Textbook-level user recommendation system.
Editor’s note: This article is from WeChat public account “All Media” (ID: Quanmeipai), author of Tencent Media.
Ad traffic is too expensive? SEM also increased prices? Through social channels, make good use of existing users, and bring new ones to achieve your unexpected results.
Simple “inviting and courteous” gameplay, played by the newsletter company Morning Brew. In 2017, it had only 100,000 subscribers, and in just 18 months, its number of users increased to 1.5 million, of which 30% came from user recommendations.
The secret of Morning Brew’s success is that it has designed a recommendation system that turns existing users into walking billboards, motivating them to recommend this product to others, resulting in a large number of new users.
This issue of the whole media (ID: quanmeipai) brought the re-distribution article of the product manager of Morning Brew, Tyler Denk, to deeply dispel the ideas and methods of user growth.
Morning Brew is a company that provides daily newsletters via email, providing young professionals with fun and interesting business news, covering the most cutting-edge information from Wall Street to Silicon Valley. Tyler, the company’s No. 2 employee, must have multiple roles and work as an engineer, product, and user.
In 2017, Tyler’s goal was clear – to raise users at all costs. At that time, user growth relied on word-of-mouth communication and the tireless efforts of the two founders. After that, Tyler focused on the existing 100,000 users and used their value to create the world’s most efficient recommendation tool.
Primary Rule: Do Growth, Start with Products
Before you decide to grow, there are two things to consider:
1. Does this product have value to users/consumers?
2. How do you build a growth mechanism in your product?
Morning Brew spreads naturally through word of mouth before adopting the recommendation mechanism. When the product is so good that you can spread it spontaneously, you can start to build a recommendation system. on the contrary,Trying to create a recommendation system for a product that is not good or useless is equivalent to pouring water into a bucket with a hole underneath.
With regard to recommendations, there are many success stories:
-
When users recommend HQ Trivia to someone else to play, they can get extra life.
-
Get free storage when you recommend Dropbox to someone else.
-
Harry’s Razors (Note: A razor product) has created a recommendation system that provides free products before the product is released.
-
Wealthfront (Note: A smart investment advisory platform) has an invitation system, and both the referee and the referee can get a free credit of $5,000, which is effective.
Morning Brew’s recommendation mechanism is based on “milestone achievements” (note: recommending products to several users and receiving rewards), and in each newsletter sent out, there is a special “share us” ( Share the Brew).
Share the Brew section example (editor screenshot from October 14, 2019)
Technical Logic: Recommended Process & Vulnerability Protection
Morning Brew’s recommendation mechanism is tailor-made, it uses the Ruby language and works with the email service provider Sailthru.
The mechanism flow is as follows:
1. Each subscriber has a referral code associated with the account, which is unique.
2. When using this code, Morning Brew will generate a share link.
3. When someone else (assuming Subscriber X) registers with your referral link, X will be permanently associated with your account. You will become the “recommendator” of X.
4. Morning Brew uses a two-way authentication mechanism to ensure that the email address is trueeffective. When subscribing user X to register, you must log in to the mailbox to confirm.
5. Once the confirmation link is clicked, the Sailthru mail system is triggered to respond and send an API request to the database.
6. The request will prompt the subscriber X to just confirm the email, meaning that you (the referee) should be rewarded for this.
7. Your referral count will be increased by 1, but you may not receive a reminder email for this number update.
It took a long time to build this system. In the beginning, Morning Brew didn’t need to subscribe to User X to confirm the email address, because it would add to the registration. However, in order to maximize the company’s interests, Tyler made a change. Currently, 85% of recommended users will be authenticated at the time of registration.
Although this means losing 15% conversion rate, it also ensures that higher quality users are participating in the subscription. When building a recommendation system, you need to make various considerations, such as profit, time, and so on.
At the same time, be aware of system vulnerabilities and establish defensive measures. Providing free rewards will always attract people who want to take advantage of the loopholes. In addition to requiring users to authenticate their mail, there are several other defenses:
1. Use a third-party email authentication system to authenticate every email address entered into the ecosystem.
2. When a user arrives at a new milestone, the company receives an email listing the list of recommended mailboxes (it is easy to identify fake or obsolete email addresses).
3. The company has a blacklist of 600 fake email addresses. This number is still growing.
4. Email addresses that are blacklisted will no longer be able to recommend others.
5. Disable specific IP addresses to submit referrals.
In addition, the system also has a management panel that adjusts when there is an error in the recommended count to ensure that the user can get the rewards they deserve.
Reward Mechanism: Milestone Progression & Very Low Customer Cost
In a few years, Morning Brew received a lot of real feedback from readers and conducted research and discussions in Facebook’s Morning Brew area. The current “milestone”-based reward system was born after careful calculation and evaluation of demand.
