Google, Airbnb, Amazon, Didi... all use conversion rate optimization secrets

Defined from an economic perspective, growth refers to the continuous change in economic facts, which means an increase or decrease per unit time.

In the Internet field, the increase and improvement of indicators related to revenue business can also be summarized into the category of growth, including the UV of the page, the number of visiting IPs, the user's visit time, stay time, and usage time. E-commerce or internal purchases of Internet products also include indicators such as GNV, LTV, average order value, etc.; the Internet social field has indicators such as sharing, likes, and reposts.

As for growth, the traditional way is to attract new customers, such as doing door-to-door marketing , providing subsidies and promotions to users, organizing some online and offline activities, advertising, and directly purchasing traffic. However, traffic is becoming more and more expensive, costs are rising, and continuous investment has not produced any actual value. This is an unsustainable way of growth.

Customer loyalty gained by stimulating user behavior is often low, and the user experience will also be poor. New or relatively mature Internet companies will use a method called growth hack to achieve growth.

In the long run, growth hackers focus more on customer experience, namely retention, activation, and word-of-mouth recommendations. This is called the Pirate model, the A AR RR model. The overall user growth model is divided into five parts: new user acquisition, activation, retention, word of mouth, marketing and revenue. Continuously optimize and improve these five links to achieve growth.

AARRR Pirate Model, 5 Steps to Improve Conversion Rate

Step 1: Attract new visitors

The conversion rate of a new user's first visit is 100%, which is all traffic, but not all visits are valuable. If the user stays on the page longer during the second visit, we can consider this to be a good new user. This type of new user may not be 100%. Only 70% of user visits are converted into relatively good visits. This type of user who has not abandoned us is only worth five cents.

Step 2: Activation

The conversion rate from attracting new users to activating users has changed a lot. Only 30% of users have taken many actions on the platform, so the conversion rate is only 30%. For users who have a good experience like this, the value may be higher, perhaps 25 cents. The number of users who can generate registrations may be even lower. Among a hundred visitors, only 5 may register, so the registration value of each user is one dollar. What would happen if the registration conversion rate increased from 5% to 10%? The final value can be doubled.

Step 3: Retention

Retention means that users not only sign up, but also use the product repeatedly. There will be a lot fewer such users, for example, 3% or 2% of the total users, and their user value is higher, probably several dollars. If we optimize the retention process from 3% to 4%, we can increase the business volume by one third.

Step 4 & 5: Word-of-mouth marketing, converting into paying users

This conversion rate may be the lowest, but the value obtained is the highest. A good word-of-mouth marketing may be worth 10 yuan, but a good paying user may bring 25 yuan or even more value. The effect would be even more obvious if the conversion rates in terms of word-of-mouth and revenue could be improved. An increase from 1% to 2% is a doubling of the increase.

Conversion rate optimization is the conversion rate from each link to the next link in the AARRR graph. In other words, if there are 100 new users and 10 of them are activated, there will be a conversion rate of 10%. If the conversion rate of 10% can be increased to 20%, there will be more than double the number of new activated users, 20. This is the concept of conversion rate optimization.

If optimization can be achieved in every conversion link, the final revenue growth will be greatly improved.

Conversion rate optimization is the most effective and cost-effective way to increase sales. Increasing the conversion rate from new user acquisition to activation from 10% to 20% will bring in twice as many new activated users. If the conversion rate in the next few links remains unchanged, the revenue will double, which is a huge growth. Without conversion rate optimization, growth is difficult.

In essence, we need to continuously improve our products, marketing strategies and operational strategies.

For enterprises, it is necessary to maximize and optimize business goals. If you are responsible for attracting new customers, make your conversion rate from new customers to activation as high as possible. If you are responsible for monetization, improve the conversion rate from retention to revenue.

For customers, optimizing conversion rates also improves user experience. 99% of users will no longer use a product due to poor user experience.

What benefits will optimizing conversion rates bring you?

1. Expand customer acquisition channels

Neither online advertising nor offline promotion has produced good conversions. Maybe you spent 100,000 yuan and attracted 10,000 users, but 90% of them left when they registered, and only 1,000 people stayed in the end. This channel can only help you get 1,000 users. But if you optimize the conversion rate and reduce the churn rate from 90% to 50%, this channel can bring you 5,000 users instead of 1,000 users, so you can actively expand your customer acquisition channels.

