Is it possible to grow by adding new features? Why not dig for treasures from old features?

Analyze user behavior, identify key features, optimize penetration, and increase user growth and retention.

Innovation seems to be a key word written into the mission, vision and values ​​of every company. In the process of product development and user growth, people are often excited about launching new features. Indeed, new features from 0 to 1 are more likely to attract new users and increase user activity in order to meet the latest needs, but they are also accompanied by high costs and high risks.

Judging from the experience of well-known technology companies such as Facebook, Netflix, Airbnb, Duolingo, etc., big growth comes from less eye-catching things: gradual and continuous optimization of core products, which can also be understood as innovation from 1 to 1.x.

Today, we discuss how to achieve user growth by optimizing and tapping into existing features: Why? How?

Whether you are an entrepreneur, product manager or marketer, you can get valuable information.

Without further ado, let's go straight to the main text. Enjoy:

01 Why growth doesn’t just rely on new features?

When we ask most product teams how they plan to drive more growth, we often hear the same answer: we need more features: more features will satisfy more user needs, which will lead to more acquisition and retention, which will in turn lead to more revenue. This is the conventional mindset.

But if we think about it carefully, where is the loophole in this logic?

First, in a clear scenario, the continuous addition of functions may make the product complex, even blur the positioning, and reduce the user experience. For example, in addition to the core service of "listening", QQ Music has been constantly adding new functions, continuously releasing new functions from social to metaverse to games, and users are asking "when will the 3A masterpiece be released?"; Alipay replaced the function at the bottom center with short videos. A tool software with payment as its original intention has also embarked on the mode of "killing time to retain users". Perhaps Alibaba is also building its own "video account".

Second, developing new functions requires a lot of R&D investment, marketing, operations and other resources to promote, which affects the optimization rhythm of existing functions. In the context of the Internet's "cost reduction and efficiency improvement", is it more appropriate to use resources where they are needed most?

Third, some research on new features may be "wishful thinking" of developers: keeping up with the trend, competitors have also done it, the super app route is not wrong, I think users will like it... New features require users' time and energy to adapt, and acceptance may not be as expected.

02 Why is it more effective to mine old features for growth?

The most important features in a product are usually the ones that are targeted first. Once a product has completed the MVP (minimum viable product), many subsequent new features are usually only targeted to a small group of users. Therefore, it is more likely to increase product impact by exploring and improving engagement with features that have already been released than by focusing on features that have not yet been prioritized.

Another reason is that we may not have yet discovered the maximum utility of old features. When you first release a feature, you often find that 80% of the feature design works well, but the remaining 20% ​​doesn’t. Some users may get stuck at steps they find confusing or require a lot of effort, and fail to complete the usage process and churn. Every layer of the funnel that uses the product is a treasure trove of data that can be improved exponentially after optimization.

For example, Netflix has continuously optimized its algorithm over the past 20 years, increasing the number of movies that users can choose from from 2% that are automatically recommended by the system to 80% today. Previously, a user had to search through hundreds of titles to find a movie they like, but today, most users only need to look at 40 titles before they can happily click the "Play" button.

03 How to efficiently mine old functions—ARIA analysis framework

If increasing engagement on existing key features is often more impactful than adding new ones, the next question is: How do we mine old feature treasure troves in a structured, repeatable way?

Ken Rudin, summarizing his growth experience at Salesforce, Facebook, Google, and ThoughtSpot (an AI-driven analytics vendor), discovered a useful ARIA framework.

ARIA has four key principles: Analyze, Reduce, Introduce, and Assist. For an existing functional analysis, this is not a one-time process, but a growth flywheel of continuous optimization.

Principle 1: Analysis

First, we needed to figure out which features to focus on. This meant analyzing the data to determine which features were most correlated with growth, and then analyzing the penetration of those features to highlight which ones weren’t getting the adoption that they deserved from users.

"Key functions" - those functions that are most relevant to attracting new users, conversion, retention or fission . Increasing the penetration rate of these functions will have the greatest impact on our business growth.