Morning BrewReward, the number indicates the number of referrals
Recommended to 3 people:
Users receive a reward called “Light Roast,” a dedicated Sunday newsletter.
The effectiveness of “light baking” is reflected in two aspects:
1. Additional exclusive content is likely to match the user’s interests;
2. There is no additional cost for the company, no need to buy or mail physical items.
Daily newsletters are pushed from Monday to Saturday, and “light baking” is sent every Sunday. Currently, the latter has more than 75,000 subscribers. This means that more than 75,000 people have shared at least 3 people with Morning Brew.
Recommended to 5 people:
Send the Morning Brew sticker to readers.
If you sell 10,000 stickers from StickerMule, the cost of a single sheet is $0.2, plus a mailing fee of $0.65 per pack. Get 5 users for a total of 1.25 US dollars, that is, the cost per new user is 0.25 US dollars (about 1.75 yuan).
In comparison, the largest users get access to Facebook and Instagram, and each new user gets a cost of $3 to $5 (about 21 to 35 yuan).
Recommended to 10 people:
Readers can enter the exclusive “inside” community. Morning Brew’s private Facebook team is about to welcome 10,000 members, where they discuss the latest business news, trends and events, as well as job opportunities and readers with similar ideas. For Morning Brew, this reward is also zero cost.
So far, in order to get 10 new users, the company has provided member content, exclusive community and stickers for a total of $1.25 (about 8.75 yuan).
Recommended to 15 people:
Mail a Morning Brew silicone mobile wallet for these users, which costs $1.50 and the cost of mailing is still $0.65.
When recommended to 15 people, each user’s acquisition cost is 0.23 US dollars (about 1.61 yuan), that is, (sticker + wallet) ÷ 15. The more people you recommend, the lower the cost of new users.
Recommended to 25 people:
Send a super soft Morning Brew T-shirt to readers. The company initially hoped that the T-shirts were as cheap as possible, the results were very uncomfortable, and no users were willing to wear them. A year ago, the company decided to buy the best t-shirt. Today, everyone likes this T-shirt very much, and at least four people wear it every day in the office.
Recommended to 50 people:
Send a reader a coffee cup with the words “Rise and Grind” (fan favorite).
Recommended to 100 people:
Send a “Brewneck” (MorningBrew Sweatshirt). Like a T-shirt, it is quite comfortable.
At the end of the day, the rewards program is designed to motivate readers to share products. In exchange, they will receive content, community and physical gifts. Gifts are also a marketing tool. When someone wears a Morning Brew short-sleeved shirt or drinks coffee in a Morning Brew cup, it is actually increasing brand exposure.
The recommendation mechanism also increases the level of engagement for users who make recommendations. If someone shares a product or service to a friend, classmate, or colleague, their chances of continuing to use the product increase.
Recommended Center: Education, Incentives, Helping Users
When technology and reward mechanisms are in place, the next step is to build a center. Here, readers can learn about rewards, see the progress of recommendations, and use tools to share products smoothly.
In a broad sense, thisThe main function of the page is to educate, motivate and help users share products.
Screenshot of the Desktop Recommendation Center
Education User
First of all, users need to know that there are reward plans and their operational logic, and this plan will work.
At the top of the page, the product shows the user the value of sharing, such as sharing the product will be rewarded. After that, the page shows each reward and how it is obtained through beautiful pictures.
Create a recommendation mechanism that looks at the problem from the perspective of the user. The readers are very busy. Most people read the newsletter in the morning, on the way to work, or at the desk to start a day’s work. It is important to make the reward clear, intuitive and repeatable.
Inspire users
The product will show the user a real-time recommendation count, encouraging them to: “Only getting the target Y is only X recommended!”
Before, the product tried to use the progress bar to pass the count information, and finally decided to replace it with a numeric style. In fact, it was mainly for aesthetics, not because of data driven changes. However, in terms of efficiency, no significant difference between the two has been found. The ultimate goal is the same: update the user to progress and encourage them to reach the next milestone.
The progress bar used by the previous version of the recommendation center
Help users
Here, users can do their best to make sharing products easier. The product hopes to remove the obstacles in the user’s use.
First, the product provides users with a unique recommendation link and how to use it. The “Copy Link” button is user-friendly. Once copied to the clipboard, they can be posted to Reddit, shared with classmates, or sent to friends via SMS.
In addition, the product allows users to recommend via email. Users can manually enter someone else’s email address or authorize the import of contacts in the address book.
When inviting by mail, the product has provided a pre-written email template explaining Morning BreWhat is w and its value proposition. This way, the user does not need to spend time writing emails. However, the template content is editable and can be written by the reader.
Finally, referral links can be shared with major social networking sites and instant messaging apps. Click the Share button and the user’s exclusive referral link and message template will be pasted into the draft and can be shared immediately.