2. Make all market activities more efficient.

By optimizing the conversion rate, new users are more likely to become loyal users or even paying users. In this case, you can invest more actively in the market.

It turned out that 10,000 yuan was enough to get 1,000 users, which ultimately brought in 100 yuan in revenue. Now ten thousand yuan can bring in 5,000 users and generate 15,000 yuan in revenue. At this time, you should be able to actively expand the market, and your market behavior will be more effective.

3. Bring new effective growth channels

Loyal users can also help you promote and convert them into word-of-mouth marketing users. As word of mouth spreads, other users will come through friends’ recommendations, which in itself is a new growth channel. Although the number of users brought in by other aspects is the same, revenue has increased due to the improvement in conversion rate.

4. Improved user experience and user happiness

The service has improved. After activation, users find that the APP is very useful to them and they will stay here longer. Why would he convert into a word-of-mouth marketing user? Because the service exceeded his expectations, he felt that he should introduce it to his friends. He knew that this product could solve certain problems and provide certain help, and that it could do it better than others. This is the greatest and most essential value of conversion rate optimization.

5. Keep looking for room for improvement

Use conversion rate as a measurement standard and always look for room for improvement to optimize the conversion rate. Let’s use a chart and a table to look at the conversion rate. From the outermost layer, the conversion at each level is gradually getting smaller, and the conversion value is constantly increasing. Each level of conversion has a higher value, and each level of increase in conversion rate will bring actual business growth.

When solving an optimization equation, the goal is simple: conversion rate. Maximizing conversion rate, the goal is very clear. There may be limitations on budget, human resources and material resources, limitations on the number of users themselves, influence of strategies, etc.

The Conversion Rate Optimization Equation

With the restrictions, what are the variables in the solution of this equation?

Product Level

From the color scheme of the product UI, product functions, and user experience process to the details of all aspects of the copywriting, how to stimulate user usage, and the layout of the UI to the technical level of recommendation algorithms, exclusion algorithms, and back-end system architecture, all of these will affect the conversion rate.

Operational level

What kind of activities are done outside of the product, online and offline, what kind of subsidies and promotions are given to users, copywriting, social sharing, dissemination, various videos, these are all variables. We can continually adjust these variables to maximize conversions within constraints; this is the optimization equation.

Suppose we had a very powerful computer, we could input the variables, constraints and goals of the equation and calculate the best method, color scheme, text and algorithm. This does not exist. Because the solution space of this optimization equation is infinite, in mathematics, it is impossible to find the optimal point in an infinite solution level.

What should I do?

We cannot just think about having a very powerful brain and then calculate what kind of UI, product, and operation can bring the best conversion rate, but we must constantly make improvements at the product and operation levels, and then continuously optimize the conversion rate until we get infinitely close to the optimal solution.

Optimize products through MVP

The initial products, operations and services may be provided to users in the simplest way. When users and consumers give us feedback, we continue to improve through technology, creativity, analysis, ideas and A/B testing.

Just as shown in the picture above, it does not mean that you start by building a wheel, then a chassis, and a car model, and then a car comes out. This is not MVP iteration.

Before the advent of cars, there were no available products at all. This is the traditional product. The concept of MVP in traditional industries is vague or they do not have the benefit to enjoy this iterative method. But if you are an Internet agent or an online product, and the new generation of products still use this method, you are likely to be eliminated by the market.

The second way is also wrong. First build a scooter, then a bicycle, then provide users with motorcycles, and finally a car . This is of course a method of continuous iteration for customers, but the cost is too high and without accumulation, a scooter cannot be transformed into a bicycle through lightweight iterations and improvements, and a bicycle cannot be easily transformed into a motorcycle or a car.

Therefore, the correct MVP iteration is to first give him a shabby little truck, then renovate it, improve it, expand it according to user needs, and finally decorate it luxuriously. This is the MVP. First, give users a usable small broken car, and then continuously iterate to add cargo space, passenger compartment, and then decorate it. This is a new generation of optimization and iteration method.

What are the optimization directions?

For example, Google will look for areas where optimization may be needed, including product UI. A change in UI design may affect users’ understanding of the product and improve conversion rates. The content of the copywriting has changed from "apply now" to "apply immediately". Don't look at this one word, it may bring very different conversion rates. For example, if you change the “add to cart” function to “click to receive a prize”, you will find that the click-through rate will increase. Improving user operation efficiency by improving page layout is also a classic optimization method.

