the pro-metric ads.

The new label started reporting data on 13th September and in 3 days we found that 22 people actually subscribed via pro-metric ads. That adds approximately $44 to Scrolller revenue in just 3 days. We will keep tracking the results and will share them with you next week.

Progress on Content Recommendation

Over the last week, we ran an analysis of our recommendations as well as a speed test of the system for the production of recommendations. in terms of speed, our recommendation system is more than capable of handling our level of users. It can't run real-time, but we will be able to start offering content recommendation emails. This will help us test our systems before we start to integrate this into other parts of our website.

This week, we attempted to assess the quality of the recommendations using the certainty analysis. However, we discovered that certainty is not a good predictor of quality. To solve this issue, our machine learning expert began screening the recommendations one by one. This is not the most scalable solution for evaluating the quality of recommendations, but for projects not yet deployed to production, it is a good habit to manually inspect your model’s output. This important step will make clear the model’s strengths and weaknesses. Once we deploy models to users we will use their engagement to help improve the quality of the recommendations.

We also discovered that the system can generate good recommendations even if you only have one favorite item. There are still a few difficulties to work out for users with little data. Our experts designed the algorithm to generate a new set of recommendations for each user every time.

Current Step

• Generate 10 great recommendations for our users based on their likes for content recommendation emails.

• Create metrics for evaluation of success for those emails.

Next Steps

• Integrate the content recommendation system in Scrolller as a "More Like This" feature.

User Experience Issues

Scroller's development team began prioritizing and working on the user experience issue in the queue this week. Scrolller believes that every time a user has a problem with the user experience, we lose an opportunity to convert that person to a premium subscriber. We prioritized and began working on the following high-priority bugs.

• Content tiles disappear while loading.

• The whole column stops loading if a video fails to render.

These problems are being worked on and will be fixed this week. We want to give our users the best experience possible, therefore before we begin working on enhancements, we want to solidify the foundation by making it completely bug-free.

Please let us know in the comment if you have encountered an issue while using Scrolller so, our team can prioritize and work on it.

Roadmap for Growth for User Accounts

• Unlocking Categories plan (Design Phase)

• Content Recommendation (In Progress)

• SFW Categories (In Progress)

• Scrolller APP (Back Log)

• Scrolller Extension and Enhancements (Blocked)

• Pro-Metric Advertising (Completed)

Revenue

Premium CTA popups and subscriptions

The premium CTA popup's impact on driving premium subscriptions has stabilized at 30+ new subscriptions each day, with a single-day high of 63 new subscriptions. Last week, we witnessed a slight decrease in its influence, with an average of 27 subscribers every day for the week. We'll keep experimenting with the copies to see which one outperforms the other, and we'll look at how we can improve the impact of premium CTA popups even more. The graph below shows the number of new subscribers in the last four weeks.

A/B Testing Copies for Premium CTA Popups

This week our developers have added the necessary labels in the google analytics and we are waiting for the week to see the results. The addition of these new labels will help us to measure the true impact of Premium CTA popups in terms of the dollar amount. The labels started to report the data from 13th Sep. We don't have any conversion via the premium CTA popup up in the reported data. We will continue to look into the data this week and will come up with an accurate number for how much revenue is brought to us via premium CTA popup.

Payment Processor Problem

We contacted high-risk payment providers to determine whether or not our business is indeed high risk. Not surprisingly, the high-risk payment providers believe that all digital entertainment sites are high-risk businesses. Unfortunately, the pricing structure of these high-risk payment providers is prohibitively expensive for the Scrolller, since they can charge up to 15% of your weekly revenue.

Next Steps:

• We are contacting a third-party PCI DSS certified vault that can store our customer data and supports different payment gateways. This will allow us to remain PCI compliant while also avoiding the risk of losing our clients' subscriptions.

• We are aiming to improve the NSFW experience for our users so that we are not categorized as a high-risk adult business and can work with standard payment providers that are not only cheaper but also have superior support. In the future newsletter, we'll go through our new NSFW experience, which will be personalized to the users' content preferences.

Let us know in the comments what you think about our plans for dealing with the Payment Processor Problem and feel free to share other payment processors you feel will work well for our situation.