The options are not as good as possible, sometimes less is more.
For example, the login conversion rate recommended by email is 75%, and the other way is only 35%.
Mobile-side recommended options
Based on the above data, perhaps the product should only provide an email invitation option, but it may also be counterproductive. This is because it is very likely that only certain groups of people will be invited by mail, and most people are still accustomed to sharing through social networks. The company is further validating this hypothesis.
After analysis, it was found that the number of registrations shared through WhatsApp and SMS was 10 times shared by LinkedIn, 5 times that of Twitter and 2 times that of Facebook. For example, with WhatsApp, every 5 shares bring in 1 subscriber; on Twitter, 25 shares are required to bring 1 subscriber.
Data shows that peer-to-peer channels (such as SMS and WhatsApp) are more effective than one-to-many platforms like Facebook, Twitter, and LinkedIn.
landing page: multiple tests to increase conversion rate
When a reader shares a proprietary link with someone else, the product wants to: turn that person into a subscriber.
The product landing page is simple and intuitive to design. Users have two choices: registration or not. However, the landing page has some details to be tested, such as overall layout, title, subtitle, bottom text, style, button color, other images and user comments.
Users from different sources have different responses to this page. Currently, Facebook ads, blog posts, and articles mentioned in the article are all traffic sources, and they have different responses to the design, language, and features of the landing page.
Tyler uses Google Optimize (Editor’s Note: A Google A/B testing tool) to test every detail of the landing page, which optimizes those through the recommendation mechanism.The incoming traffic is especially important. The company divides test users into two categories: users who are invited by email, and users who are otherwise obtained.
First, the latter, through the improvement of the title, subtitle, bottom text and bottom color, the product conversion rate increased by more than 4%, which means more than 4,500 users per month.
For example, the new landing page emphasizes “friends recommend you to register”:
Original landing page
New landing page
The conversion rate for the new version is 3.1% higher than the original.
The same new and old design tests, the conversion rate by email invitation increased by 7%.
Google Optimize screenshots, testing different terms and designs
The conversion rate for email invitations is much higher than the other methods, which is why the Referral Center prioritizes the “Share via email” option and puts it above the social network share button.
“Share Us”: Refine the Conditional Logic of Incentive Rendering
Draw the entire recommendation system into a funnel diagram, as follows:
Recommended Funnel
Familiar with the specifics of the referral program, now you can start talking about the top layer of the funnel – the newsletter sharing area.
Each newsletter has a dedicated “Share Us” section to remind readers of the existence of a recommendation mechanism. For those who have already made recommendations, this section is to stimulate more recommended behavior.
The importance of this section is:
1. Every morning, the newsletter reaches 1.5 million users
2. About 45% of people will open newsletters
3. About X% of people will click the “Share” button
After the third step is implemented, the recommendation mechanism discussed earlier will become possible. If the reader does not click the “Share Us” button, then the many features of the recommendation center, the rewards provided, and the conversion rate of the landing page are meaningless. This means that the product needs to optimize the “X” in the third step.
A test version of the newsletter sharing area
There are several ways to achieve conversion goals. The common sharing area is similar to the recommendation center: it displays the values at the top, shows the readers the number of recommendations, and presents all the rewards in a visual form.
In order to reduce the barriers to the recommendation mechanism, the product provides readers with two ways to share. When the user clicks the “Share Us” button, they will jump to the recommendation center; or, the user can copy and paste the exclusive recommendation link. This is similar to the recommendation center approach that I talked about before. It is actually a simplified version of the recommendation center.
The company used Sailthru’s scripting language Zephyr, which adds more conditional logic to the email. For example, if the number of user recommendations is less than 3, it means that the TA has not received a “light baking” reward. Therefore, the sharing area will focus on the “light baking” reward. If the user already has 20 recommendations, the TA may see a T-shirt that is available for 5 recommendations.
“Light Bake” reward
T-shirt rewards
In order to maintain the freshness of the plate, the product often changes the content according to the user’s recommendation number, and the GIF animation and the display of the specific rewards take turns.
For example, if your recommendation is zero, you will see a “light baking” bonus on Monday, a sticker reward on Tuesday, a T-shirt on Wednesday, a coffee cup on Thursday, etc. .
Different people like different rewards, so try to make more changes. This also serves to inform the reader and motivate the action.
Marketing Email: Unlock Multiple User Behavior Nodes
In addition to daily newsletter emails, users will receive marketing emails. The main purpose of these emails is to make the recommendation system more effective. It can be divided into three main categories: initial breakthroughs, milestones and boosts. In addition, there are sweepstakes.