In terms of product functions, we would like to add some new functions to see if we can provide a better user experience, increase retention, and increase conversions. It will also include some backend technical aspects, such as whether the architecture can be improved to enhance user efficiency and how the recommendation algorithm can be improved. All of these can be adjusted and optimized.

Use the comparative experimental method mentioned earlier to see if there is any improvement after the change. If there is any improvement, publish it.

Registration rate , retention rate , sharing rate, user activity, conversion, and user behavior are all directions for optimization.

A/B testing is a single-blind experiment in which only one condition is changed. Why do A/B testing?

What are the benefits of A/B testing?

This is Airbnb's data. You can see that the data has been growing. It can be seen that there is a red time period in the middle, which lasted for about a month. What happened during this period? Airbnb launched a new product feature, but took it offline by the end of the day.

During this red time period, the data has been growing. A big problem is that the external environment has such a huge impact on the data that we have no idea from this picture whether this product feature that has been online for a month has had an impact on user growth, and what kind of growth it has.

Therefore, if there is no A/B testing and scientific verification, some unscientific decisions may be made in products, operations and marketing strategies. Maybe there was a decision that had a negative impact of -20%, but we don't know because we haven't done the experiment to verify it. Or the backend technicians may easily develop a recommendation algorithm that increases the order volume by 5%. But we don't know that because they bring in 5% of the orders. In this way, we lose this valuable experience.

From this picture, we can see that only A/B testing can tell us what impact a product function will have before and after it is launched. If every product iteration and strategy launch is optimized through A/B testing, then the growth curve may be the same as this blue line. Maybe not our favorite exponential or linear growth, but it’s always growth.

Without A/B testing, the data may fluctuate continuously like the red line. In the short term, the red line and the blue line may be similar, but over time, the red line will be far behind the blue line. This is how A/B testing drives continuous growth, and it is also Google's best practice.

How does A/B testing work?

We need to consider the choice of random traffic. How to support multiple variables, single variable, multiple experiments, and single experiment at the same time? How can we choose the experimental version more conveniently and know how long it will take to get reliable results? Without good statistical tools , the execution interval will not be calculated correctly and the experimental results will be inaccurate.

At this time, you can consider using a third-party A/B testing tool. In the product or marketing industry, if you want to access a third-party A/B testing SDK, you can focus on designing iterative plans. These solutions can be released to users or consumers through third-party platforms, and then the experimental traffic can be adjusted freely, allowing 1% of users to try this solution and 5% of users to try that solution. Based on data feedback, the pros and cons of the solutions can be analyzed. This is the method used in A/B testing practice.

What does correct, efficient A/B testing look like?

A product manager can do dozens of A/B tests a day, and hundreds in seven days. Sixty to seventy out of a hundred experiments neither increase nor decrease the conversion rate. There may be a dozen or twenty things that you think are good, but are actually bad, and you can also learn from the lessons. Only a few of them are good enough that you can release them to users and ultimately bring in 20% of the revenue. If you have an experiment that improves your results by 20% every week, or if several experiments add up to 20%, you will be twice as far ahead of your competitors after a month and a half.

Through a third-party A/B testing platform, you can easily conduct experiments, including implementing a simple revision, publishing with one click to do a lot of QA testing, grabbing traffic, rolling back with one click, and immediately closing the experiment if a bug is found. Obtain accurate experimental results in real time, accelerate the convergence of the execution interval, support a large number of concurrent experiments, and more.

Try not to build this system by yourself. Many companies that have successfully built one have spent at least two years of manpower and material resources, tens of millions of dollars in expenses, and may also encounter various bugs.

How do Airbnb and Google do A/B testing?

Airbnb

All changes to Airbnb's important pages and process optimizations will be released to 1%-5% of users first. Take a look at the actual data to see if there is any improvement in visit time, retention, and orders, and then decide whether to fully release it or cancel it.

The new generation of companies have this awareness and ability from day one. They will tell their colleagues why they should do A/B testing and why this method can achieve sustainable growth.

Google

I did a lot of A/B testing experiments at Google to help grow Google's advertising revenue. I found some very interesting phenomena. For example, if the ad space is moved one pixel to the left, it will bring revenue growth. If it is moved two items to the left and then one pixel, there will be a loss.