Progress on Faster Hosting for Premium Users

In the last newsletter, we talked about how separating the servers for premium and non-paid users will allow us to give our premium users a speedier experience. The enhanced infrastructure will be available for dev testing by the end of this week, and we want to stress test it to ensure that it is ready and delivers on the promises we make to our premium users.

Our developers are collaborating with our data scientists to define the key measures that will be utilized to assess performance improvement. In our future newsletter, we'll go into more detail regarding that metric and our testing.

Roadmap for Growth in Revenue

• Faster Hosting for Premium Users (In Progress)

• Exclusive Features for Premium User (Design Stage)

• Multiple Payment Processors (Design Stage)

Social Traffic

User Generated Collections Progress

Shareable favorites along with shareable follows are moving along in the testing phase and we are refining these two features so when we launch it is absolutely bug-free and ready to be used by end-users.

While resolving the bug that was reported last week our developers ran into a blocker that they cannot merge the user-defined collections and regular collection on follows page as they have different data structures and implementations. Our developers suggested creating a separate page for user-defined collections but it will bring more complexity.

Scrolller believes that content is the king and users should be able to enjoy it without any hassle or complexity. Our developers are working on creating a homogeneous experience for our users and hopefully, this blocker will be resolved this week.

Progress

• Shareable Favorites collection (Bug Fixes)

• Shareable Following collection (Bug Fixes)

• Leaderboard (In Progress)

• Incentives for top 100 users in the leaderboard. (Backlog)

Social Media Update

As highlighted last week, our social channels are picking stable growth with an increased no. of followers. On profile reach, our September Tweet impressions have reached 42.5K so far with the top tweet earning 1,464 impressions. We however saw a reduction in reach on Facebook and Instagram which reached 25.1k and 24k respectively. We'll keep on testing and experimenting with social strategies to see which ones are effective in driving traffic to the Scrolller website.

🔘 Instagram 🔘 Facebook 🔘 Twitter

Thank for you being our fans/followers on social media. If you have strategies that have worked for you on social media, you're welcome to share them with us in the comments section.

Roadmap for Growth in Social Traffic

• User Generated Collections: Create and share your favorites and follows (In Progress)

• Facebook Automation (Back Log)

• Twitter Automation (Back Log)

Team Highlights

Issues Refinement and Estimation Sessions

From this week our developers and product owners have started an Issue refinement and Estimation session. It will not only help with estimating the deliveries but also will help the developers to point out what possible blockers they might face in the development of certain features.

It will help reduce rework in development and testing but also identifies the dependencies within the team and helps to foresee risks. Our developers have welcomed the change and they feel it will allow them to commit to the work for the whole week they can get done and will help in transparency across the team. We believe that regular refinement and estimation sessions will help increase the throughput of the team.

Team Updates

The Scrolller Team Highlights for the week!

Mannan — Catching up with the progress of projects as well as user experiences to resolve any blockers.

Nick — Completed front-end setup, API, backend server, and database. Also, updated Prisma schema to make it compatible to run it on new hosting.

Karol — Enhancing display collections and shareable favorites

Valentin — Social media management, strategies & community support

Tim — Finished analyzing the recommendation system to get the selection right for most users and wrote a first version of the recommendation pipeline.

Matt — Working towards finding metrics to measure product performance

Betty — Social media strategies, management & community support

Marité — Working on improving UI/UX designs

Chris — Investor in-person and engineering meetings as well as weekly team management activities.

Have any feedback for us? Don't hesitate to share it with us in the comments sections. We like hearing any feedback you have for us.

Road Map

You can share feedback with us about any of the new features below. Using this Airtable Form. Looking forward to your comments

This week

• Improve quality and user experience of pro-metric advertisements (Completed)

• User Generated Collections: Create and share (In Progress)

• Content Recommendation (In Progress)

• Turning off old Scrolller (In Progress)

• Exclusive Features for premium users (In Progress)

• New native advertisements (Back Log)

• Monetization Update and Optimization (Back Log)

• Categories page for SFW collections + Unlock (Blocked)

• Enhancements to Auto Scroll Feature (Blocked)

P.S. As a reminder, subscribe to our newsletter if you are yet to, walk with us on this Building Scroller exciting yet challenging journey.

Sign Up To Help → https://airtable.com/shrrRkFqMSZBqF24Q us translate Scroller to Russian.

To help us with Scroller features testing, Go Here

Until next time, happy scrollling!