The initial breakthrough from 0 to 1
The reality is that most readers recommend 0. Convincing a person to recommend is challenging, so the product needs to focus on a mutation from 0 to 1. Once the reader has recommended it, they will also experience these:
1. Know the existence of the reward program
2. Incentives for reward programs
3. Know how to share their exclusive links
4. Harvest satisfaction
Although there are “share us” in the newsletter, this does not guarantee that the reader knows or understands the rewards program. To solve this problem, readers with a recommended number of 0 will receive an email 7 days after his subscription. This is based on the assumption that after a week of reading the newsletter, readers’ excitement peaks, and they are most likely to talk and share with others.
Tyler believes that there is a strong connection between the user’s freshness and the willingness to share: I love my iPhoneX, but after a few weeks, I no longer tell people how much I like it.
Let the first sharing of readers is the biggest obstacle, so the company tested this “seventh day mail” approach from different angles, such as light baking, stickers, showing all rewards, greetings from CEOs, offerings Starbucks gift cards, etc.
Using SailtHru, the mail system can test 5 versions of the mail at the same time, select the winning one, think about the reason for winning; based on this, think of 5 different versions to further improve the conversion rate.
5 test versions
The figure below shows the number of recommendations since the user self-registration, and the results are as expected.
After user registration, use the number of days associated with the recommended number
Only by observing a small percentage of users, you will find that this one-time email has a significant effect on recommendations. As long as you can increase the conversion rate by 1% through this email, you can increase the number of recommendations by thousands per month.
Milestone Mail
The milestone email will be sent after the user has successfully reached a certain number of recommendations. The recommended number is 5, 10, 15, 25, 50, 100, 1000. This is to let the reader know what they have achieved, how to get rewards, and motivate them to reach the next milestone.
This whole process is called the “Referral Pipeline”. Product hope, let users recommend 3 times, 5 times, 10 times after recommending 1 time… In short, maximize the value of user contribution.
In order to increase clickthrough rate, this type of auto-issued email goes through 5 tests.
Each message will contain the following information: Congratulations on reaching this milestone, thanks to sharing Brew… but do you know that only X recommendations are missing from unlocking the next achievement? How can you stop at this time?
For those milestone rewards that need to ship physical items, the company collects user addresses and related information through forms and encourages them to work toward the next milestone.
Boost mail
In the entire recommendation pipeline, the easiest target to achieve is “boost mail.” Those who have only one recommendation from the next milestone (such as 2, 4, 9 recommended) will receive an email if they have stopped for two weeks, encouraging them to get the final one.The number of recommendations.
For example, if a user stagnate for 4 recommendations for two consecutive weeks, they will receive an email “Hey, congratulations on recommending 4 users, thank you very much for sharing it with others… but you know you are away Is it only one recommendation number to get the sticker? If you have already come here, it is better to go one step further, right?”
Sweepstakes: Achieving a breakthrough of 0
Morning Brew will conduct a large sweepstake every 4-5 weeks. In the next X hours, each time a user wins a recommendation, they will receive a lucky draw ticket. The more people you recommend, the more tickets you get. This is not an innovation, but a huge stimulus for the recommendation mechanism.
Recommended Center will go online lottery activity map
As mentioned above, the biggest challenge for the product is how to make the user’s recommendation change from 0 to 1. A huge incentive will transform thousands of users and achieve a breakthrough of zero. In other words, those who have viewed it many times but have ignored “Share Us” many times now have enough motivation to click this button and jump to the Referral Center.
Once you enter the Referral Center, users will see the Control Panel and find that sharing is easy. With just a few clicks, you can share with many friends and receive encouragement emails. Although the possibility of winning is very low, they can “teach” the behavior they share.
Events include: Apple’s Mac Pro, iPhone, Samsung Galaxys, $2,000 in cash, travel in Singapore, a trip to New York with Brew employees and an exclusive visit to Nasdaq. Gifts must be fresh and have real value to the reader.
Summary down, these methods have undergone countless hands-on tests and are indeed effective in bringing user growth. Currently, more than 225,000 users have recommended at least 1 person to subscribe to Morning Brew. In the summer of 2018, the team averaged more than 1,000 recommended users per day;Since then, user growth has become more significant.
Although the product recommendation system has been effective, there is still much room for improvement. The team is evaluating the following questions:
1. What kind of rewards can the product offer, and it is also cost-effective while improving demand?
2. How can I improve the language in the recommendation system so that more users can participate?
3. Should I add more sharing options to readers, or should I reduce the options and simplify the process?
4. Can I combine the rewards system with social media marketing to present what the subscribers read, use the Morning Brew app, and the surrounding?
5. Is there any other place that needs to be boosted? Is there a redundant item that can be removed?
The optimization of any detail in the product can bring thousands of new users every month. The most important thing is to constantly test new ideas and choose the most appropriate and effective methods to achieve rapid growth.