There is absolutely no theoretical explanation for this. If a product manager told me that moving one pixel to the left can bring growth, I would definitely not believe it. However, A/B testing will tell me that this is the fact. Therefore, any product changes and optimizations require A/B testing before they can be launched.

8 Practical Experiences for Efficient AB Testing

1. Amazing results

Small changes lead to huge marketing KPIs. For example, pixels, colors, and copy can have a huge impact.

2. Patience test

But most of the changes will not be drastic improvements, which is easy to understand. If the user doesn't care about this at all, changing it again and again won't have any effect.

3. Twyman’s Law

Any chart that looks unexpected is usually because the data is statistically incorrect. If the experimental results are very good, it is very likely that the data statistics are wrong.

4. You are different

Copying other people's experience often doesn't produce any results.

5. Speed ​​is critical

Any change that can speed up user response time will definitely lead to a huge improvement in KPI. So if a technician, product, or marketing agency says they can speed up your H5 loading time and increase user response time, you should support them in doing so by all means.

6. Focus on product quality

It is easy to increase click-through rate, but what is important is to improve the user's real experience.

For example, sometimes e-commerce companies emphasize price display, which will reduce the click-through rate of adding items to the shopping cart. An e-commerce company enhanced the way product prices were displayed to allow users to see the prices more clearly. However, it was found that the click-through rate of adding items to the shopping cart or the product browsing rate dropped by more than half, which seemed to be a very bad improvement. But this is actually a very successful improvement, which improves product quality and increases user purchase rate and GME a lot. Why? It is easier for users to find the products they want to buy, which improves the experience and increases sales.

7. Fast and lightweight iteration

Try not to do large, complex experiments that involve a lot of changes. This makes it easier to trace the cause, such as what change produced what effect, rather than changing 10 places to produce one effect. Will these 10 changes have a positive effect on me? uncertain.

8. The number of users is the base number

It takes thousands or tens of thousands of users to carry out efficient predictions.

Ink Weather

They conducted an interesting experiment, hoping to increase the sharing rate of weather news. The share button has been improved and three different improvement methods have been found, each of which has a change compared to the original method. There is a method that may appear to have improved by 5%, but the execution range is -1% to +10.4%. What does this mean? This means that the sharing rate may have dropped by 1%. However, the third option is particularly ideal. It can increase the sharing rate by an average of 18%, with an execution range of 11.9%-23.6%. This means that if the sharing button is released online, the sharing rate can be increased by at least 11.9%.

Didi Taxi

When Didi recruits drivers, it will think of various optimization plans to improve the driver conversion rate. The middle plan allows you to complete one order per day and earn gas money easily, which increases the registration rate by 18%. The last program, car owner recruitment, increased the registration rate by 21%. The results of such an A/B test on the platform not only produced results, increasing the driver registration rate, but also helped product managers and marketers.

The original marketing strategy was to show a handsome driver and his family. Such a strategy may not necessarily attract drivers. Why? Maybe many people who want to take a taxi clicked in thinking it was a taxi advertisement. This method can attract some part-time drivers to register.

For the middle solution, people who click in will find it troublesome to upload their vehicle registration, driver’s license, bank account number, etc., and the conversion rate is not ideal.

But recruiting car owners is different. It is very clear that you want to be a Didi driver. At this time, the registration rate will increase, which will be of great help to future market strategies.

Amazon

Amazon's shopping cart button remains unchanged no matter how the page changes. When you click on it, a green prompt pattern will be displayed. Why does Amazon use such a design? Because it has conducted a large number of A/B tests and found that only this kind of text and UI can achieve the highest conversion rate.

Amazon has also conducted many interesting A/B tests, including one to increase the conversion rate of credit card applications. At the beginning, Amazon credit card was available on the product display page. If you added a piece of clothing to the shopping cart, you could also add an Amazon credit card to the shopping cart, but such credit card promotion was basically ignored. Finally, the operation team came up with a very interesting idea: could they promote credit cards when users were checking out? First, it saves advertising space and does not need to be displayed on the product page. Second, it generates a very high conversion rate. Finally, this helps Amazon earn $100 million every year.

These experiences that everyone knows are all produced through A/B testing and continuous optimization and iteration.

Mobile application product promotion service: APP promotion service Qinggua Media information flow

The author of this article is @Mantou Business School Wang Ye Compiled and published by (APP Top Promotion), please indicate the author information and source when reprinting!