For example: YouTube’s “Subscribe” feature notifies users when a channel they subscribe to releases a new video. Their data analysis showed that subscribing to a channel is strongly correlated with user retention, because notification updates are an important means of recall. So they made this a key feature and found a better time to increase the penetration of the “Subscribe” feature: it was better to wait until users watched the channel on several different days before reminding them to subscribe, and clearly emphasized the benefits of subscribing to the channel (rather than assuming that users would obviously feel the benefits).

Through such optimization, the number of users subscribing to the channel increased by nearly 10%, and the first-month retention rate of these users increased by about 8%.

After determining the key feature sets, the next step is to collect data about their existing user penetration. For example, how many YouTube users have subscribed to the channel? What is the usage? How to determine the usage points of different users? You can refer to the following table for data analysis:

Final tip: When analyzing these metrics, stratify users based on the time period they have been using our product (e.g. 0-30 days, 31-90 days, and 91+ days): This can help us understand which features are recognized in the early stages of user contact with the product and which ones need more time.

If a feature is already being used successfully by most users, there is not much room for improvement. Once we have identified the features to focus on, we need to increase their penetration.

Principle 2: Reduce

The first effective way to increase the penetration of a feature is not necessarily to increase promotion, but to reduce the friction associated with using it: reduce the number of steps, reduce the effort required to complete each step, and reduce the cost of understanding.

For example, in Calendly (a software that helps schedule meetings), we can improve meeting availability. Users have multiple types of meetings (internal meetings, client meetings, 1-1 meetings), each with different scheduling requirements. But if you don’t have to tediously schedule meetings for each day of the week, but instead set the duration of one type of meeting and then copy it to other days at once, you can save a lot of steps.

Once you minimize the steps required to use the function, check whether you can reduce the effort required for each step in the remaining steps. This is not limited to common measures such as optimizing UI processes, imitating similar processes in other popular software, and optimizing descriptions. You can also try to let users edit instead of creating from scratch, intelligently pre-fill default values, and let users click to select instead of typing.

For example, in 2023, video messaging tool Loom launched an AI feature, Auto Message Composer, which can automatically summarize videos and share them with colleagues, which is why Loom promoted it so heavily through free trials.

Finally, we also need to find out which features in the product confuse users. We need to capture some new users and observe their hesitation points when they try these features for the first time. Avoid some common terms/common logos in the industry, which look very advanced but are confusing to ordinary people.

Principle 3: Introduction

Reducing the friction associated with a feature is critical, but it only solves part of the problem. If no one knows about the feature, then there will be insufficient traffic at the top of the funnel and conversion data at the bottom will no longer be significant.

So, we need to raise users’ awareness of key features and motivate them to use them: not just introduce newly launched features, but all key features.

The simplest way is to remind key functions during the first use and re-use. Common UI elements for promotion: pop-up windows, banners, red dot reminders, etc. But the key question is not how we should remind, but when.

We often rely on “onboarding” to introduce key features, but the amount of features we can show is limited and its effectiveness is questionable when new users feel overwhelmed. (Perhaps they will subconsciously close the pop-ups, especially when they completely take up the screen and block the main flow.)

There is a better way. The right time to introduce a new feature is when users are most likely to want to use it. When users are told about a feature when they need it, they are more likely to use it and remember it than if they are just told about it.

For example, Google Maps introduced the "Put Stops" feature (the ability to add stops for gas, food, or rest stops to an existing navigation route). When the feature was first launched, they displayed a pop-up window to all users after opening the app, but the penetration rate was average. No one needed to add a temporary stop immediately after opening the map, so users closed the pop-up window and quickly forgot about the feature.

Later, they changed the feature to introduce it immediately after users searched for long-distance directions, because that’s when they were most likely to be interested in adding a stop. Even though this meant only people on long drives discovered the feature, penetration increased by nearly 15% almost immediately.

To increase users’ awareness of features and motivate them to use them, another important strategy is to let other users introduce these features to them. Common strategies include: old users bring new users to referral mechanism, sharing process and interface optimization.

Principle 4: Assistance

If we have lowered the friction of using a feature and brought in more users, they may try to use the feature but run into problems, and this is when the Assistance principle comes into play. Assistance is not only about helping users solve the problems they encounter when using a feature, but also another way to encourage them to try using the feature.

Common solution 1: The product has a built-in help center to provide answers to common questions and use intelligent customer service to help answer users' personalized questions.

Common solution 2: Strengthen the construction of educational and guiding content for complex functions , such as blog posts and video tutorials. However, the most important thing is that we need to decide which method is most suitable based on the user's usage and the complexity of the product.

Common solution three: Incentives for users to try features , incentives such as coupons, discounts, or rewards. Although incentives can be very effective, they are usually only suitable for short-term goals (such as boosting sales). Therefore, we may also need to use other methods to motivate users, such as creating emotional links (such as through emotional content or stories) or providing social recognition (such as through user ratings and comments).

04 A successful case of practicing this principle - duolingo

Duolingo is a magical existence. As a language learning platform, it has won the love of users in a less popular track. In the past five years, Duolingo's DAU (daily active users) has increased from 5 million to nearly 30 million (a 6-fold increase!)

The company's stock price has nearly tripled in the past two years, thanks in large part to methodical product iteration and experimentation. Its public winning formula: winning streaks, notifications, and leaderboards.

Although gamification is embedded in the learning process, Duolingo still keeps the product clean and simple as its principle, and optimizes each function in a restrained and effective manner. Today, let's take a look at how Duolingo digs out growth treasures from existing functions.

1. The treasure of winning streak mechanism

When a product manager on Duolingo’s retention team discovered that users were significantly less likely to quit if they logged in for 10 consecutive days, they thought the finding was interesting enough to be worth further optimization, even though it was largely just correlation and selection bias.

The first major advancement they made was the introduction of “Streak Protection,” a notification that alerts users who are about to lose their login streak.

After that, they launched a number of improvements: such as calendar view, animation effects, adjustment of the continuous login "freeze" function and continuous login rewards.

These improvements all build on the existing streak concept and significantly improve user retention. One of the reasons is that it motivates users over time: the longer the streak, the stronger the motivation to maintain it. From a broader perspective, their success with streaks also shows us that we can achieve major breakthroughs with existing features.

2. The Treasure of Notification System

Through a large number of A/B tests in the past few years, Duolingo found that push notifications are an important means to promote growth. However, this effect has stabilized over time. Therefore, they gave the push team the freedom to optimize the sending time, templates, pictures, copywriting, localization, etc. of the notifications, and tested the effects of different messages through algorithms to maximize the click-through rate.

Other companies bombard us with spam-like messages that can get annoying or even turn off notifications, but Duolingo makes it so that every notification it sends feels like it’s tailor-made for us, and always appears when we’re most likely to click on it.

This kind of anthropomorphic notification has also sparked a hot topic of "encouraging learning" on Xiaohongshu and Bilibili.

3. Treasures of the Leaderboard

When it comes to gamification, leaderboards are definitely a common feature. The Duolingo team not only recruited talents from the gaming industry, but also drew on successful leaderboards from other games, such as Gardenscapes, Golf Clash, Toon Blast, etc.

And the most effective leaderboard (its fourth iteration) is an opt-out experience that algorithmically groups people into a new group of 30 each week, with some of them promoted or demoted to higher or lower leagues each week, and leagues automatically adjusting people's difficulty level to make every rank feel engaging.

It took a lot of product and design effort to adapt these ideas to Duolingo’s simple, introductory system; the ranking system was the most complex feature they’d added, but they had to make sure people could figure it out without instructions popping up.

Fortunately, it did have a very good effect: D1 retention increased by 1%, D7 retention increased by 2%, and D14 retention increased by 3%. In addition, people's learning time increased by about 17%, and once the weekend came, they were "rolled" on the leaderboard.

05 Conclusion

Innovation is not only from 0 to 1, but also from 1 to 1.x.

When "reducing costs and increasing efficiency" becomes the main theme of the new stage of the Internet, innovation also requires low costs and high returns. Improving the penetration rate of existing key functions may be more influential than launching new functions.

The above is all about "Exploring Existing Features to Accelerate Growth". The reference comes from growth work practice & research. If you have different opinions, please leave a message in the comment area below to discuss.

*Reference: Ken Rudin "How to accelerate growth by focusing on the features you already have"

Author: Jiayu Teacher

WeChat public account: Notes on adding fish and meat (ID: gh_e5033136a891)